Advanced Powder Technology 23 (2012) 228–233

Contents lists available at ScienceDirect

Advanced Powder Technology

journal homepage: www.elsevier.com/locate/apt

Original Research Paper

Evaluation of natural gas dehydration in supersonic swirling separators applying

the Discrete Particle Method
Chuang Wen, Xuewen Cao ⇑, Yan Yang, Jing Zhang
Department of Oil and Gas Engineering, China University of Petroleum, No. 66, Changjiang West Road, Qingdao 266555, China

a r t i c l e i n f o a b s t r a c t

Article history: The natural gas flow fields and particles separation characteristics were numerically calculated with the
Received 4 August 2010 RNG k  e turbulence model and Discrete Particle Method (DPM) in the supersonic swirling separator. An
Received in revised form 14 February 2011 experimental system was set up for testing the separation efficiency of three new designed separators
Accepted 18 February 2011
with wet air. The numerical results showed that the new annular nozzle not only expanded the natural
Available online 4 March 2011
gas to supersonic velocity with resulting in low temperature (72 °C), but also strengthened the swirling
flow with the centrifugal field of 640 000g (g is the acceleration of gravity), both of which created good
Keywords:
conditions for natural gas dehydration. Under the strong swirling flow field, most particles collided with
Natural gas dehydration
Supersonic swirling separator
the walls or entered into the liquid-collection space directly, while only few particles escaped together
Separation efficiency with the gas flow. The separation efficiency reached over 95%, when the length of the cyclone separation
Discrete Particle Method (DPM) section was about 10 times of the diameter of the wall at throat. The experimental results indicated that
the water can be efficiently removed from the wet air. The numerical results were in good agreements
with the experimental findings, which demonstrated that the Discrete Particle Method (DPM) was accu-
rate and stable enough to evaluate the dehydration characteristics of the supersonic swirling separator.
Ó 2011 The Society of Powder Technology Japan. Published by Elsevier B.V. and The Society of Powder
Technology Japan. All rights reserved.

1. Introduction
The demand for natural gas has motivated the oil and gas indus-
try to discover natural gas reservoirs in remote areas. However, in
the process of producing and transporting natural gas, the presence
of water in natural gas can cause corrosion, excessive pressure
drop, hydrate, the decrease of its heating value and the reduction
in gas transmission efficiency. The possibility of pipeline obstruc-
tion due to the formation of hydrate is one of the most serious
problems in the gas industry. Therefore, it is important to assure
that water is removed as the gas is transported from the wellhead
to a processing facility.
Supersonic swirling separators have been introduced to treat
the problems in natural gas for offshore applications [1,2]. Alferov
et al. [3] and Betting et al. [4] proposed a method and apparatus for
the separation and liquefaction of the gas mixtures, respectively.
Alferov et al. [5] introduced a technological process for supersonic
separator and compared the effectiveness of the 3-S separator,
Joule–Thomson valve, and turbo-expander in extracting C3+ from
natural gas. Liu et al. [6] described a supersonic swirling
dehydration system for natural gas and the indoor experiment
was carried out to test the unit performance. The effects of the
⇑ Corresponding author.
E-mail address: caoxw@upc.edu.cn (X. Cao).
temperature, flow rate and pressure loss ratio on the dehydration
characteristics were analyzed. The key components of Twister sep-
arator and hydrocarbon recovery were investigated by Betting and
Epsom [7]. Jiang et al. [8] developed a mathematical model to
investigate the one-dimensional transonic flow of two-component
gas mixture with spontaneous condensation. The mixing flow field
of the nitrogen and water vapor is numerical simulated in a super-
sonic separator.
The computational fluid dynamics technique was used to study
the behavior of high-pressure natural gas in supersonic nozzles by
Jassim et al. [9,10]. The effects of real gas and nozzle geometry on
the natural gas flow behavior in the nozzle were discussed. The
influences of vorticity on the performance of the nozzles and shock
wave positions were studied. Shock wave with reasonable strength
was beneficial to the particles separation. Selective dehydration of
high-pressure natural gas through supersonic nozzles was investi-
gated by Karimi and Abdi [11]. A computational model linked to
MATLAB and HYSYS package was presented to predict the effect
of different parameters such as the inlet pressure, inlet
temperature and flow rate on the behavior of the working fluid.
However, few researchers focused on the characteristics and
separation efficiency of the liquid particles in the supersonic
swirling separator.
In the two-phase flow models, the governing equations of
continuous phase are generally written in Eulerian form, whereas,

0921-8831/$ - see front matter Ó 2011 The Society of Powder Technology Japan. Published by Elsevier B.V. and The Society of Powder Technology Japan. All rights reserved.
doi:10.1016/j.apt.2011.02.012
 C. Wen et al. / Advanced Powder Technology 23 (2012) 228–233
Fig. 1. Flow chart of the supersonic swirling separator. 1–compressor, 2–heat
exchanger, 3–humidifier, 4–supersonic swirling separator, 5–dry gas outlet,
6–surge tank, 7–gas/liquid separator, 8–liquid outlet.
the governing equations of the discrete phase can be described in
either Eulerian or Lagrangian form [12]. Comparing with Euler–
Euler model, the Euler–Lagrange model, in which the motion of
each particle can be individually tracked, provides the most
detailed insight into single particle dynamics.
In this paper, we investigate the particle trajectories and sepa-
ration efficiency by using the Discrete Particle Method (DPM),
which can simulate a discrete second phase in a Lagrangian frame
of reference. The experimental study was carried out for wet air
based on our three new designed supersonic separators.
2. Experimental setup
An experimental system was set up to test the separation effi-
ciency according to experimental requirements and the experi-
mental flow chart of the supersonic swirling separator was
shown in Fig. 1. The working fluid was the wet air, which could
be obtained by humidifying the unsaturated air using the humidi-
fier. The supersonic swirling separator was constructed of stainless
steel. The Atlas Copco GA45VSD type screw compressor provided a
maximum working pressure of 1.3 MPa. Its air displacement is
360 m3/h. Vaisala HMT3608 type humidity/temperature transduc-
ers were installed both at the inlet and dry gas outlet of the super-
sonic swirling separator. The accuracy of the transducer is ±1.5%
and ±0.2 °C for relative humidity and temperature, respectively.
Three Rosemount3051 type pressure sensors were located in the
supersonic separator inlet, dry gas out1et and liquid outlet. The
precision of these pressure sensors is 0.075%. YOKOGAWA DY040
type vortex flow meters were set in the inlet and dry gas outlet.
The precision of these flow meters is ±0.75%.
3. Numerical conditions
FLUENT software is a CFD (Computational Fluid Dynamics) sol-
ver of choice for complex flows ranging from incompressible (low
subsonic) to mildly compressible (transonic) to highly compress-
ible (supersonic and hypersonic) flows. The flows, in the super-
sonic swirling separator, are very complicated, including swirling
flow and supersonic velocity. The commercially available FLUENT
code was utilized here for the investigation.
In a supersonic swirling separator, the gas phase is treated as
continuous and the locally averaged Navier–Stokes equation is
used to describe the mass, momentum and energy conservation
equations for the continuous phase [13–15]. A particle motion
can be described as translation and rotation, which are completely
determined by Newton’s second law of motion. During the move-
ment, the particle may collide with its neighbor particles or wall at
the contact points and interact with the surrounding fluid, through
which the momentum and energy are exchanged.
The model for the discrete liquid particles is at any individual
particle level, whereas that for the fluid flow is at the computa-
tional cell level. Therefore, it needs to couple the numerical scales.
229
Fig. 2. New geometry of the supersonic swirling separator.
This is actually achieved by considering the gas-particle interaction
force based on the Newton’s third law of motion.
In the supersonic swirling separator, small liquid particle sus-
pended in the gas which has a great temperature gradient experi-
ences a force in the direction opposite to that of the gradient.
Therefore, the gas-particle interaction force includes the drag force
and the thermophoretic force [16,17].
3.1. Geometry
In our supersonic swirling separator, the Laval nozzle was de-
signed to be a new annular nozzle, which incorporates a central
body, allowing the principle of conservation of angular momentum
to be harnessed. The channel, between the wall and the central
body, forms a Laval nozzle, which is composed of three sections:
the convergent (subsonic zone), throat section (critical zone) and
divergent section (supersonic zone), as seen in Fig. 2.
The pressure, flow rate, and temperature of natural gas for the
design of supersonic separator are 4 MPa, 10  104 Nm3/d and
30 °C, respectively. The supersonic separator is about 587.4 mm
long, 80 mm and 40 mm in the inlet and outlet diameter. The
diameters of the central body and wall at the throat are 12.00
and 18.65 mm, respectively.
3.2. Grid system
The quality of the mesh plays a significant role in the accuracy
and stability of the numerical calculation. The polyhedral meshes
allow the flexibility of an unstructured mesh to be applied to a
complex geometry. The swirling vanes were twisted and the cy-
clone separation section was complicated. Therefore, the super-
sonic swirling separator is meshed by the hexahedral and
tetrahedral grids, and divided by the coarse, moderate and fine
grids.
In our initial study, the grid independency was investigated
with the mesh number of 920,825 and 193,738, respectively. The
numerical results show that the axial distributions of the Mach
number are consistent. Therefore, grid independency studies have
shown that larger grids do not necessarily influence the accuracy in
the case of the supersonic swirling flow, while the mesh size is in a
certain range.
3.3. Turbulence model and solution algorithm
In the supersonic swirling separator, the strong swirling flow is
generated by the swirling vanes. Therefore, the RNG k  e turbu-
lence model is used here, because it provides an option to account
for the effects of swirl by modifying the turbulent viscosity
appropriately [18,19].
Mathematical calculation adopted the finite volume method,
second-order upwind scheme and wall function applied in along-
wall, while coupling numeration of velocity field and tress was
based on SIMPLE algorithm [20]. The second-order scheme was used
for the Pressure Interpolation Scheme as a result of the compressible
230 C. Wen et al. / Advanced Powder Technology 23 (2012) 228–233
swirling flows in the separator. The time step was 1e-6 in the numer-
ical simulation. The convergence criterion is 106 for energy equa-
tion and 103 for all other equations.
3.4. Particle size distribution
Gas expands in the supersonic swirling separator to supersonic
velocities with resulting in lower temperatures, which leads to
the nucleation and condensation of the water. The typical size
of the liquid particle is 0.1–4 lm. In the calculation, the
Rosin–Rammler model [21] is used to describe the distribution
of the liquid particle size. If the size distribution is of the
Rosin–Rammler type, the mass fraction of droplets of diameter
greater than d is given by:
 n
Y d ¼ eðd=dÞ
where Yd is the mass fraction of droplets; d is the droplet diameter;
 is the size constant and n is the size distribution parameter. In our
d ugal field of about 640 000g (g is the acceleration of gravity).
simulation, the particle size constant d  is 1.4 lm; the size distribu-
tion parameter n is 1.62.
3.5. Initial and boundary conditions
The Discrete Particle Method (DPM) contains the assumption
that the second phase is sufficiently dilute that particle–particle
interactions and the effects of the particle volume fraction on the
gas phase are negligible. In the numerical simulation, the liquid
particles were injected along with the gas phase. It was supposed
that the particles speed were equal to the gas velocity calculated
by the simulation results. Variable numbers of the liquid particles
(10,000, 40,000 and 100,000) were used to predict the particle tra-
jectories and separation efficiency. The results were in reasonable
agreement with each other. Therefore, the number of 10,000 liquid
particles was utilized for our simulations.
Pressure boundary conditions are imposed for the inlet and out-
let of the supersonic swirling separator. Total pressure (absolute
pressure) and total temperature at the nozzle inlet are set to be
4 MPa and 303 K, respectively. The static pressure and stagnation
temperature are chosen for outlet boundary conditions, which
are 2.68 MPa and 290 K, respectively. Turbulence intensity and vis-
cosity ratio are assigned as turbulence parameters. No-slip and adi-
abatic boundary conditions are specified for the walls.
In addition, for the discrete phase, the escape boundary condi-
tions are assigned for the inlet and dry gas outlet. That is, the par-
ticles will not return the computational domain, if they reach the
inlet due to the backflow. The trap boundary conditions are used
for the liquid outlet and the walls of the cyclone separation section
and the liquid-collection space. We appoint the reflect boundary
conditions for all other walls.
3.6. Equation of state
In this paper, it is assumed that the gas obeys the model equa-
tion of state which enables one to take into account the effect of
super-compressibility in Eq. 2.
PV ¼ mRTZ
where P, V, m, R, T, Z are the gas pressure, volume, mass, gas con-
stant, temperature, coefficient of super-compressibility, respec-
tively. The coefficient of super-compressibility Z is a function of
pressure and temperature.
The mole composition of natural gas is as follows: 2.04% N2,
0.45% CO2, 0.03% H2O, 91.36% CH4, 3.63% C2H6, 1.44% C3H8, 0.26%
i-C4H10, 0.46% n-C4H10, 0.17% i-C5H12, and 0.16% n-C5H12.
4. Results and discussion
4.1. Natural gas flow field
The natural gas flow field was numerical simulated in the
supersonic swirling separator based on the above mentioned
numerical methods. The average static temperature and tangential
velocity of cross section was analyzed along the axis. Numerical re-
sults are presented in Fig. 3.
In the supersonic swirling separator, natural gas expands in the
annular Laval nozzle to supersonic velocities with resulting in low-
er temperatures. The static temperature decreases slowly in the
convergent section of the nozzle, while gas expansion induces a ra-
pid decrease in the divergent part. The average static temperature
is about 72 °C at the nozzle exit. The swirling motion is generated
ð1Þ by the vanes at the entrance of the nozzle. The swirl strength in-
creases strongly, due to the contraction in the annular nozzle and
the second convergence of the central body, resulting in a centrif-
As a result of the low temperature and the strong centrifugal
field, the nucleation and condensation of water will occur, and
the liquid droplets can be centrifuged onto the walls with creating
a liquid film. Then the gas liquid mixture passes through the cy-
clone separation section, in which the static temperature increases
and tangential velocity decreases slowly because of the wall fric-
tion. The liquid particles enter into the liquid-collection space for
removal. The dry gas moves across the diffuser where a shock wave
occurs. The remaining kinetic energy in the gas flows is trans-
formed to increase the static pressure.
4.2. Liquid particle trajectory
As above-mentioned the water will be condensed from natural
gas mixtures duo to the low temperature in the supersonic separa-
tor. In this section, the Discrete Particle Method (DPM) was em-
ployed to predict the flow characteristics of the liquid particles,
with the following assumption:
(1) It is assumed here that the particle is a sphere, and the evap-
oration process of the liquid particles could be ignored.
(2) The initial velocity of the liquid particles is equal to the gas
speed, obtained by the numerical calculation.
(3) Thermal equilibrium exists between gaseous and liquid
particles.
(4) The liquid particles are removed as soon as they are cap-
tured on the wall.
ð2Þ
Fig. 3. Average static temperature and tangential velocity along the axis.
 C. Wen et al. / Advanced Powder Technology 23 (2012) 228–233
The liquids particles are released and tracked at the nozzle out-
let. Fig. 4 shows the predicted liquid particle trajectories in the
supersonic separator. Obviously, the liquid particles move in circu-
lar motion under the combined effects of the drag force, the ther-
mophoretic force and the centrifugal force. Fig. 4 demonstrates
that most of the droplets enter into the liquid-collection space
and then are removed from natural gas mixtures, while only few
liquid particles pass through the diffuser with the dry gas.
Fig. 5 presents a single particle trajectory. It can be seen that the
particles have three representative moving states, which are: (a)
the particle colliding with the wall, (b) the particle entering into
the liquid-collection space and (c) the particle passing through
the diffuser. For the first kind trajectory in Fig. 5(a), the particle
is centrifuged onto the walls, creating a thin liquid film, which gets
into the liquid-collection space along the wall. Some particles enter
into the liquid-collection space directly without collision with the
wall. For other few particles, the drag force plays a much more sig-
nificant role than other forces, as shown in Fig. 5(c). Therefore,
these particles pass through the diffuser together with the gas
flow, which cause the gas phase entraining few liquids. In addition,
some slip-gas flows through the liquid-outlet.
4.3. Separation efficiency
The flow velocity is supersonic at the nozzle outlet in the super-
sonic separator. The gas speed is so high that it results in a
Fig. 4. Liquid particles trajectories.
Fig. 5. Single particle trajectory. (a) Particle colliding with the wall. (b) Particle entering into the liquid-collection space. (c) Particle passing through the diffuser.
231
maximum gas residence time less than 2 ms. Therefore, the length
of the cyclone separation section greatly affects the gas/liquid sep-
aration performance. Fig. 6 shows the effects of the length of the
cyclone separation section on the separation efficiency. The dimen-
sionless parameter l represents the various lengths, which is the
ratio of the length of the cyclone separation section to the diameter
of the wall at throat. For a given design condition, the dimensions
of the throat in the supersonic separator is determined. Therefore,
the dimensions of the throat can be chosen as a characteristic
parameter.
The separation efficiency g for the liquid particles is calculated
according to:
mtrap
g¼  100% ð3Þ
mtrap þ mescape
where mtrap and mescape are the mass flow of the liquid particles pass-
ing through the liquid-outlet and dry-gas outlet, respectively.
It can be seen that the separation efficiency is significant im-
proved with the increase of the cyclone section length. When the
ratio of the length to the diameter l is 10, the separation efficiency
reaches over 95%. This is because the increasing length not only in-
creases the wall surface area for particles collision, but also extends
the particles residence time for gas/liquids separation. However,
the gas/liquid temperature will increase slowly due to the presence
of the friction loss, in the cyclone separation section. As a result of
the rise of the static temperature, the liquid particles, condensed
232 C. Wen et al. / Advanced Powder Technology 23 (2012) 228–233
Fig. 6. Effects of the cyclone section length on the separation efficiency.
by the low temperature, will re-evaporate into the gas phase,
which decease the particles separation efficiency. Therefore, we
suggest that the length of the cyclone separation section is about
10 times of the diameter of the wall at throat for high separation
efficiency.
4.4. Experiment
Three new supersonic swirling separators were designed for the
experimental investigation. The length l of the cyclone section is
3.33, 6.67 and 10.0, respectively. The separation efficiency for the
liquid particles in wet air has been tested.
The separation efficiency of the liquid particles in the three
supersonic swirling separators was tested based on the above
mentioned experimental system. Meanwhile, the air–water parti-
cles two-phase flow was numerically calculated with the Discrete
Particle Method (DPM). The experimental and numerical results
are shown in Fig. 7.
The numerical and experimental results present that the length
of the cyclone separation section has a great effect on the separa-
tion performance. The results of repeated experiments have shown
that the separation efficiency is about 66%, 80%, and 94% in the
three different supersonic swirling separators with the length l of
3.33, 6.67 and 10, respectively. In most cases, the calculated values
are more than the experimental data, and the maximum error is
about ±6.5%. We consider that this is because we neglected the ef-
fects of the re-evaporation of the liquid particles. The good agree-
ment with the computational and experimental results indicate
that the Discrete Particle Method (DPM) can effectively evaluate
the dehydration characteristics of natural gas, and high separation
performance of the water can be obtained in our new designed
supersonic swirling separator.
5. Conclusion
In the supersonic swirling separator, a new annular Laval nozzle
with a central body was designed for natural gas dehydration. The
water was condensed from natural gas mixtures due to the low
temperature (72 °C). The contraction of the central body can
strengthen the swirling flow with the centrifugal field of 640 000g.
The Discrete Particle Method (DPM) was applied to predict the
particle trajectories and separation efficiency. Most particles col-
lided with the walls or entered into the liquid-collection space di-
rectly, while only few particles escaped together with the gas flow.
The separation efficiency reached over 95%, when the length of the
Fig. 7. Separation efficiency with wet air.
cyclone separation section was about 10 times of the diameter of
the wall at throat.
The separation efficiency of the supersonic swirling separator
was experimental tested with saturated wet air for working fluid.
The results of repeated experiments have shown that the separa-
tion efficiency can reach about 94% and the maximum error is
about ±6.5%. The high separation performance of the water can
be obtained in our new designed supersonic swirling separator.
Acknowledgements
This study is supported by National High Technology Research
and Development Program of China (‘‘863’’, No.2007AA09Z301)
and National Major Science & Technology Specific Projects
(No.2008ZX05017-004).
References
[1] D.F. Okimoto, J.M. Brouwer, Supersonic gas conditioning, World Oil 8 (2002)
89–91.
[2] J.M. Brouwer, H.D. Epsom, Twister supersonic gas conditioning for unmanned
platforms and subsea gas processing, in: Proc. of the Offshore Europe
Conference, 2003, pp. 219–225.
[3] V.I. Alferov, L.A. Baguirov, V. Feygin, A. Arbatov, S. Imaev, L. Dmitriev, V.I.
Rezunenko, US Patent, 6372019, Apr 16, 2003.
[4] M. Betting, T. Van Holten, J. Van Veen. US Patent, 6524368, Feb 25, 2003.
[5] V.I. Alferov, L.A. Baguirov, L. Dmitriev, V. Feygin, S. Imaev, J.R. Lace, Supersonic
nozzle efficiently separates natural gas components, Oil & Gas Journal 103
(2005) 53–58.
[6] H. Liu, Z. Liu, Y. Feng, K. Gu, T. Yan, Characteristic of a supersonic swirling
dehydration system of natural gas, Chinese journal of chemical engineering 1
(2005) 9–12.
[7] M. Betting, H. Epsom, Supersonic separator gains market acceptance, World Oil
4 (2007) 197–200.
[8] W. Jiang, Z. Liu, H. Liu, J. Zhang, X. Zhang, Numerical simulation of two-
component mixture in one-dimension supersonic separator, Acta petrolei
sinica petroleum processing section 6 (2008) 697–701.
[9] E. Jassim, M.A. Abdi, Y. Muzychka, Computational fluid dynamics study for flow
of natural gas through high-pressure supersonic nozzles: Part 1. Real gas
effects and shockwave, Petroleum Science and Technology 26 (2008) 1757–
1772.
[10] E. Jassim, M.A. Abdi, Y. Muzychka, Computational fluid dynamics study for
flow of natural gas through high-pressure supersonic nozzles: Part 2. Nozzle
geometry and vorticity, Petroleum Science and Technology 26 (2008) 1773–
1785.
[11] A. Karimi, M.A. Abdi, Selective dehydration of high-pressure natural gas using
supersonic nozzles, Chemical Engineering and Processing 48 (2009) 560–568.
[12] C. Cortes, A. Gil, Modeling the gas and particle flow inside cyclone separators,
Progress in Energy and Combustion Science 5 (2007) 409–452.
[13] T.B. Anderson, R. Jackson, Fluid mechanical description of fluidised beds,
equation of motion, Industrial and Engineering Chemistry Fundamentals 6
(1967) 527–539.
[14] D. Gidaspow, Multiphase Flow and Fluidization, Academic Press, San Diego,
1994.
 C. Wen et al. / Advanced Powder Technology 23 (2012) 228–233 233

[15] R.B. Bird, E.S. Warren, E.L. Edwin, Transport phenomena, John Wiley & Sons,
New York, 1960.
[16] H. Ounis, G. Ahmadi, J.B. McLaughlin, Brownian diffusion of sub-micrometer
particles in the viscous sub-layer, Journal of Colloid and Interface Science 143
(1991) 266–277.
[17] L. Talbot, R.K. Cheng, R.W. Schefer, D.R. Willis, Thermophoresis of particles in a
heated boundary layer, Journal of Fluid Mechanics 101 (1980) 737–758.
[18] S.B. Pope, Turbulent Flows, Cambridge University Press, Cambridge, 2000.
[19] A. Gupta, R. Kumar, Three-dimensional turbulent swirling flow in a cylinder:
Experiments and computations, International Journal of Heat and Fluid Flow 2
(2007) 249–261.
[20] S.V. Patankar, Numerical heat transfer and fluid flow, Taylor & Francis, New
York, 1980.
[21] M. Peleg, Determination of the parameters of the Rosin–Rammler and beta
distributions from their mode and variance using equation-solving software,
Powder Technology 2 (1996) 181–184.