Third International Conference on CFD in the Minerals and Process Industries

CSIRO, Melbourne, Australia

10-12 December 2003

NUMERICAL STUDY OF GAS-SOLID FLOW IN A CYCLONE SEPARATOR

1

B. WANG , D. L. XU , G. X. XIAO , K. W. CHU and A. B. YU

1. School of Materials Science and Engineering, Xian University of Architecture and Technology, Xian 710055, P.
R. China
2.

Centre for Simulation and Modelling of Particulate Systems, School of Materials Science and
Engineering, University of New South Wales, Sydney, NSW 2052, Australia

ABSTRACT
INTRODUCTION

This paper presents a numerical study of the gas-powder

flow in a typical Lapple cyclone. The gas flow is obtained
by the use of the Reynolds stress model. The resulting
pressure and flow fields are verified by comparison with
the measured results and then used in the determination of
powder flow that is simulated by the use of a Stochastic
Lagrangian model. The separation efficiency and
trajectories of particles from the simulation are shown to
be comparable to those observed experimentally. The
effects of particle size and gas velocity on the separation
efficiency are quantified and the results are shown to agree
well with experiments.

Gas cyclone separators are widely used in industries to

separate dust from gas streams or for product recovery
because of its geometrical simplicity, relative economy in
power consumption and flexibility with respect to high
temperature and pressure. The conventional method of
predicting the flow field and the collection efficiency of
cyclone separator is empirical. During the past decades,
application of computational fluid dynamics (CFD) for the
numerical calculation of the gas flow field in a cyclone is
becoming more popular. One of the first CFD simulations
was done by Boysan [1]. He found that the standard k
turbulence model is inadequate to simulate flows with
swirl because it leads to excessive turbulence viscosities
and unrealistic tangential velocities. Recent studies
suggest that the accuracy of numerical solution can be
improved by using Reynolds Stress Model (RSM) [2-4].

NOMENCLATURE
CD
d
Fk
g
p'
rp
Re
t
u
u

drag coefficient
particle diameter, m
the momentum transport coefficient, t-1
acceleration due to gravity, m s-2
dispersion pressure, Pa
radius of particle, m
Reynolds number
time, s
instantaneous velocity, m s-1
dispersion velocity, m s-1

u
u

time average velocity in axial direction, m s-1

particle instantaneous velocity in axial direction,

m s-1
particle instantaneous velocity in radial direction,

Currently, particle turbulent dispersion due to interaction

between particles and turbulent eddies of fluid is generally
dealt with by two methods [5]: mean diffusion which
characterizes only the overall mean (time-averaged)
dispersion of particles caused by the mean statistical
properties of the turbulence, and structural dispersion
which includes the detail of the non-uniform particle
concentration structures generated by local instantaneous
features of the flow, primarily caused by the spatialtemporal turbulent eddy features and evolutions. For
prediction of the mean particle diffusion in turbulent flow,
both Lagrangian and Eulerian techniques can be used. The
stochastic Lagrangian model employs a stochastic
sampling approach to simulate the turbulent fluctuations
of the fluid at particle location. Since the early work of
Yuu et al. [6] and Gosman and Ioannides [7], the
stochastic Lagrangian model has shown significant success
in describing the turbulent diffusion of particles. It has
been reported that it is necessary to trace up to 3105
particle trajectories in order to achieve statistically
significant solutions even for two-dimensional flows [8,9].
In order to bring Stochastic Lagrangian model into
industrial applications, some modified models were
proposed. Sommerfeld et al. [10] proposed Langevin
stochastic differential equation models by making use of
Possibility Density Function (PDF). Litchford and Jeng
[11] developed a stochastic dispersion-width transport
model, where the dispersion-width is explicitly computed
through the linearized equation of motion using the
concept of particle-eddy interactions. Moreover, Chen and

m s-1

v
w

time average velocity in radial direction, m s-1

particle instantaneous velocity in tangential
direction, m s-1

w
x

time average velocity in tangential direction, m s-1

axis, m
Kronecker factor
fluid viscosity, kgm-1 s-1
density, kg m-3

SUBSCRIPTS
g
i ,j,k
p

gas
1,2,3
particle

Copyright 2003 CSIRO Australia

371

of fluid and the drag force caused by dispersion velocity of

fluid. Then the momentum equation of a particle in the
two phase flow with ambient temperature can be expressed
as:

Pereira [12] reported a SPEED model where a combined

stochastic-probabilistic method is used to describe the
turbulent motion of discrete particles so that only a small
number of particle trajectories is required.
In this paper, we used RSM and Stochastic Lagrangian
model in Fluent to study the gas-solid flow in a typical
Lapple cyclone separator. The model is verified by
comparing the simulated and measured results in term of
gas pressure and flow fields, solid flow pattern and
collection efficiency. The effects of particle size and gas
velocity are discussed.

du

dv

(2)

2
p

(3)

dw p
dt

There are usually three models used in cyclone simulation:

k model, algebraic stress model (ASM) and RSM.
The k model adopts the assumption of isotropic
turbulence, so it is not suitable for the flow in cyclone
which has anisotropic turbulence. ASM can not predict the
recirculation zone and Rankine vortex in strongly swirling
flow since it ignores or underestimates the effect of stress
convection. RSM forgoes the assumption of isotropic
turbulence and solves a transport equation for each
component of the Reynolds stress. It is thought as the most
applicable turbulence model for cyclone flow field even
though it has the disadvantage of being more
computationally expensive than other unresolved-eddy
turbulence models [2-4].

= F v + v v +

dt

MODEL DESCRIPTION

= F u + u u g

dt

F =

where

= Fk w + w w 2p

Re
18
C
24
d

is

v p wp

(4)

rp

the momentum transport

coefficient between fluid and particles, and the drag

coefficient is given as:

24

Re

24 1 + 0 .15 Re
C =
Re

0 .44

0 .687
p

Re 1
p

1 < Re 1000

where Re =

Re > 1000
p

d
p

is the Particle Reynolds

In RSM, the transport equation is written as [13]:

number, can be u , v and w . When the particle

p

interacts with fluid eddy, u

ij

ij

ij

ij

where the left two terms are the local time derivative of
stress and convective transport term, respectively. The
right five terms are:
the stress diffusion term:

SIMULATION CONDITIONS

u u u + pu + pu uu
x i j
i j k
j ik i jk

003 fluent6.0.12

De

the shear production term:

h

the pressure-strain term : = p(

ij

the dissipation term: = 2

u
u i
+ j )
x j
xi

u j
u i
+ u j u k
Pij = u iu k
x k
x k

ij

standard deviation of 2k / 3 . Particle-eddy interaction

time and dimension should not be larger than the lifetime
and size of a random eddy.

D =
ij
x
k

w is obtained by
sampling from an isotropic Gaussian distribution with a

( u u ) +
( u u u ) = D + P + + S (1)
t
x
i

u i u j
x k x k

(a)

(b)

and the source term: S

Figure 1. Schematic and grid representation of the
cyclone.

In the modelling of particle dispersion, the interactions

between particles are neglected since only dilute flow is
considered in this model. The virtual mass force, the
Basset force, the Magnus force and the Saffman force are
not considered. Only the gravity force and gas drag force
on particles are calculated. The gas drag force is
decomposed as the drag force caused by average velocity

Figure 1(a) shows the notations of the cyclone dimensions

and Table 1 indicates the dimensions of the typical Lapple
cyclone. Figure 1(b) shows the computational domain,
containing 45750 CFD cells. The whole computational
domain is divided by structured hexahedron grids. At the

372

zone near wall and vortex finder the grids are dense, while
at the zone away from wall the grids are sparse. The gas
pressure at the top of the vortex finder is 1 atm. Unless
otherwise specified, the inlet gas velocity and the particle
velocity are both 20m/s.

Dynamic Pressure
501.993
478.695
455.397
432.099
408.801
385.503
B
362.205
338.907
315.609
292.311
269.013
245.716
222.418
199.12
175.822
152.524
129.226
105.928
82.6298
59.3319
36.0339
12.736

Physical experiment has also been conducted to validate

the numerical model. The particles used are the cement
raw materials. The particle size distribution can be well
described by the Rosin-Rammler equation, with the
characteristic diameter equal to 29.90m and the
distribution parameter 0.806. The particle density is
3320kg/m3.

Table 1. Geometry of the cyclone considered (D =0.2m)

a/D

b/D

De/D

S/D

h/D

H/D

B/D

0.5

0.25

0.5

0.625

2.0

4.0

0.25

A-A

B-B

Figure 3. Contour of dynamic pressure

5500
5000

Numerical
Experimental

4500

RESULTS AND DISCUSSION

) 4000
aP
( 3500
po
rd 3000
2500
er
us 2000
se 1500
r
P 1000

Gas flow field

Figure 2 shows that the static pressure decreases radially
from wall to centre, and a negative pressure zone appears
in forced vortex. This means there is a negative pressure
zone in the center of the cyclone. The black line in Figure
2 is the dividing line between the positive static pressure
and negative static pressure. The pressure gradient is the
largest along radial direction, as there exists a highly
intensified forced vortex.

500
0

A-A

15

20

25

30

35

Figure 4. Experimental results of pressure drop compared

with calculated results
Figure 4 shows the relation between the pressure drop and
the inlet gas velocity. With the increase of the inlet gas
velocity, the pressure drop increases. The experimental
data obtained agree reasonably well with the calculated
results although consistently slightly higher.
28
26

24
22

Tangential Velocity (m/s)

10

The inlet gas velocity(m/s)

Static Pressure
845.024
798.656
715.192
671.915
619.363
542.082
498.805
440.071
368.973
325.695
260.779
195.863
152.585
81.4867
22.7531
-8.15939
-63.8018
-107.079
-187.452
-236.912
-280.189
-366.744

B-B

20
18
16
14
12
10

experimental
numerical

8
6
4
2

Figure 2. Contour of static pressure

0
-0.10

Figure 3 shows that the dynamic pressure is the largest in

the CS surface (the interface between forced vortex and
quasi-free vortex). In the quasi-free vortex zone, the
dynamic pressure increases with the radius while the
dynamic pressure decreases and tends to zero in the forced
vortex zone. Figure 3 shows that the distribution of
dynamic pressure is asymmetrical due to the nonsymmetry of the tangential velocity.

-0.08

-0.06

-0.04

-0.02

0.00

0.02

0.04

0.06

0.08

0.10

X ( m)

Figure 5. Experimental results of tangential velocity

compared with calculated results
Figure 5 shows the experimental and calculated tangential
velocities at the cylindrical section of the cyclone. The
simulation results are in good agreement with the
experimental results. The flow field in cyclone indicates
the expected forced/free combination of the Rankine

373

vortex finder will help weaken the chaotic flow and reduce
pressure drop. Figure 7 also shows the diameter of forced
vortex is a little larger than that of the vortex finder.
Moreover, since much gas flow inbursts the vortex finder,
the axial velocity reaches a peak value when gas flow into
the vortex finder.

vortex. Moreover, because the cyclone has only one gas

inlet, the axe of the vortex does not coincide with the axe
of the geometry of cyclone.
Figure 6 shows the calculated tangential velocity
distribution in detail. The tangential velocity distribution
is similar to the dynamic pressure distribution. This means
the tangential velocity is the dominant velocity in cyclone.
The value of tangential velocity equals zero on the wall
and the centre of the flow field. The high speed gas enters
the inlet and is accelerated up to 1.5~2.0 times of the inlet
velocity at point A. Then the velocity decreases as the gas
spins down along the wall. Before it goes below the vortex
finder, the gas flow collides with the follow-up flow and
forms a chaotic flow close to the vortex finder outside wall
(B point). The gas velocity decreases sharply at point B.
This is the main cause of the short-circuiting flow and
often results in a high pressure drop.
Tangential Velocity
29.4518
27.5895
25.7272
C
23.8649
22.0025
20.1402
18.2779
B
16.4155
14.5532
12.6909
10.8286
8.96624
7.10391
5.24158
3.37925
1.51693
-0.3454
-2.20773
-4.07005 A
-5.93238
-7.79471
-9.65704

A-A

From Figure 8 (A-A), we can see that the forced vortex in

the central is a twisted cylinder. The axis of the forced
vortex is not coincided at the geometrical axis of cyclone,
and also is not a line but a curve. There is a zone right
under the vortex finder where gas flows into the vortex
finder directly instead of spinning down to the conical
section and then flowing upward. This is the shortcircuiting flow, which does harm to cyclone performance.
In the conical section, the radial velocity is much larger
than that of cylinder section. This will drag some particles
into the forced vortex and these particles will not be
collected. From Figure 8 (B-B), we see that the
distribution of radial velocity is nearly uniform in the
quasi-free vortex area. The distribution of the radial
velocity in the forced vortex is eccentric because of the
non-symmetrical geometry of the cyclone. Figure 8 (C-C)
shows that the radial velocity is negative or inward flow in
the gas inlet and then becomes zero rapidly. Afterwards it
becomes positive due to the effect of centrifugal force
around the vortex finder. At point A, the radial velocity
becomes negative again, directing to the centre, because of
the collision among gas.

C-C

B-B

Radial Velocity
6.16873
5.29714
4.86135
4.3633
3.55397
3.11817
2.55787
1.81079
1.375
0.752432
0.0676124
-0.368182
-1.053
-1.67557
-2.11136
-2.85844
-3.41874
-3.85454
-4.66387
-5.16192
-5.59771
-6.4693

Figure 6. Contour of tangential velocity

Axial Velocity

A-A

21.4941
19.9052
18.3163
16.7274
15.1385
13.5496
11.9607
10.3718
8.78294
7.19405
5.60515
4.01625
2.42736
0.838464
0
-1.54488
-3.13377
-4.72267
-6.31157
-7.90046
-9.48936
-11.0783

A
C-C
A

A-A

B-B

Figure 8. Contour of redial velocity distribution

Particles flow pattern

Figure 9 shows the change in location with time of 15000
particles with five diameters within 1 second. Red, orange,
green cyan and blue respectively represent five diameters
of particles, that is, 1 10-4m, 3 10-5m 7 10-6m, 2
10-6m and 2 10-7m. From this figure, it can be seen that
the trajectory of the largest particles (red) is at the upside
of the cone, the trajectory of the smallest particles (blue) is
at the downside of the cone. The other three sized particles
are largely in-between the two extremes.

B-B

Figure 7. Contour of axial velocity

Figure 7 shows that the forced vortex is a twisted cylinder
and not completely axially symmetric, especially in the
conical section. From Figure 7 (B-B) we see the centre of
the forced vortex dose not coincided with the geometrical
centre of cylindrical body of the cyclone, which deflected
to the gas inlet. This should be one of the main reasons
why there is eccentric vortex finder in some revised
cyclone separators and some modified inlet shapes were
proposed [14]. From Figure 6 (C-C), we see eccentric

374

There is therefore a critical value to distinguish the flow

pattern of particles of different diameters. If the particle
diameter is larger than the critical value, the particle will
keep a circular motion in the cone of cyclone. In contrast,
if the particle diameter is less than this critical value, the
particles will be collected directly or escape from the
cyclone. The critical value is related to the geometry of
cyclone, the gas inlet velocity and the properties of
particles. In this cyclone, the critical diameter is
approximately 1 10-5m.

t=0.05s

t=0.1s

t=0.15s

t=0.2s

t=0.3s

t=0.35s

t=0.4s

t=0.45s

In order to verify the numerical simulation results,

physical experiments have been done by use of two types
of ceramic balls whose density is similar to the cement raw
material. The diameter of yellow ceramic ball is 1mm, and
the green ones are 2mm. The experimental results are
shown in Figure 10. In Figure 11 (a), the pure cement raw
material was used. It is observed that the particles flow
downward at the cone section and display a certain
descending angle. On the other hand, ceramic ball
(Figure11-b) kept spinning at a certain height and did not
show the descending angle. This phenomenon supported
the simulation results. In the real industry, these bigger
particles will be eventually collected due to their
interactions with other particles.

t=0.25s

t=0.5s

2
t=0.55s

t=0.6s

t=0.8s

t=0.85s

t=0.65s

t=0.9s

t=0.7s

t=0.75s

t=0.95s

t=1.0s

10 m 210 m 710 m 310 m 110 m

-7

-6

-6

-5

-4

Figure 10. The trajectories of particles with different

diameters

Figure 9. Animation of particle flow

As shown in figure 9, large particles are collected while
small particles escape from the cyclone. The particles with
the smallest diameter can not move outward to the wall of
cyclone since the centrifugal force on them is not bigger
than the gas drag force on them. The particles with
diameters of 2 10-6m and 7 10-6m can spin down to the
conical section of cyclone and then should be collected
while the bigger particles with diameter of 3 10-5m and 1
10-4m spin downward first and then keep spinning near
the wall at a certain horizontal level.

Figure 11. Photos showing the trajectories of tracing

particles of different diameters

375

100

6.

Separation efficiency ( % )

98

5.

LOTH E., Numerical approaches for motion of

dispersed particles, droplets and bubbles, Progress
in Energy and Combustion Science 26 (2000) 161223.

6.

YUU S., YASUKOUCHI N., HIROSAWA, Particle

turbulent diffusion in a dust laden round jet,
AIChE Journal 24 (1978), 509-519.

7.

GOSMAN A.D. and IOANNIDES E., Aspects of

computer simulation of liquid-fuelled combustors.
AIAA 19th Aerospace Science Mtg., Paper 81-0323
(1981) St. Louis, Mo..

8.

STURGESS G.J., SYED S.A., Calculation of a

hollow-cone liquid spray in uniform airstream.
Journal of Propulsion and Power, 1 (1985) 360-369.

9.

MOSTAFA A.A., MONGIA H.C., MCDONELL,

V.G. and SAMUELSEN, G.S., Evolution of
particle-laden jet flows: a theoretical and
experimental study, AIAA Journal, 27 (1989)167183.

96
94
92
90
88
86

Numerical
Experimental

84
82
80
0

10 12 14 16 18 20 22 24 26 28 30 32 34

Inlet gas velocity(m/s)

Figure 11. Experimental results of separation efficiency

compared with calculated results
The most important economical parameters of a cyclone
separator are separation efficiency and pressure drop.
Generally, The increase of gas inlet velocity will increase
the separation efficiency, but it will also increase the
pressure drop. In this work, physical and numerical
experiments have both been done to find the effect of gas
inlet velocity on separation efficiency and pressure drop.
As shown in Fig. 4, the pressure drop increases with the
inlet gas velocity, and there is a good agreement between
predicted and measured results. Figure 11 shows that the
collection efficiency can be enhanced with the increase of
inlet gas velocity, as expected. The prediction matches the
measurement reasonably well.

10. SOMMERFELD M, KOHNEN G and RUGER M.

Some open questions and inconsistencies of
Lagrangian particle dispersion models, Proc. Ninth
Symp. on Turbulent Shear Flows, Kyoto, Japan,
Paper (1993) .
11. LITCHFORD R.J. and JENG SM, Efficient
statistical transport model for turbulent particle
dispersion in sprays. AIAA Journal 29 (1991)
1443-1451.

CONCLUSIONS

12. CHEN XQ, PEREIRA JCF, Efficient computation

of particle dispersion in turbulent flows with a
stochastic-probabilistic model, Int. J. Heat and
Mass Transfer, 40 (1997) 1727-1741,.

RSM has been used to simulate the anisotropic turbulence

flow in a Lapple cyclone. Its applicability has been
verified by the good agreement between the calculated and
measured pressures and flow fields. On this basis, a
stochastic Lagrangian model has been used to predict the
flow pattern of particles in the cyclone and its validity is
confirmed by comparing the predicted and measured solid
flow trajectories and collection efficiency. The proposed
model provides a convenient way to study the effects of
variables related to operational conditions, cyclone
geometry and particle properties, which is important to the
optimum design and control of cyclone process.

13. SHUN R. and LI Z.Q., Simulation of strong swirling

flow by use of different turbulence model, Power
Engineering 22 (2002).
14. SUASNABAR, D. J., Dense medium cyclone
performance enhancement via computational
modelling of the physical processes, Ph.D thesis
(2000).

REFERENCES
1.

BOYSAN F, AYER WH, SWITHENBANK JA,

Fundamental mathematical-modeling approach to
cyclone design, Transaction of Institute Chemical
Engineers, 60 (1982) 222-230.

2.

HOEKSTRA A.J., DERKSEN J.J., H.E.A. VAN

DEN AKKER An experimental and numerical
study of turbulent swirling flow in gas cyclones,
Chemical Engineering Science 54 (1999) 2055-2056.

3.

PANT K., CROWE C.T., IRVING P., On the design

of miniature cyclone for the collection of
bioaerosols, Powder Technology 125 (2002) 260265

4.

SOMMERFELD M., HO C. H., Numerical

calculation of particle transport in turbulent wall
bounded flows, Powder Technology 131 (2003) 1-

376