SPE 71538

Oil-Water Separation in Liquid-Liquid Hydrocyclones (LLHC) Experiment and Modeling

Carlos Gomez, Juan Caldentey, Shoubo Wang, Luis Gomez, Ram Mohan and Ovadia Shoham, SPE, The University of
Tulsa

Copyright 2001, Society of Petroleum Engineers Inc.

This paper was prepared for presentation at the 2001 SPE Annual Technical Conference and
Exhibition held in New Orleans, Louisiana, 30 September 3 October 2001.
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, as
presented, have not been reviewed by the Society of Petroleum Engineers and are subject to
correction by the author(s). The material, as presented, does not necessarily reflect any
position of the Society of Petroleum Engineers, its officers, or members. Papers presented at
SPE meetings are subject to publication review by Editorial Committees of the Society of
Petroleum Engineers. Electr onic reproduction, distribution, or storage of any part of this paper
for commercial purposes 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 where and by whom the paper was presented. Write Librarian, SPE, P.O.
Box 833836, Richardson, TX 75083-3836, U.S.A., fax 01-972-952-9435.

Abstract
The liquid-liquid Hydrocyclone (LLHC) has been widely used
by the Petroleum Industry for the past several decades. A
large quantity of information on the LLHC available in the
literature includes experimental data, computational fluid
dynamic simulations and field applications. The design of
LLHCs has been based in the past mainly on empirical
experience.
However, no simple and overall design
mechanistic model has been developed to date for the LLHC.
The objective of this study is to develop a mechanistic model
for the de-oiling LLHCs, and test it against available and new
experimental data. This model will enable the prediction of
the hydrodynamic flow behavior in the LLHC, providing a
design tool for LLHC field applications.
A simple mechanistic model is developed for the LLHC.
The required input for the model is: LLHC geometry, fluid
properties, inlet droplet size distribution and operational
conditions. The model is capable of predicting the LLHC
hydrodynamic flow field, namely, the axial, tangential and
radial velocity distributions of the continuous-phase. The
separation efficiency and migration probability are determined
based on swirl intensity prediction and droplet trajectory
analysis. The flow capacity, namely, the inlet-to-underflow
pressure drop is predicted utilizing an energy balance analysis.
An extensive experimental program has been conducted
during this study, utilizing a 2 MQ Hydroswirl hydrocyclone.
The inlet flow conditions are: total flow rates between 27 to
18 gpm, oil-cut up to 10%, median droplet size distributions
from 50 to 500 m, and inlet pressures between 60 to 90 psia.
The acquired data include the flow rate, oil-cut and droplet
size distribution in the inlet and in the underflow, the reject
flow rate and oil concentration in the overflow and the

separation efficiency. Additional data for velocity profiles

were taken from the literature, especially from the Colman and
Thew (1980) study. Excellent agreement is observed between
the model prediction and the experimental data with respect to
both separation efficiency (average absolute relative error of
3%) and pressure drop (average absolute relative error of
1.6%).
Introduction
The petroleum industry has traditionally relied on
conventional gravity based vessels, that are bulky, heavy and
expensive, to separate multiphase flow. The growth of the
offshore oil industry, where platform costs to accommodate
these separation facilities are critical, has provided the
incentive for the development of compact separation
technology. Hydrocyclones have emerged as an economical
and effective alternative for produced water deoiling and other
applications. The hydrocyclone is inexpensive, simple in
design with no moving parts, easy to install and operate, and
has low maintenance cost.
Hydrocyclones have been used in the past to separate
solid-liquid, gas-liquid and liquid -liquid mixtures. For the
liquid -liquid case, both dewatering and deoiling have been
used in the oil industry. This study focuses only on the latter
case, namely, using the liquid-liquid hydrocyclones (LLHC) to
remove dispersed oil from a water continuous stream.
Oil is produced with significant amount of water and gas.
Typically, a set of conventional gravity based vessels are used
to separate most of the multiphase mixture. The small amount
of oil remaining in the water stream, after the primary
separation, has to be reduced to a legally allowable minimum
level for offshore disposal. LLHCs have been used
successfully to achieve this environmental regulation.
There is a large quantity of literature available on the
LLHC, including experimental data sets and computational
fluid dynamic simulations. However, there is still a need for
more comprehensive data sets, including measurements of the
underflow droplet size distribution. Additionally, there is a
need for a simple and overall mechanistic model for the
LLHC.
The objective of the present study is two fold: to develop
a mechanistic model for the LLHC that can predict the flow
behavior in the hydrocyclone and the oil/water separation
efficiency; and to acquire new experimental data for the

C. Gomez, J. Caldentey, S. Wang, L. Gomez, R. Mohan, O. Shoham

LLHC, including detailed measurements of the droplet size

distributions in the inlet and underflow streams.
The
developed mechanistic model can be utilized for the design of
LLHCs, providing the flexibility of designing alternative
LLHC geometries for the same operating conditions for
optimization purposes. It will also allow detailed analysis and
performance prediction for a given LLHC geometry and
operating conditions, including separation efficiency and flow
capacity (pressure drop flow rate relationship).
LLHC Hydrodynamic Flow Behavior. The LLHC, shown in
Figure 1, utilizes the centrifugal force to separate the dispersed
phase from the continuous fluid. The swirling motion is
produced by the tangential injection of pressurized fluid into
the hydrocyclone body. The flow pattern consists of a spiral
within another spiral moving in the same circular direction
(Seyda and Petty, 1991). There is a forced vortex in the region
close to the LLHC axis and a free-like vortex in the outer
region. The outer vortex moves downward to the underflow
outlet, while the inner vortex flows in a reverse direction to
the overflow outlet. Moreover, there are some re-circulation
zones associated with the high swirl intensity at the inlet
region. These zones, with a long residence time and very low
axial velocity, have been found to be diminished as the flow
enters the low angle taper section (see Figure 1).
An explanation of the characteristic reverse flow in the
LLHC is presented by Hargreaves (1990). With high swirl at
the inlet region, the pressure is high near the wall region and
very low toward the centerline, in the core region. As a result
of the pressure gradient profile across the diameter, which
decreases with downstream position, the pressure at the
downstream end of the core is greater than at the upstream,
causing flow reversal.
As the fluid moves to the underflow outlet, the narrowing
cyclone cross-sectional area increases the fluid angular
velocity and the centrifugal force. It is due to this force and the
difference in density between the oil and the water, that the oil
moves to the center, where it is caught by the reverse flow and
separated, flowing into the overflow outlet. Instead, if the
dis persed phase is the heavier, like solid particles, it will
migrate to the wall and exit through the underflow.
The amount of fluid going through the different outlets
differs with heavy and light dispersion. It means that for these
two different separation cases, two different geometries are
needed (Seyda and Petty, 1991). In the deoiling case, usually
between 1 to 10 percent of the feed flow rate goes to the
overflow.
Another phenomenon that may occur in a hydrocyclone is
the formation of a gas core. As Thew (1986) explained,
dissolved gas may come out of solution because of the
pressure reduction in the core region, migrating fast to the
LLHC axis, and eventually emerging through the overflow
outlet. A significant amount of gas can be tolerated but
excessive amounts will disturb the vortex. An experimental
study on this topic is found in Smyth and Thew (1996).

SPE71538

LLHC Geometry. The deoiling LLHC consists of a set of

cylindrical and conical sections. Colman and Thews (1988)
design has four sections, as shown in Figure 2. The inlet
chamber and the reducing section are designed to achieve
higher tangential acceleration of the fluid, reducing the
pressure drop and the shear stress to an acceptable level. The
latter has to be minimized to avoid droplet breakup leading to
reduction in separation efficiency. The tapered section is
where most of the separation is achieved. The low angle of
this segment keeps the swirl intensity with high residence
time. An integrated part of the design is a long tail pipe
cylindrical section in which the smallest droplets migrate to
the reversed flow core at the axis and are being separated
flowing into the overflow exit. This configuration gives a very
stable small diameter reversed flow core, utilizing a very small
overflow port.
Young et al. (1990) achieved similar results to ColmanThews LLHC, in terms of separation efficiency, with a
different hydrocyclone configuration. Three sections were
used instead of four. The reducing section was eliminated and
the angle of the tapered section was changed from 1.5 to 6.
Later, Young et al. (1993) developed a new LLHC design,
which resulted in an improvement in the separation
performance. The principal modification of the enhanced
design was a small change in the tail pipe section. A minute
angle conical section was used rather than the cylindrical pipe.
Another important parameter in the LLHC geometry is the
inlet configuration, as shown in Figure 3. Rectangular and
circular, single and twin inlets have been most frequently used
by different researchers. The main goal is to inject the fluid
with higher tangential velocity, avoiding the rupture of the
droplets. The twin inlets have been considered to maintain
better symmetry and for this reason maintain a more stable
reverse core (Colman et al., 1980; Thew et al., 1984). Good
results have also been achieved with the involute single inlet
design.
The last element of the LLHC is the overflow outlet. This
is a very small diameter orifice that plays a major role in the
split ratio, defined as the relationship between the overflow
rate and the inlet flow rate. Most of the commercial LLHCs
permit changing the diameter of this orifice, depending on the
range of operating conditions.
Literature Review
There are hundreds of literature references on the LLHC,
including experimental studies, CFD simulations and
modeling. Detailed review of these previous studies can be
found in Caldentey (2000) and Gomez (2001). In this section
only pertinent mechastudies are reviewed briefly
Two textbooks that condense pioneering works on
hydrocylones
and
fundamental
theories,
including
experimental data, design, and performance aspects, are
Bradley (1965) and Svarovsky (1984). Both refer in most of
the chapters to solid-liquid hydrocyclones, with only a small
section available on liquid-liquid separation and other
application areas.

SPE71538

OIL-WATER SEPARATION IN LIQUID-LIQUID HYDROCYCLONES (LLHC)- EXPERIMENT AND MODENLING

Experimental Studies. Only a representative sample of

previous experimental studies is summarized here. The review
is divided into laboratory studies and design and application
studies, as follows.
Laboratory Studies. Earlier studies were presented by Simkin
and Olney (1956), Sheng (1974), Johnson et al. (1976), Smyth
et al. (1980), Colman et al. (1980) and Colman and Thew
(1980).
A general revision of the hydrocyclone developed at
Southampton University was carried out by Thew (1986), who
also discussed some issues presented previously by Moir
(1985). Other studies were published by Gay (1987),
Bednarski and Listewnik (1988), Woillez and Schummer
(1989) and Beeby and Nicol (1993).
Young et al. (1990) measured the flow behavior in a
Colman and Thew (1980) type hydrocyclone, and later
proposed a new modified design. In 1991, Weispfennig and
Petty explored the flow structure in a LLHC using a
visualization technique (laser induced fluorescence). The
performance of a mini hydrocyclones, of 10 mm-diameter,
were studied by Ali et al. (1994) and Syed et al. (1994).
Design and Applications. A summary of the selection, sizing,
installation and operation of hydrocyclones was provided by
Moir (1985). Meldrum (1988) described the basic design and
principle of operation of the de-oiling hydrocyclone.
Choi (1990) tested a system of six hydrocyclones (35 mm
diameter) operating in parallel for produced water treatment
(PWT). The performance of three commercial liquid-liquid
hydrocyclones (two static and one dynamic) in an oil field was
evaluated by Jones (1993).
CFD Simulations. Numerical simulations or CFD are used
widely to investigate flow hydrodynamics. As expressed by
Hubred et al. (2000), the solution of the Navier Stokes
Equations for simple or complex geometry for non-turbulent
flow is feasible nowadays. But current computational
resources are unable to attain the instantaneous velocity and
pressure fields at large Reynolds numbers even for simple
geometries. The reason is that traditional turbulence models,
such as k-?, are not suitable for this complex flow behavior.
On the other hand, more realistic and complicated turbulence
models increase the computational times to inconvenient
limits.
The flow in hydrocyclones has been numerically
simulated by Rhodes et al. (1987), Hsieh and Rajamani (1991)
(see also Rajamani and Hsieh, 1988; Rajamani and
Devulapalli, 1994) and He et al. (1997). In most of these
studies the models were evaluated through comparison with
laser-doppler anemometry (LDA) data. Many researchers
have used this technique to measure the velocity field and
turbulence intensities (Dabir, 1983; Fanglu and Wenzhen,
1987; Jirun et al., 1990; and Fraser and Abdullah, 1995).
Modeling. Although widely used nowadays, the selection and
design of hydrocyclones are still empirical and experience
based. Even though quite a few hydrocyclone models are
available, the validity of these models for practical
applications has still not been established (Kraipech et al.,
2000). A thorough review of the different available models

can be found in Chakraborti and Miller (1992) and Kraipech et

al. (2000).
The LLHC models can be divided into empirical and
semi -empirical, analytical solutions and numerical simulations
(Chakraborti and Miller, 1992). The empirical approach is
based on development of correlations for the process key
parameters, considering the LLHC as a black box. The semiempirical approach is focused on the prediction of the velocity
field, based on experimental data. The analytical and
numerical solutions solve the non-linear Navier-Stokes
Equation. The former one is a mathematical solution, which is
achieved neglecting some of the terms of the momentum
balance equation. The numerical solution uses the power of
computational fluid dynamics to develop a numerical
simulation of the flow. As Svarovsky (1996) comments, it
seems that the analytical flow models have been abandoned in
favor of numerical simulations due to the complexity of the
multiphase flow phenomena.
From extensive experimental tests, Colman and Thew
(1983) developed some correlations to predict the migration
probability curve, which defines the separation efficiency for a
particular droplet size in a similar way that the grade
efficiency does for solid particles. Seyda and Petty (1991)
evaluated the separation potential of the cylindrical tail pipe
section. A semi-empirical model to predict the velocity field in
a cylindrical chamber was developed to calculate the particle
trajectories, and hence, the grade efficiency.
Wolbert et al. (1995) presented a computational model to
determine the separation efficiency based on the analysis of
the trajectories of the oil droplets. An extension of Bloor and
Ingham (1973) LLHC model was presented by Moraes et al.
(1996). The modification takes into account the difference in
the split ratio for liquid-liquid and solid-liquid hydrocyclones.
The literature review confirms the need for accurate
experimental data utilizing appropriate sampling procedure
and including the measurements of the droplet size
distributions at the inlet and underflow sections, and the need
to develop a simple mechanistic model for the LLHC. These
deficiencies are addressed in the present study.
Experimental Program
This section describes the experimental facility, working
fluids, definitions of pertinent separation parameters, and the
experimental results of the LLHC.
Experimental Facility. The experimental three-phase, oilwater-gas, flow loop is shown Figure 4. The oil-water-gas
indoor flow facility is a fully instrumented state-of-the-art
two-inch flow loop, enabling testing of single separation
equipment or combined separation systems. The test loop
consists of four main components: storage and metering
section, LLHC test section, downstream oil-water separation
facility, and data acquisition system. Following is a brief
description of these sections.
Storage and Metering Section. Oil and water are stored in two
tanks of 400 gallons capacity each. Each tank is connected to
two pumps. The first one is a 3656 model pump, 1x2-8 size,

C. Gomez, J. Caldentey, S. Wang, L. Gomez, R. Mohan, O. Shoham

cast iron construction with bronze impeller, John Crane Type

21 mechanical seal, 10 HP motor and operates at 3600 rpm. It
delivers 25 gpm @ 108 psig. The second one is a 3656 Model
pump, 1.5x2-10 size, cast iron construction with bronze
impeller, John Crane Type 21 mechanical seal, 25 HP motor
and operates at 3600 rpm. It delivers 110 gpm @ 150 psig.
Both pumps are equipped with return lines. Each fluid is
pumped from the storage tank to the metering section. The
metering section comprises of pressure gages, control valves,
variable speed controllers and state-of-the-art Micromotion
net oil computers (NOC), which provide the total mass flow
rate, water-cut, temperature, and mixture density. The signals
from the flow meters and control valves are fed to the data
acquisition system, which will be described later. Check
valves to prevent any back-flow are installed downstream of
the control valves. The metered oil and water are then
combined in a mixing-tee to obtain oil-water dispersion.
Additionally, a static mixer is available in parallel to the
mixing-tee for homogenization of the flow.
Test Section. Figure 5 shows a schematic of the LLHC test
section and Figure 6 presents a photograph of the LLHC
prototype installed in the test section. The LLHC is a 2-inches
NATCO MQ Hydro Swirl Hydrocyclone mounted vertically
with a total height of 62 inches. Water flows into the test
section through a 2 pipe coming from the water tank. This
pipe has a split section where the water split stream mixes
with oil in order to get thorough mixing with the desired oil
concentration. The split section is a inch pipe composed of
a water wheel paddle meter, a mixing tee and a static mixer.
Oil for the mixture is pumped from a 55 gallons barrel with a
gear pump, and metered by means of a gear flow meter. Once
the oil and water are mixed, they pass through a static mixer in
order to get a desired droplet size distribution. After this point
the mixture is directed to the main stream pipe entering it by
means of an inverse pitot tube. Once the mixture enters the
main stream line, it can either flow directly to the test section
or be subjected to an additional mixing loop where smaller
droplet size distributions can be achieved. The mixture can be
sent to either the MQ steel hydrocyclone or the MQ acrylic
hydrocyclone.
The latter LLHC, which has the same
characteristics as the steel one, is placed for observation
purposes.
In order to measure the droplet size dis tribution, a special
isokinetic sampler is designed and operated in order to get
representative accurate measurements of the distributions, as
shown in Figure 7. Samples from both the inlet and underflow
streams can be obtained. Once the sample is taken, it is placed
in the droplet size distribution analyzer. For this purpose, a
Laser scattering device, namely, the Horiba LA -300 analyzer
is used to analyze the samples. . It may be noted that a
surfactant-based additive is utilized, as shown in Figure 7, to
avoid coalescence in the sample when transferred and run in
the droplet size analyzer
The flow in the LLHC is split into two streams: The
overflow stream, with mainly oil, and the down-flow stream,
with mainly water. The overflow is discharged into a 55
gallons barrel and the underflow is sent to the downstream

SPE71538

three-phase separator. Pressure transducers are located on the

upper and the lower outlets of the LLHC. The underflow
stream passes through a metering section, located upstream of
the three-phase separator, where flow rate, density,
temperature and water cut are measured using a liquid
Micromotion coriolis mass flow meter. Due to the small oil
concentration in some of the experiments, a special oil content
analyzer is utilized to measure the oil concentration of the
underflow. This equipment is a Horiba OCMA 220 model
that uses infrared spectroscopy technique.
Downstream Oil-Water Separation Section. The 528 gallon
three-phase flow separator located downstream of the LLHC
test section operates at 10 psig. It consists of three
compartments. In the first compartment the oil-water mixture
is stratified and the oil flows into the second compartment
through flotation. In this compartment, there is a level control
system that activates a control valve discharging the oil into
the oil storage tank. The water flows from the first
compartment to the third compartment through a channel
located below the second compartment. In this compartment,
there is also a level control system, allowing water to flow into
the water storage tank.
Data Acquisition System. IDM variable speed controllers
installed on all the 4 pumps control the oil and water flow
rates into the test section. The flow loop is also equipped with
several temperature sensors and pressure transducers for
measurement of the in-situ temperature and pressure
conditions.
All output signals from the sensors, transducers, and
metering devices are collected at a central panel. A state-ofthe-art data acquisition system, built using LabView, is used
to both control the flow in the loop and also to acquire data
from analog signals transmitted from the instrumentation. The
program provides variable sampling rates. The sampling rate
was set at 2 Hz for a 2 minutes sampling period. The final
measured quantity results from an arithmetic averaging of 120
readings, after steady-state condition is established.
A regular calibration procedure, employing a highprecision pressure pump, is performed on each pressure
transducer at a regular schedule, to guarantee the precision of
measurements. The temperature transducer consists of a
Resistance Temperature Detector (RTD) sensor and an
electronic transmitter module.
Working Fluids. Tap water and mineral oil were chosen and a
dye (red) was added to the oil to improve flow visualization
between the phases. The oil has low emulsification, fast
separation, appropriate optical characteristics, non-degrading
properties, and is non-hazardous. The properties of the oil are
0
API=33.7 and O = 13.6 cP at 1000 F. During all the
experimental runs the average temperature in the flow loop
varied between 700 and 800 F.
Definition of Separation Parameters. Following are the
definitions of two important parameters used in this study to
define the total separation efficiency:

SPE71538

OIL-WATER SEPARATION IN LIQUID-LIQUID HYDROCYCLONES (LLHC)- EXPERIMENT AND MODENLING

Split Ratio: The split ratio is the ratio of the overflow rate
to the inlet flow rate, as given below:
q
overflow
F=
100%
(1)
q
inlet
where F is the split ratio, qoverflow is the total flow rate at the
upper outlet of the LLHC , and q inlet is the total inlet flow rate.
Oil Separation Efficiency: Practical interpretation of
separation data is concerned with the purity of individual
discharge streams. Many references quantify the relative phase
composition of the separated streams in the form of a
percentage by volume measurement. In this study a widely
used definition is adapted for the oil separation efficiency,
namely,

ff

q oil overflow
q oil inlet

100 %

(2)

where qoil-overflow is the flow rate of oil at the overflow, qoil-inlet

is the flow rate of oil at the inlet. Utilizing continuity
relationship, Equation (2) becomes

ff

= (1

c
underflow oil underflow
) 100% (3)
qinlet coil inlet

Note that when coil-underflow tends to zero, the separation

efficiency is maximum.
Experimental Results. A total of 124 runs were conducted in
this study. The data is analyzed and presented, so as to
demonstrate the effect of the flow variables on the separation
efficiency, as given in the following sections.
Effect of Pressure Drop and Flow Rate. The separation of oil
droplets in the swirl chamber of the hydrocyclone is a result of
the forces imposed on the oil droplets in the spinning fluid and
the residence time in the chamber. Lower flow rates mean
longer residence times but lower acceleration forces.
Conversely higher flow rates result in higher acceleration
forces and smaller residence times. As shown in Figure 8, the
MQ Hydroswirl performance is independent of flow rate in the
range tested. For hydrocyclones of similar geometries, the
literature reports similar results.
Effect of Underflow Pressure. Back pressure must be applied
at the hydrocyclone underflow, in order to force the core
stream containing the oil to the overflow; otherwise, all the
flow will exit through the underflow and no separation would
occur. For a given underflow backpressure, if the overflow
pressure is slowly increased, the core diameter increases,
ultimately resulting in part of the oil core discharging out
through the underflow. The MQ Hydroswirl performance is
independent of the underflow pressure, as shown in Figure 9,
provided there is sufficient backpressure to force enough flow
out of the overflow (Young et al. 1990). It is critical that
constant back pressure be applied, since swings in
backpressure result in the oil in the core being rapidly
discharged with the cleaned water.

Effect of Overflow Diameter. Separation efficiency of

LLHCs is independent of overflow diameter (Young et al.
1990). This is confirmed by the results of this study, as shown
in Figure 10. However, the minimum overflow rate to make
an effective separation increases with increasing overflow
diameter. The minimum flow rate for each orifice opening
size is a result of a minimum velocity required for the oil to
move to the overflow (Young et al. 1990). This minimum
velocity multiplied by the cross sectional area of the overflow
results in a minimum flow rate for effective separation for
each overflow opening size. Increasing overflow size results
in an increased amount of water, which must be removed with
the oil to obtain the same removal efficiency. This of course
means that a greater flow rate of oily wastewater must be
reprocessed.
The major advantage of larger overflow
diameters is that it allows more oil to be removed without
affecting the purity of the underflow water stream when large
slugs of oil are encountered in field operations. Furthermore,
larger outlets are not as susceptible to blockage as the smaller
ones. Figure 10: Effect of overflow diameter on effic iency
Effect of Inlet Oil Concentration. Field reports indicate that
with increased oil concentrations, the performance of the MQ
Hydroswirl hydrocyclone improves and can handle the
additional oil. As can be observed in Figure 11, separation is
independent of inlet oil concentration when there is adequate
flow at the overflow to remove the required amount of oil.
The improved separation of field installations with increasing
oil content is probably due to the presence of larger oil droplet
sizes.
Effect of Oil Droplet Size Distribution. The variable having
the greatest impact on oil-water separation is the oil droplet
size distribution. Figure 12 shows the separation performance
of the MQ Hydroswirl hydrocyclone for several droplet size
distributions, with the median droplet size shown. As can be
seen, the oil separation efficiency increases with increase in
the droplet size. This can be intuitively expected as the larger
oil droplets coalesce faster than the smaller ones.
Typical results for the droplet size distributions in the
inlet and underflow streams are given in Figure 13. This
figure demonstrates the removal of the large droplets from the
feed stream. Also, the underflow stream contains smaller
droplets sizes, as compared to the inlet stream, due to breakup
of droplets in the LLHC.
Mechanistic Modeling
The following sections provide details of the mechanistic
model developed for the LLHC in the present study.
Swirl Intensity. The swirl intensity is defined as the ratio of
the local tangential momentum flux to the total momentum
flux.
The swirl intensity equation given below is a
modification of the Mantilla (1998) correlation, based on
Erdal (2001) CFD simulations, given by

C. Gomez, J. Caldentey, S. Wang, L. Gomez, R. Mohan, O. Shoham

= 0.49 Re

0.118

Mt 2

M I
T

0.35
1 M
1
EXP t I 4
2 M T
Re z

as shown in Figure 14. This type of behavior is known as a

Rankine Vortex. Algifri et al. (1988) proposed the following
equation for the tangential velocity profile:

0.93

( 1 + 1.2 tan( )0.15 )

0.16

Dc

0.7

(1 + 2 tan( ) )
0.12

(4)
where M t /M T is the ratio of the momentum flux at the inlet slot
to the axial momentum flux at the characteristic diameter
position, calculated as:

Mt
m& V is
m& / c Ais
A
=
=
= c
&
&
M T mU avc
m / c Ac
Ais

(5)

The variables in the above equations are: O is the swirl

intensity, Re is the Reynolds number, is the semi-angle of
the conical sections, Dc is the characteristic diameter of the
LLHC (measured where the angle changes from the reducing
section to the tapered section in the Colman and Thews
Design, and at the top diameter of the 3 tapered section of the
Youngs Design), z is the axial position starting from Dc, Vis
is the velocity at the inlet, Uavc is the average axial velocity at
& is the mass flow rate, Ac is the cross sectional area at
Dc, m
Dc and A is is the inlet cross sectional area.
The Reynolds number is defined in the same way as for
pipe flow with the caution that it refers to a given axial
position, yielding:

Rez =

cU avz D z
c

(6)

where c is the viscosity of the continuous fluid.

The inlet factor, I, as suggested by Erdal (2001), is
defined as:

n
I = 1 EXP
2

SPE71538

(7)

where n = 1.5 for twin inlets and n = 1 for involute single

inlet.
Velocity Field. The swirl intensity is related, by definition, to
the local axial and tangential velocities. Therefore, it is
assumed that once the swirl intensity is predicted for a specific
axial location, it can be used to predict the velocity profiles.
Both the tangential and axial velocities are calculated
following a similar procedure as proposed by Mantilla (1998).
The radial velocity, which is the smallest in magnitude, is
computed considering the continuity equation and the wall
effect.
Tangential Velocity. It has been confirmed experimentally
that the tangential velocity is a combination of a forced vortex
near the hydrocyclone axis, and a free-like vortex in the outer
wall region, neglecting the effect of the wall boundary layer,

r 2
w
T

= m 1 EXP B
U avc r
Rc

Rc

(8)

where w is the local tangential velocity, which is normalized

with the average axial velocity, Uavc, at the characteristic
diameter; Rc is the radius at the characteristic location and r is
the radial location. The term Tm represents the maximum
momentum of the tangential velocity at the section and B
determines the radial location at which the maximum
tangential velocity occurs. The following expressions were
obtained by curve-fitting several sets of the experimental data.

Tm =

(9)

Involute Single Inlet:

B = 55. 7 1.7

(10)

Twin Inlets:

B = 245. 8 2.35

(11)

It can be seen that the above equations are only functions of

the swirl intensity, O. Thus, for a given axial position, the
tangential velocity is only function of the radial location and
the swirl intensity.
Axial Velocity. In swirling flow the tangential motion gives
rise to centrifugal forces which in turn tend to move the fluid
toward the outer region (Algifri 1988). Such a radial shift of
the fluid results in a reduction of the axial velocity near the
axis, and when the swirl intensity is sufficiently high, reverse
flows can occur near the axis. This phenomenon causes a
characteristic reverse flow around the LLHC axis, which
allows the separation of the different density fluids.
A typical LLHC axial velocity profile is illustrated in
Figure 15. Here, the positive values represent downward flow
near the wall, which is the main flow direction, and the
negatives values represent upward reverse flow near the
LLHC axis. The flow reversal radius, rrev, is the radial position
where the axial velocity is equal to zero.
To predict the axial velocity profile, a third-order
polynomial equation is used with the proper boundary
conditions. The general form is as follows:

u( r ) = a1 r 3 + a 2 r 2 + a 3 r + a4

(12)

where a1 , a2 , a3 and a4 are constants. The boundary conditions

considered are:
1.

du( r = R z )
=0
dr

The velocity is maximum at the

wall;

SPE71538

OIL-WATER SEPARATION IN LIQUID-LIQUID HYDROCYCLONES (LLHC)- EXPERIMENT AND MODENLING

2. u( r = rrev ) = 0

3.

du( r = 0 )
=0
dr

Zero velocity at the location of

reverse flow, rrev;
The velocity is symmetric about the
LLHC axis; and

4. 2 u( r )rdr = U avz cR
Rz
c 0

2
z

Mass conservation.

Substituting the boundary conditions in Equation (12)

yields the axial velocity profile, which is a function of the
swirl intensity, O only:

u
U avz

During a differential time dt, the droplet moves at velocity Vr

= dr/dt in the radial direction and Vz = dz/dt in the axial
direction. Combining these two equations and solving for the
axial distance yields the governing equation for the droplet
displacement:

dz
dz
V
= dt = z
dr dr
Vr
dt

z =

Vz
dr
Vr

(17)

3
2
2 r
3 r 0.7

+
+1
C Rz C R z
C

(13)

r
3 2 rev 0. 7
R

(14)

Neglecting the axial buoyancy force (no-slip condition), the

droplet axial velocity Vz is equal to the axial velocity of the
fluid, u. This simplification is reasonable when the
acceleration due to the centrifugal force in the radial direction
is thousand times larger than the acceleration of gravity. Due
to this aspect, the LLHC is not sensitive to external
movements and it can be installed either horizontally or
vertically.

(15)

The droplet velocity in the radial direction is equal to the

fluid radial velocity, v, plus the slip velocity, Vsr. Rearranging
Equation (17) yields the total trajectory of the droplet, namely:

r
C = rev
Rz

rrev
= 0.21 0.3
Rz

Several assumptions are implicit in these equations. First,

axisymetric geometry is imposed. Then, the effects of the
boundary layer are neglected, and finally the mass
conservation balance does not consider the split ratio. The last
assumption can be considered a good approximation for small
values of split ratios used in the LLHC, usually less than 10%.
Radial Velocity. The radial velocity, v, of the continuous
phase is very small, and has been neglected in many studies.
In our case, in order to track the position of the droplets in
cylindrical and conical sections, the continuity equation and
wall conditions suggested by Kelsall (1952) and Wolbert
(1995) are used for the radial velocity profile, yielding:

v=

r
u tan( )
Rz

u
r
z = rr ==rr21
v + Vsr

The only unknown parameter in Equation (18) is the slip

velocity, which can be solved from a force balance on the
droplet in the radial direction, as shown Figure 16.
Assuming a local equilibrium momentum yields:

( c d )

2
w 2 d 3 1
2 d
= C D cV sr
r 6
2
4

Droplet Trajectories. The droplet trajectory model is

developed using a Lagrangian approach in which single
droplets are traced in a continuous liquid phase. The droplet
trajectory model utilizes the flow field presented in the
previous section. Figure 16 presents the physical model. A
droplet is shown at two different time instances, t and t + dt.
The droplet moves radially with a velocity Vr and axially with
Vz . It is assumed that in the tangential direction the droplet
velocity is the same as the continuous fluid velocity, as no
force acts on the droplet in this direction. Therefore, the
trajectory of the droplet is presented only in two dimensions,
namely r and z.

(19)

where the left side of the equation is the centripetal force, and
the right side is the drag force. Solving for the radial slip
velocity, results in:

(16)

The radial velocity is a function of the axial velocity and

geometrical parameters. In the particular case of cylindrical
sections, where tan() = 0, the radial velocity, v, is equal to 0.

(18)

4 d
V sr = c
3
c

w2 d 2

r C

(20)

where d is the droplet diameter, ?d is the density of the

dispersed phase, ?c is the density of the continuous phase and
CD is the drag coefficient calculated using the following
relationship (Morsi and Alexander, 1972 and Hargreaves,
1990):

C D = b1 +

b2
b
+ 32
Red Red

(21)

where the coefficients b are dependent on the Reynolds

Number of the droplets, defined as:

Red =

c d V sr
c

(22)

C. Gomez, J. Caldentey, S. Wang, L. Gomez, R. Mohan, O. Shoham

The values for the b coefficients, as functions of the range

of Re d , are shown in Table 1:
Finally, a numerical integration of Equation (18)
determines the axial location of the droplet as a function of the
radial position. The trajectory of a given size droplet is mainly
a function of the LLHC velocity field and the physical
properties of the dispersed and continuous phases.
Separation Efficiency. The separation efficiency of the
LLHC can be determined based on the droplet trajectory
analysis presented above. Starting from the cross sectional
area corresponding to the LLHC characteristic diameter, it is
possible to follow the trajectory of a specific droplet, and
determine if it is either able to reach the reverse flow region
and be separated, or if it reaches the LLHC underflow outlet,
dragged by the continuous fluid and carried under.
As illustrated in Figure 17, the droplet that starts its trajectory
from the wall (r = Rc) does not reach the flow reversal radius,
and thus is not separated but rather carried under. However, if
the starting location is at r < Rc, the chance of this droplet to
be separated increases. When the starting point of the droplet
trajectory is the critical radius, rcrit, the droplet reaches the
reverse radius, rrev , and is carried up by the reverse flow and is
separated.
Therefore, assuming homogeneous distribution of the
droplets, the efficiency for a droplet of a given diameter, e(d),
can be expressed by the ratio of the area within which the
droplet is separated, defined by rcrit, over the total area of flow.
This assumption has also been applied by other researchers
(Seyda and Petty, 1991; Wolbert et al., 1995 and Moraes et al.,
1996). As proposed by Moraes et al. (1996), the efficiency is
given by:

0 , if rcrit = rrev
2
2
r rrev
( d ) = crit2
, if rrev < rcrit < R c (23)
2
R c rrev
1, if r = R

crit
c

Repeating this procedure for different droplet sizes, the

migration probability curve is obtained as shown in Figure 18.
This function has an S shape and represents the separation
efficiency, e(d), vs. the droplet diameter, d. It can be seen that
small droplets have an efficiency very close to zero and as the
droplet size is increased, e(d) increases sharply until it reaches
d 100, which is the smallest droplet size with a 100% probability
to be separated.
The migration probability curve is the characteristic curve
of a particular LLHC for a given flow rate and fluid
properties. This curve is independent of the feed droplet size
distribution and is used in many cases to evaluate the
separation of a given LLHC configuration.
Using the information derived from the migration probability
curve and the feed droplet size distribution, the underflow
purity, eu , can be determined as follows:

u =

SPE71538

( di )Vi
i

(24)

i Vi

where eu is expressed in %, and Vi is the percentage

volumetric fraction of the oil droplets of diameter di . The
underflow purity is the parameter that quantifies the LLHC
capacity to separate the dispersed phase from the continuous
one.
Pressure Drop. The pressure drop from the inlet to the
underflow outlet is calculated using a modification of the
Bernoullis Equation:

1
1
Pis + c Vis2 = Pu + c U 2u + c (h cf + h f ) + cg sin L
2
2
(25)
where ?c is the density of the continuous phase; Pis and Pu are
the inlet and outlet pressures, respectively; Vis is the average
inlet velocity and Uu is the underflow average axial velocity; L
is the hydrocyclone length, ? is the angle of the LLHC axis
with the horizontal; hcf corresponds to the centrifugal force
losses and h f is the frictional losses.
The frictional losses are calculated similar to that of pipe
flow:

hf ( z ) = f ( z )

z V r2 ( z )
D( z ) 2

(26)

where f is the friction factor and Vr is the resultant velocity.

In the case of conical sections, all parameters in Equation
(26) change with the axial position, z. The conical section is
divided into m segments and assuming cylindrical geometry
in each segment, the frictional losses can be considered as the
sum of the losses in all the m segments, as follows.

h f ( conical )

V r2
?z
= f (z )
Dn 1 + Dn
n =1
2
m

( at ( 2 n 1 )

Z
2

(27)

The resultant velocity, Vr, is calculated as the vector sum of

the average axial and tangential velocities, The annular
downward flow region is only considered, as presented in the
following set of equations:

VR2 ( z ) = U Z2 + WZ2

(28)

2 R

Wz =

z
Wrdrd
0 rrev
2 R

z
rdrd
0 rrev

(29)

For simplification purposes, the average axial velocity in

Equation (28), Uz, is calculated assuming plug flow, namely,
Uz is equal to the total flow rate over the annular area from the
wall to the reverse radius, rrev. The Moody friction factor is
calculated using Halls Correlation (Hall, 1957).

SPE71538

OIL-WATER SEPARATION IN LIQUID-LIQUID HYDROCYCLONES (LLHC)- EXPERIMENT AND MODENLING

1/ 3


10 6

+
f ( z ) = 0. 00551 + 2 x10 4
(30)

D ( z ) Re( z )

where e is the pipe roughness and Re is the Reynolds Number,

calculated based on the resultant velocity computed in
Equation (28).
The centrifugal losses are the most important ones in
Equation (25), and account for most of the total pressure drop
in the LLHC. They are calculated using the following
expression:
u
hcf = rRrev

( nWu ) 2 ( r )
dr
r

(31)

where Wu is calculated from Equation (29) at the underflow

outlet and the centrifugal force correction factor, n = 2 for
twin inlets, and n = 3.2 for involute single inlet.
The centrifugal force correction factor compensates for
the use of Bernoullis Equation under a high rotational flow
condition. Its meaning is similar to the kinetic energy
coefficient used to compensate for the non-uniformity of the
velocity profile in pipe flow (Munson et al., 1994).
Rigorously, the Bernoulli equation is valid for a streamline
and the summation of the pressure, the hydrostatic and the
kinetic terms can only be considered constant in the entire
flow field if the vorticity is equal to zero.
Numerical Solution. The simulation code based on the
developed mechanistic model uses mainly two different
numerical methods to obtain the results. The tangential
velocity, given by Equation (29), is solved using the
Trapezoidal Rule, and for the droplet trajectory, a fourth-order
Runge-Kutta method is used to solve Equation (18). Also, a
commercial program (Mathematica 4.0) was used to verify the
resulting numerical values given by the computer code.
Resusults and Discussion
This section presents comparison between the LLHC
mechanistic model predictions and experimental data taken
either at the present study or from the literature. Comparisons
are made for the swirl intensity, velocity profiles, migration
probability, pressure drop, droplet size distribution and global
separation efficiency.
Swirl Intensity. The swirl intensity, which is the ratio of the
local tangential momentum flux to the total momentum flux,
can be obtained from Equation (4). Figure 19 provides the
comparison between the model predictions and the Colman
and Thew (1980), Case 2 data. Note that only 1 data point is
plotted, due to availability of axial and tangential velocity
measurements at specific axial location. The results display
the swirl intensity versus the dimensionless axial position,
where z is the axial distance from the characteristic diameter,
that is the location where the tapered section begins. Good
agreement is observed between the data point and the model

predictions. It has been experimentally proven by several

researchers that the swirl intensity decays exponentially with
axial position due to the wall frictional losses (Mantilla, 1998).
The model predictions show the same trend
Velocity Profile. The velocity field predicted by the
mechanistic model is compared with the same experimental
data set used for the swirl intensity comparison, namely, Case
2. Figure 20 presents the comparison between the
experimental data and model prediction for the tangential
velocity. The y-axis corresponds to the axis of the LLHC, and
the x-axis represents the radial position. The units used
originally were conserved, namely, millimeters per second for
the tangential velocity, and millimeters for the radial position.
The model predicts with acceptable accuracy the tangential
velocity at the wall, the peak velocity and the radius where it
occurs. The experimental data and the model display a
Rankine Vortex shape, namely, a combination of forced
vortex near the LLHC axis and a free like vortex at the outer
region.
The axial velocity profile predicted by the model is next
compared with the experimental data in Figure 21. The
positive values of axial velocities correspond to downward
flow, which is the direction of the main flow, while the
negative values represent the reverse flow. The mechanistic
model performance is excellent with respect to the axial
velocity in the downward flow region, and not so good in the
reverse flow region. Considering the calculations that the
model follows to compute the separation efficiency, the
prediction of the reverse flow velocity profile is not so
important. What is really important is the prediction of the
radius of zero velocity (rrev) since beyond this point the droplet
is assumed to be separated, moving upwards to the overflow
exit.
Migration Probability: A comparison between the model
predictions of the migration probability curve as compared
with experimental data of Colman and Thew (1980) is given
in Figure 22. Fair agreement is observed with the data.
Pressure Drop. A comparison between the predicted pressure
drop and experimental data from the present study is shown in
Figure 23, while Figure 24 shows the model predictions of
pressure drop vs. flow rate as compared with the experimental
data taken by Young et al. (1990). Very good agreement is
observed in both cases, with an average absolute relative error
of 1.6%.
Droplet Size Distribution. Figure 25 shows a comparison
between the model predictions and experimental data of the
droplet size distribution for runs 101. As shown in the figures,
good agreement is observed with experimental results. The
model prediction curves for the underflow droplet size
distribution are shifted to the right, which means that the
model predicts efficiency smaller than the experimental one.
Also there is a discontinuity in the model curve because the
model doesnt consider either breakup or coalescence. This

10

C. Gomez, J. Caldentey, S. Wang, L. Gomez, R. Mohan, O. Shoham

means that the smallest droplet that enters the LLHC is also
the smallest one that is found in the underflow stream. On the
other hand, the largest droplet in the underflow stream is the
largest droplet with a calculated efficiency below 100%.
Global Separation Efficiency. Both the underflow purity and
the migration probability curve predicted by the model are
evaluated through comparisons with experimental data. Table
2 presents a comparison with the experimental data taken at
the present study for a representative sample of the 124 runs,
and Table 3 shows a comparison with literature experimental
data, where cases 9 to 22 are part of the set of experiments
published by Colman et al. (1980). These experimental data
sets are for the LLHC configuration given in Table 4. The
characteristic diameter and operational conditions are reported
in Table 3.
As can be seen from both tables 2 and 3, the model
predictions are in excellent agreement with both data sets,
with an average absolute relative error of 3%. The results are
also plotted in Figures 26 and 27, respectively.
Summary and Conclusions
A new facility for testing LLHCs was designed, constructed
and installed in an existing three-phase flow loop. The test
section is fully instrumented to measure the important flow
and separation variables, including flow rates (inlet, underflow
and overflow) and the respective oil concentrations; droplet
size distributions (inlet and underflow streams); pressures
(inlet and underflow) and temperature. A mixer bypass loop
enables the generation of a wide range of droplet size
distributions.
A set of 124 experimental runs was conducted, with inlet
total flow rates between 18 to 26 GPM, inlet oil cuts between
0 to 10%, inlet droplet size distributions with droplet medians
between 30 to 160 microns, inlet pressures from 60 to 90 psia,
underflow pressures between 35 to 63 psia, temperature
between 65F 80F, and overflow reject diameter of 3mm
and 4mm. The collected data permitted the calculation of the
LLHC separation efficiency for each of the runs.
The collected data reveals that LLHCs can be used up to
10% inlet oil concentrations, maintaining high separation
efficiency. However, the performance of the LLHC is best for
very low oil concentrations at the inlet, below 1%. For low
concentrations, no emulsification of the mixture occurs in the
LLHC. However, high inlet concentrations, up to 10%,
promote emulsification posing a separation problem in the
overflow stream.
A simple mechanistic model is developed for the LLHC.
The model is capable of predicting the LLHC hydrodynamic
flow field, namely, the axial, tangential and radia l velocity
distributions of the continuous-phase.
The separation
efficiency and migration probability are determined based on
swirl intensity prediction and droplet trajectory analysis. The
flow capacity, namely, the inlet-to-underflow pressure drop is
predicted utilizing an energy balance analysis.
The prediction of the LLHC model was compared against
the data from both the present study and published data for

SPE71538

velocity profiles from the literature, especially from the

Colman and Thew (1980). Good agreement is obtained
between the model predictions and the experimental data with
respect to both separation efficiency (average absolute relative
error of 3%) and pressure drop (average absolute relative error
of 1.6%).
Nomenclature
A = cross sectional area
B = peak tangential velocity radius factor (Eqs. 10 and 11)
c = Concentration
CD = drag coefficient
d = droplet diameter
D = diameter
Dc = LLHC characteristic diameter
f = friction factor
F = Split ratio
g = gravity acceleration

h = losses
I = inlet factor
L = length
m = N of segments
& = mass flow rate
m
Mt = momentum flux at the inlet slot
MT = axial momentum flux at the characteristic diameter
position
n = centrifugal force correction factor, number of inlets
P = pressure
q = volumetric flow rate
r = radial position
R = LLHC radius
Re = Reynolds Number
t = time
Tm = maximum tangential velocity momentum (Eq. 9)
u = continuous phase local axial velocity
U = bulk axial velocity
v = continuous phase local radial velocity
V = volumetric fraction / velocity
Vr = droplet radial velocity
Vsr = droplet slip velocity in the radial direction
Vz = droplet axial velocity
w = continuous phase local tangential velocity
W = mean tangential velocity
z = Axial position
Greek Letters
O = swirl intensity
= taper section semi -angle
e = pipe roughness
eff = efficiency / purity
? = axis inclination angle to horizontal
= viscosity
? = density
= Horizontal plane angle
Subscripts
av = average

SPE71538

OIL-WATER SEPARATION IN LIQUID-LIQUID HYDROCYCLONES (LLHC)- EXPERIMENT AND MODENLING

c = characteristic diameter location / continuous phase

cf = centrifugal
crit = critical
d = dispersed phase / droplet
f = frictional
i = inlet
is = inlet section
o = overflow
r = resultant
rev = reverse flow
sr = Slip radial velocity
u = underflow
z = axial position
Acknowledgments
The authors thank Mr. Grant Young from Vortex Fluid
Systems Inc. and Dr. Charles Petty from Michigan State
University for their help and advise during this study.

References
1.

2.

3.

4.

5.

6.
7.

8.

9.
10.

11.

12.
13.

Algifri, A., Bhardwaj, R. and Rao, Y., 1988, Turbulence

Measurements in Decaying Swirl Flow in a Pipe. Applied
Scientific Research, Vol. 45, pp. 233-250.
Ali, S., Wesson, G., Petty, C. and Parks, M., 1994, The
Use of Small Hydrocyclones for Produced Water
Clarification. Hydrocyclone Development Consortium,
Michigan State University.
Bednarski, S., and Listewnik, J.: "Hydrocyclones for
Simultaneous Removal of Oil and Solid Particle from
Ships' Oily Waters", Filtration and Separation, March/April
1988, pp. 92-97.
Beeby, J.P., and Nicol, S.K.: "Concentration of Oil-inWater Emulsion Using the Air-Sparged Hydrocyclone",
Filtration and Separation, March/April 1993, pp. 141-146.
Bloor, M. and Ingham, D., 1973, "Theoretical Investigation
of the Flow in a Conical Hydrocyclone". Trans. Instn.
Chem. Engrs., Vol. 51, pp. 36-41.
Bradley, D., 1965, "The Hydrocyclone". Pergamon Press.
Caldentey, J., 2000, A Mechanistic Model for Liquid
Hydrocyclones. M.S. Thesis. The University of Tulsa,
U.S.A.
Chakraborti, N. and Miller, J., 1992, "Fluid Flow in
Hydrocyclones: A Critical Review". Mineral Processing
and Extractive Metallurgy Review, Vol. 11, pp. 211-244.
Choi, M.S, 1990, "Hydrocyclone Produced Water
Treatment for Offshore Developments", SPE 20662.
Colman, D. and Thew, M., 1980, "Hydrocyclone to Give a
Highly Concentrated Sample of a Lighter Dispersed
Phase". In International Conference on Hydrocyclones,
BHRA, Cambridge, United Kingdom, paper 15, pp. 209223.
Colman, D. and Thew, M., 1983, "Correlation of
Separation Results From Light Dispersion Hydrocyclones".
Chem. Eng. Res. Des., Vol. 61, pp. 233-240.
Colman, D. and Thew, M., 1988, "Cyclone Separator".
U.S. Patent 4764287.
Colman, D., Thew, M. and Corney, D., 1980,
"Hydrocyclones for Oil/Water Separation". In International
Conference on Hydrocyclones, BHRA, Cambridge, United
Kingdom, paper 11, pp. 143-165.

11

14. Dabir, B., 1983, "Mean Velocity Measurements in a 3''Hydrocyclone Using Laser Doppler Anemometry". Ph.D.
Thesis. Michigan State University, Michigan.
15. Erdal, F., 2001, Local Velocity Measurements and CFD
Simulations in GLCC Separators. Ph.D. Dissertation. The
University of Tulsa, U.S.A.
16. Fanglu, G. and Wenzhen, L., 1987, "Measurements and
Study of Velocity Field in Various Cyclones by Use of
Laser Doppler Anemometry". In 3rd International
Conference on Hydrocyclones, Wood, P. (ed), Elsevier,
Oxford, England, pp. 65-74.
17. Fraser, S. and Abdullah, M., 1995, "LDA Measurement on
a Modified Cyclone". ASME Laser Anemometry, FEDVol. 229, pp. 395-403.
18. Gay, J.C., 1987, "Rotary Cyclone Will Omprove Oily
Water Treatment and Reduce Space Requirement/Weight
on Offshore Platforms", SPE 16571.
19. Gomez, C., 2001, Oil-WaterSeparation in Liquid-Liquid
Hydrocyclones (LLHC) Experiment and modeling. M.S.
Thesis. The University of Tulsa, U.S.A.
20. Hall, N., 1957, Thermodynamics of Fluid Flow.
Longmans, Green, New York.
21. Hargreaves, J., 1990, Computing and Measuring the Flow
field in a Deoiling Hydrocyclone. Ph.D. Thesis.
University of Southampton, England.
22. He, P., Salcudean, M., Branion, R. and Gartshore, I., 1997,
"Mathematical Modeling of Hydrocyclones". In ASME
Fluids Engineering Division Summer Meeting, FEDSM973315.
23. Hsieh, K. and Rajamani, R., 1991, "Mathematical Model of
the Hydrocyclone Based on Physics of Fluid Flow". AIChE
Journal, Vol. 37, No. 5, pp 735-746.
24. Hubred, G., Mason, A., Parks, S. and Petty, C., 2000,
"Dispersed Phase Separations: Can CFD Help?".
Proceeding of ETCE/OMAE Conference, New Orleans,
Louisiana.
25. Jirun, X., Qian, L. and Qui, J., 1990, "Studying the Flow
Field in a Hydrocyclone With no Forced Vortex I, II".
Filtration and Separation, July/August, pp. 276-278,
September/October, pp. 356-359.
26. Johnson, R., Gibson, W.E., and Libby, D.R, 1976,
"Performance of Liquid-Liquid Cyclones", Ind. Eng. Chem.
Fundam, Vol. 15, No. 2.
27. Jones, P.S.: "A Field Comparison of Static and Dynamic
Hydrocyclone", SPE Production and Facilities, May 1993,
pp. 84-90.
28. Kelsall, D., 1952, "A Study of the Motion of Solid Particles
in a Hydraulic Cyclone". Trans. Instn. Chem. Engrs., Vol.
30, pp. 87-108.
29. Kraipech, W., Chen, W. and Parma, F., 2000, "Prediction
of Hydrocyclone Performances - How Much Can the
Models Do?". American Filtration & Separation Society
Annual Conference, Myrtle Beach, SC, March 14-17.
30. Mantilla, I., 1998, Bubble Trajectory Analysis in GasLiquid Cylindrical Cyclone Separators. M.S. Thesis. The
University of Tulsa.
31. Meldrum, N., 1988, "Hydrocyclones: A Solution to
Produced-Water Tratment". SPE Production Engineering,
November, pp. 669-676.
32. Moir, D.N.: "Selection and Use of Hydrocylones", The
Chemical Engineer, January 1985, pp. 20-27.
33. Moraes, C., Hackenberg, C., Russo, C. and Medronho, R.,
1996, "Theoretical Analysis of Oily Water Hydrocyclones".

12

C. Gomez, J. Caldentey, S. Wang, L. Gomez, R. Mohan, O. Shoham

34.

35.

36.

37.

38.

39.

40.

41.

42.

43.
44.

45.
46.

47.

48.

49.

50.

In Hydrocyclones 96, Claxton, D., Svarovsky, L. and

Thew, M. (eds), M.E.P., London, England, pp 383-398.
Morsi, S. and Alexander, A., 1971, An Investigation of
Particle Trajectories in Two-Phase Flow Systems. Journal
of Fluid Mechanics, Vol. 55, part 2, pp. 193-208.
Rajamani, K. and Devulapalli, B., 1994, "Hydrodynamic
Modeling of Swirling Flow and Particles Classification in
Large-Scale Hydrocyclones". KONA Powder and Particle,
No. 12, pp. 95-104.
Rajamani, K. and Hsieh, K., 1988, "Hydrocyclone Model:
A Fluid Mechanic Approach". In Society of Mineral
Engineers Annual Meeting, Phoenix, Arizona, preprint #
88-163.
Rhodes, N., Pericleous, K. and Drake, S., 1987, "The
Prediction of Hydrocyclone Performance with a
Mathematical Model". In 3rd International Conference on
Hydrocyclones, Wood, P. (ed), Elsevier, Oxford, England,
pp. 51-58.
Seyda, B. and Petty, C., 1991, "Separation of a Light
Dispersion in a Cylindrical Vortex Chamber". Technical
Report No. HDC-R6. Hydrocyclone Development
Consortium, Michigan State University.
Sheng, H.P., "Liquid-Liquid Separation in a Conventional
Hydrocyclone", The Canadian Journal of Chemical
Engineering, Vol. 52, August 1974.
Simkin, D.J., and Olney, R.B., 1956, "Phase Separation and
Mass Tranfer in a Liquid-Liquid Cyclone", AICHE Journal,
Vol. 2, No. 4, pp. 545-551.
Smyth, I. and Thew, M., 1996, A Study of the Effect of
Dissolved Gas on the Operation of Liquid-Liquid
Hydrocyclones. In Hydrocyclones 96, Claxton, D.,
Svarovsky, L. and Thew, M. (eds), M.E.P., London,
England, pp 357-368.
Smyth, I., Thew, M., Debenham, P. and Colman, D., 1980,
Small-Scale Experiments on Hydrocyclones for Dewatering Light Oils. In International Conference on
Hydrocyclone, Cambridge, England, paper 14, pp. 189-208.
Svarovsky, L., 1984, "Hydrocyclones". Holt, Rinehart &
Winston.
Svarovsky, L., 1996, "A Critical Review of Hydrocyclones
Models". In Hydrocyclones 96, Claxton, D., Svarovsky, L.
and Thew, M. (eds), M.E.P., London, England, pp 17-30.
Syed, K.A., 1994, "The Use of Small Hydrocyclones for
Produced Water Clarification", Michigan State University.
Thew, M., 1986, "Hydrocyclone Redesign for LiquidLiquid Separation". The Chemical Engineer, July/August,
pp. 17-21.
Thew, M., Wright, C. and Colman, D., 1984, "R.T.D.
Characteristics of Hydrocyclones for the Separation of
Light Dispersions". In 2nd International Conference on
Hydrocyclones, BHRA, Bath, England, paper E1, pp. 163176.
Weispfennig, K. and Petty, C., 1991, "Flow Visualization
in a Confined Vortex Flow". Technical Report No. HDCR5. Hydrocyclone Development Consortium, Michigan
State University.
Woillez, J., Schummer, P., 1989, "A New High Efficiency
Liquid/Liquid Separator", BHRA, Multi-Phase Flow
Proceedings of the 4th International Conference, pp. 117132.
Wolbert, D., Ma, B. and Aurelle, Y., 1995, "Efficiency
Estimation of Liquid-Liquid Hydrocyclones Using

SPE71538

Trajectories Analysis". AIChE Journal, Vol. 41, No. 6, pp

1395-1402.
51. Young, G., Taggart, D. and Hild, D., 1993, Improved
Understanding of Deoiling Hydrocyclones Leads to
Significant Performance Improvement. Produced in
Amoco Production Company Research Department Tulsa
Production Research Division.
52. Young, G., Walkley, W., Taggart, D., Andrews, S. and
Worrel, J., 1990, Oil-Water Separation Using
Hydrocyclones: An Experimental Search for Optimum
Dimensions. American Filtration Society, Advances in
Filtration and Separation Technology, Vol.3, Conference
held in Baton Rouge, Louisiana.

SPE71538

OIL-WATER SEPARATION IN LIQUID-LIQUID HYDROCYCLONES (LLHC)- EXPERIMENT AND MODENLING

13

Table 1: Drag Coefficient Constants

Range

b1

b2

b3

Red < 0.1

24

0.1 < Red < 1

3.69

22.73

0.0903

1 < Red < 10

1.222

29.1667

-3.8889

0.6167

46.5

-116.67

10 < Re d < 100

Table 2: Efficiency Comparison with Present Study

Test
(#)
1
3
6
8
11
13
16
18
21
23
101
102
103
106
111
121
122
123

Droplet Size Distribution

dmin (m)
2.3
1.7
1.9
1.7
1.9
1.7
1.9
1.7
2.3
1.7
5.1
5.1
7.7
11.6
5.9
8.8
6.7
8.8

d50( m)
51
33
53
31
56
33
62
30
68
37
133
133
185
140
133
136
143
181

dmax(m)
200
116
200
116
200
133
229
116
262
133
592
517
592
592
517
517
592
592

Pinlet

Temp.

Q inlet

Oilinlet

(psia)

(psi)

(F)

(GPM)

(%)

(%)

(mm)

Efficien.

Efficien.

92
71
91
70
91
71
90
70
90
70
90
79
70
90
90
90
79
70

27
18
26
18
27
18
28
19
28
19
27
23
19
27
27
28
23
19

80
80
79
79
80
80
79
80
72
72
77
77
78
78
79
72
72
72

25
21
25
21
25
21
25
22
25
22
25
23
22
26
26
25
23
22

1
1
3
3
5
5
7
7
10
10
1
1
1
3
5
10
10
10

6
6
6
5
12
6
14
9
11
12
6
8
13
12
11
11
12
12

3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4

87
64
87
60
89
63
92
64
96
73
96
98
99
99
98
99
98
97

89
66
90
62
91
66
93
68
96
73
96
98
98
99
98
99
98
96

Split Ratio Reject Dia. Experim.

Table 3: Efficiency Comparison with Literature Data

Case Dc (mm)
9
10
11
12
13
14
15
16
17
18
19
20
21
22

Flowrate
(lpm)

30
30
30
30
30
58
58
58
58
58
58
58
58
58

60
40
50
60
70
160
190
220
250
220
250
220
250
220

Oil
Mean Experimental
Model
Density Drop Size Underflow Underflow
(g/cc)
(mc)
Purity (%)
Purity (%)
0.87
41
88
89
0.84
35
78
79
0.84
35
82
84
0.84
35
84
88
0.84
35
88
90
0.84
35
72
66
0.84
35
74
72
0.84
35
78
75
0.84
35
81
79
0.84
17
43
48
0.84
17
47
52
0.84
70
96
92
0.84
70
97
94
0.87
41
80
80

Table 4: Geometrical Parameters for Literature Data (Runs 9 to 22)

Case
8

Design
IV

Dc(mm)
20

a1
10

a2
0.75

D2
0.5Dc

L2
30Dc

Ds
2Dc

Ls
2Dc

Di
0.35Dc

Model

C. Gomez, J. Caldentey, S. Wang, L. Gomez, R. Mohan, O. Shoham

SPE71538

OIL METERING SECTION

PG4

PG3

DV2
V5

V6

V7

TT2

DV3
PG6

PG5

TANK

PG7

MM

V13
V24

MV3

V11

CV3

V8

V9

V10

TT3

BY PASS LINE

OIL

3-PHASE
GRAVITY
SEPARATOR

CV2

MV2

PUMP

OIL SKIMMER

AIR

MIXING UNIT

MM

STORAGE SECTION

V18
V14

WATER METERING SECTION

V25

PUMP

WATER

V15

TANK
V26

V16

V17

V12

V20

WATER LINE

OIL LINE

V23

V21

TEST
SECTION

Figure 4: Schematic of Experimental LLHC Flow Loop

Figure 1: LLHC Hydrodynamic Flow Behavior

Pressure Transducer
Overflow Stream

Underflow Stream

Thermometer

Static mixer

Water Stream

Isokinetic Sampler
System

Speed Controller
Mixing Loop

Oil Stream

Overflow Discharge

Acrylic Hydrocyclone

Pressure Transducer

Steel MQ Hydrocyclone

14

Gear Flow Meter

Gear Pump

Oil Stream

Figure 5: Schematic of LLHC Test Section

Figure 2: Colman and Thews Hydrocyclone Geometry

Figure 3: LLHC Inlet Design

Figure 6: Photograph of LLHC Test Section

Oil Tank

V19

SPE71538

OIL-WATER SEPARATION IN LIQUID-LIQUID HYDROCYCLONES (LLHC)- EXPERIMENT AND MODENLING

Flow Direction

Flow Direction

15

100

80

1 Inlet
Efficiency

Underflow 2

40

Surfactant

Sample Holder

60

C oil-inlet = 1 - 10 %
d 50 = 130-150 m
Qinlet = 25-21 GPM
P = 27-18 psig

20

Do= 3 mm
Do= 4 mm

0
0

10

12

14

16

Overflow Volume Percent

7
6

Figure 10: Effect of overflow diameter on efficiency

5
100

80

Efficiency

Figure 7: Schematic of Isokinetic Sampling Probe

100

Efficiency

80

60

40

Cinlet = 1%
Cinlet = 3%
Cinlet = 5%
Cinlet = 7%
Cinlet = 10%

C oil-inlet = 1 - 10 %
d 50 = 130-150 m
Qinlet = 25-21 GPM
P = 27-18 psig

20

60

0
1

40

10

100

1,000

10,000

Inlet Oil Concentration, mg/lt

DP= 27 psig, Q= 25 GPM
20

Dp= 22 psig, Q= 23 GPM

Dp= 18 psig, Q= 21 GPM

C oil-inlet = 1 - 10 %

Figure 11: Effect of oil concentration on efficiency

d 50 = 130-150 m

0
0

10

12

14

16

18

100

Overflow Volume Percent

Figure 8: Effect of pressure drop or flow rate on

Efficiency
Efficiency

100

60

40

Feed d50 = 30 um, test 3

Feed d50 = 130 um, test 101
Feed d50 = 60 um, test 1

20

80

60

10

12

14

Overflow Volume Percent

40

Pi= 90 psia, Pu= 63 psia

Coil-inlet = 1 - 10 %
d 50 = 130-150 m
Q inlet = 25-21 GPM

20

Figure 12: Effect of Droplet Size Distribution

on Efficiency

Pi= 80 psia, Pu= 57 psia

Pi= 70 psia, Pu= 52 psia

12

Inlet

0
0

10

12

14

16

Overflow Volume Percent

Figure 9: Effect of underflow pressure on efficiency

C = 1%
P = 27 psig
Q = 25 GPM
d50 = 130 m

Underflow
10

18

Volume Fraction

Efficiency

Coil-inlet = 1 %
Qinlet = 25-21 GPM
P = 27-18 psig

80

0
1

10

100

1000

Microns, um

Figure 13: Typical Measured Droplet Size Distributions

C. Gomez, J. Caldentey, S. Wang, L. Gomez, R. Mohan, O. Shoham

SPE71538

Figure 14: Rankine Vortex Tangential Velocity Profile

Figure 17: Schematic of Droplet Trajectory

and Separation Efficiency

Figure 15: Axial Velocity Diagram

Figure 18: Migration Probability Curve
3.5

Experimental Data
Model

2.5

Swirl Intensity

16

1.5

0.5

Figure 16: Schematic of Droplet Trajectory Model

0
0

200

400

600

800

1000

1200

Z (mm)

Figure 19: Swirl Intensity Comparison (Colman

and Thew, 1980, Case 2 Data)

1400

SPE71538

OIL-WATER SEPARATION IN LIQUID-LIQUID HYDROCYCLONES (LLHC)- EXPERIMENT AND MODENLING

16000

17

70

z / Dc = 10.5

P range = 45 -4 psig

60
50

Model
Flow Rate, GPM

Tangential Velocity (mm/sec)

Experimental Data
12000

8000

40
30
20

4000

Experimental
Model

10
0
0

10

20

30

40

50

60

70

80

90

100

Pressure Drop, psi

0
0

12

16

Radius (mm)

Figure 23: Comparison of Pressure Drop vs. Flow Rate

(Present Study Data)

Figure 20: Tangential Velocity Comparison (Colman

and Thew, 1980, Case 2 Data)
300

7000
250

z / Dc = 10.5

Experimental Data
200

Flowrate (lpm)

Tangential Velocity (mm/sec)

Model

150

100

Young et al (1990)
LLHC Model

50

0
0

50

100

150

200

250

Pressure Drop (psi)

-7000
0

12

16

Radius (mm)

Figure 24: Comparison of Pressure Drop vs. Flow Rate

(Young et al., 1990, Data)

Figure 21: Axial Velocity Comparison (Colman

and Thew, 1980, Case 2 Data)
12

Underflow

100

Inlet
LLHC Model

10

90

70

Volume Fraction

Separation Efficiency (%)

80

60
50
40
30
20

4
Coil-inlet = 1 %
d50 = 130 m
Qinlet = 25 GPM
P = 27 psig

Experimental Data

Model

10
0
0

16

24

32

40

48

56

Droplet Diameter (microns)

64

0
0.1

10

100

1000

Microns, um

Figure 22: Migration Probability Comparison,

Colman and Thew (1980) Data

Figure 25: Comparison of Droplet Size Distribution

Results for Test 101

18

C. Gomez, J. Caldentey, S. Wang, L. Gomez, R. Mohan, O. Shoham

100
99

Experimental
Model

Model Efficiency (%)

98
97
96
95
94
93
92
91
90
90

91

92

93

94

95

96

97

98

99

100

Experimental Efficiency (%)

Figure 26: Comparison of Model Efficiency with Present

Study Experimental Data
100
90

Model
Experimental

Model Efficiency (%)

80
70
60

50
40
30

20
10

0
0

10

20

30

40

50

60

70

80

90

Experimental Efficiency (%)

Figure 27: Comparison of Model Efficiency with

Literature Experimental Data

100

SPE71538