chemical engineering research and design 8 9 ( 2 0 1 1 ) 603610

Contents lists available at ScienceDirect

Chemical Engineering Research and Design

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

Simulation and experiment of ow eld in axial-ow

hydrocyclone
Wang Zhen-bo, Ma Yi , Jin You-hai
College of Mechanical and Electronic Engineering, China University of Petroleum (East China), Dongying 257061, China

a b s t r a c t
Axial-ow hydrocyclone is a new type of hydrocyclone which is proposed on the basis of tangential hydrocyclone.
Its key part is rotating and accelerating component the guided vane. The structural particularity made it necessary
to pay attention to the internal ow eld of axial-ow hydrocyclone. In view of the progress of CFD technique and
mature of calculation condition, numerical simulations were carried on single-phase ow in axial-ow hydrocyclone
by using Fluent software and Reynolds Stress Model. The initial conditions were that uid medium was water and the
overow rate was 15%. The pressure and velocities elds were analyzed in the whole hydrocyclone and the specic
effects of guided vane were studied. At the same time, through the verication of LDV experiment, the simulated
values of water phase matched well with the experimental values.
Crown Copyright 2010 Published by Elsevier B.V. on behalf of The Institution of Chemical Engineers. All rights
reserved.
Keywords: Hydrocyclone; Computational uid dynamics; NavierStokes equation; RSM model; Turbulence

1.

Introduction

Hydrocyclone was a device which was used to separate two

components with different densities with the aid of the strong
centrifugal force created by the swirling ow (Vieira et al., 2005;
Nowakowski et al., 2004). It is compact in structure, small in
size, light in weight, easy to design and install, and has low
maintenance cost. These make hydrocyclone share a wide
range of applications in many industries, particularly in the
petrochemical eld. Axial-ow hydrocyclone is a new type of
hydrocyclone which is proposed on the basis of the structure
of tangential hydrocyclone. The typical structure was shown
in Fig. 1.
The work principle of axial-ow hydrocyclone is that after
entering the hydrocyclone, along the guided vane the mixture
gradually ow into a high-speed rotation, in the role of centrifugal force (Narasimha et al., 2005), the heavy phase move to
the wall, form the downward outer swirl, and ow away from
underow outlet, while the light phase move to the axis, form
the upward inner swirl, and ow away from overow outlet.
The experiments (Xuming et al., 1995) had found that compared to the traditional tangential hydrocyclone, axial-ow

hydrocyclone has the advantages of lower pressure drop and

larger handling capacity.
The ow behavior in hydrocyclone is quite complex, which
has led designers to rely on empirical equations to research
its performances (Lynch and Rao, 1975; Rajamani and Milin,
1992). But it is proved that this method is not accurate and general. In view of this shortcoming, strengthening the theoretical
study of the separate principle is particularly important. The
development of computer technology and computational uid
dynamics provides powerful tool for people to study hydrocyclone. Simulation can enhance the understanding of ow
elds within the hydrocyclone and bring about improvements
in its design (Delgadillo and Rajamani, 2007). In addition,
a number of investigators reported their measurements of
hydrocyclone using Laser Doppler Velocimetry (LDV) (Hsieh
and Rajamani, 1991; Dai et al., 1999; Hsieh, 1988). It has proved
that LDVs measurement results have been widely recognized.
Therefore, in order to deeply understand the ow characteristics of hydrocyclone, this paper used numerical simulation
and LDV measurements to study the internal ow eld in
axial-ow hydrocyclone. The distributions of static pressure,
axial velocity and tangential velocity were obtained and the

Corresponding author. Tel.: +86 13506369521.

E-mail addresses: wangzhenbote@sina.com (W. Zhen-bo), mylittleant@sina.com (M. Yi).
Received 17 February 2009; Received in revised form 27 June 2010; Accepted 8 September 2010

0263-8762/$ see front matter Crown Copyright 2010 Published by Elsevier B.V. on behalf of The Institution of Chemical Engineers. All rights reserved.

doi:10.1016/j.cherd.2010.09.004

604

chemical engineering research and design 8 9 ( 2 0 1 1 ) 603610

is stress production term; Fij is rotation production term; the

remaining items were as follows:
The turbulent diffusion term:

DTij =
xk

 
t ui uj
k xk


(5)

The pressure strain term:

ij = C1 

ui uj

2
k
3 ij

C2 pij

2
p
3 ij


(6)

The dissipation term:

Fig. 1 The typical structure of axial-ow hydrocyclone.

2
ij = ij 
3

accuracy of simulated results were veried by LDV experimental results.

2.

ui = u i + ui

(1)

And the governing equations for the velocity eld in incompressible uid can be written as in Eqs. (2) and (3):
(ui )
=0
xi


where C1 and C2 were constants, k is turbulent kinetic

energy, is turbulent dissipation rate, and t is turbulent
viscosity coefcient.

Theoretical considerations

The swirling ow in hydrocyclone (Barbosa et al., 2003) is a

classical problem of hydromechanics which could be simplied as single-phase ow in the calculation. In this study,
the simulations were carried out using the computational
uid dynamics (CFD) technique. The numerical treatment
of the NavierStokes equations is the backbone of any CFD
technique. NavierStokes incompressible equations supplemented by a suitable turbulence model were appropriate for
modeling the ow in hydrocyclone.
In Reynolds averaging, the solution variables in the instantaneous NavierStokes equations were represented by the
mean and uctuating components. The velocity eld is
dened as the sum of its mean value (u i ) and uctuation value
(ui ), and it can be shown in Eq. (1):

(u u ) =
+
xj i j
xi
xj

(2)




uj
xj


ui uj

(3)

Eqs. (2) and (3) are called Reynolds-averaged NavierStokes

(RAN) equation. They have the same general form as the
instantaneous NavierStokes equations, with the velocities
and other solution variables now representing ensembleaveraged (or time-averaged) values. Additional terms represent the effects of turbulence. These Reynolds stresses (ui uj )
must be modeled in order to calculate Eq. (3).
In this model, the transportation equations used to solve
ui uj is dened as Eq. (4):

(uk ui uj ) = DT,ij + DL,ij + pij + ij + ij + Fij

(ui uj ) +
t
xk

(7)

These constitute the basic governing equations of a threedimensional turbulent ow problem. Since the RSM accounts
for the effects of streamline curvature, swirl, rotation, and
rapid changes in strain rate in a more rigorous manner, it has
a greater potential to accurately predict complex ows, as in
the case of hydrocyclone.

3.

Model description

3.1.

Geometry

The hydrocyclone geometry used for simulation and for experimental studies was presented in Fig. 2. The body is cylindrical
with 50 mm diameter. A circular feed inlet with 30 mm diameter is connected to the cylindrical surface. A guided vane
of four runners with 50 mm length and 11 outlet angle is
below the inlet. The cone with 3 angle is connected to the
main cylindrical body with 65 mm length. And the bottom
of hydrocyclone is a tail tube. A cylindrical vortex nder
protrudes into the main cylindrical body extending a long
length inside and a small length above the top closed surface
(Svarovsky, 1984).

3.2.

Meshing scheme

Earlier reports (Wang and Yu, 2008) have indicated that the
results using a 3D model to simulate the hydrocyclone much
better match with the experimental data compared to the
results using axisymmetric geometry. Thus 3D computational
model was selected for the study.
To improve the meshs quality (Cullivan et al., 2003),
structurized mesh was used as far as possible, and part of
ow regions were divided more precisely. The total number of the structured mesh was 112,350. Fig. 3 showed
the mesh structure in different parts of the axial-ow
hydrocyclone.

(4)

where the two terms in the left of Eq. (4) were the local time
derivative of the stress and convective transportation term,
respectively. And DL,ij is molecular viscous diffusion term; pij

3.3.

Boundary and initial conditions

In Fluent computations, the boundary conditions were set as

follows:

chemical engineering research and design 8 9 ( 2 0 1 1 ) 603610

605

lent kinetic energy k and turbulent dissipation rate were

obtained.
(2) Outlet boundary: Vortex nder and outow tube were set as
outow, where the overow split was 15% in the light of
actual experimental data.
(3) Wall boundary: The wall was non-slipping boundary, and
the walls roughness was 0. For the grid spacing near
wall without strict limits, the wall approximate method
of standard function was used in Reynolds stress models
concrete computation.

4.

Simulation

In this paper, the behavior of single phase was simulated.

Simulations were carried out using Cartesian coordinate system, steady state and segregated solver. Reynolds stress model
(RSM) was adopted with the consideration of aeolotropic turbulent transportation. Many models, such as k model were
so difcult to consider the inuences resulting from rotating
ow and surfaces curvature change of ow direction. But RSM
model overcomes these shortcomings effectively (Mainza et
al., 2006; Souza et al., 2000). That was why RSM was used in
simulating hydrocyclone.
In numerical solution, the nite volume method was used
to control the diffusion of the governing equation, and the
SIMPLE algorithm was used for coupling the continuity and
momentum equations, and PRESTO! was adopted by pressure
interpolation scheme to predict swirling ow characteristics,
and a third order accurate QUICK scheme was used for spatial
discretization.
NavierStokes equations were solved sequentially till convergence factor (105 ) was met. The simulation results were
shown as follows.
Fig. 2 The basic model of axial-ow hydrocyclone.

(1) Inlet boundary: A velocity inlet condition was used to prescribe uids inow through the circular feed inlet, and
the inlet velocity was set as 1.18 m/s. Accordingly, turbu-

4.1.

Fluid ow path

The ow paths of water were shown in Fig. 4. It could be

seen that water moved in the form of spiral ow in hydrocyclone, rotated around axis after entering the cavity, and went

Fig. 3 Grid diagram of axial-ow hydrocyclone.

606

chemical engineering research and design 8 9 ( 2 0 1 1 ) 603610

Fig. 4 Fluid ow path in hydrocyclone.

Fig. 5 Velocity vector chart of the vane outlet cross-section.

Fig. 6 Complete picture chart of ow eld distribution. (a) Static pressure; (b) axial velocity; (c) tangential velocity.

chemical engineering research and design 8 9 ( 2 0 1 1 ) 603610

downward gradually, then in somewhere of the cone a small

portion of water owed upward through inner swirl to vortex
nder, while the majority of water continued to ow downward to underow tube. It could be seen that the main motion
types include the main motion types in hydrocyclone include
inner vortex and outer vortex. Besides, it was clear that the
turbulence near vortex nder and underow tube was more
serious. These kinds of phenomena investigated existence of
short-circuit ow and circular ow in axial-ow hydrocyclone.

4.2.

The vectors of the vane outlet

The guided vanes played an important role in uids rotation

and acceleration. The vectors of the vane outlet were shown
in Fig. 5. The velocity vectors of the four-vane outlets and vortex nder could be seen clearly. The same swirl directions have
shown that the consistency of the directions of inner swirl and
outer swirl. Moreover, the vectors displayed a symmetrical distribution. So the guided vanes could prevent the eccentricity
effectively.

4.3.

XY plot of ow eld distribution

Fig. 6 was a complete picture chart of ow eld distribution

in longitudinal section of axial-ow hydrocyclone. As radial
velocity was very small, it was not taken into consideration.
For the convenience of accurate comprehension, three critical cross sections were chosen to draw comparisons more
directly (as shown in Fig. 6): the interface of cylinder and cone
Section 154, middle section in cone Section 450, and the
section near the spigot outlet Section 1000. The main ow
parameters including static pressure, axial velocity and tangential velocity of these sections were shown in Fig. 7(a)(c)
respectively. Concrete analyses are as follows.

4.3.1.

607

Static pressure

The distributions of static pressure on various sections

(Fig. 6(a)) were basically in accordance with those of composed vortex, and appeared as axial symmetrical distribution
in main separation area. And static pressure decreased gradually along the radial direction from the side wall to the axis
center, that was, there were a nearby axis area of low pressure
and a nearby wall of high pressure. The curves of static pressure in every section were almost so coincident that the radial
pressure gradients were very small which could be seen in
Fig. 7(a). From the gure, in the axis of hydrocyclone, the values
of static pressure showed a slight decline with the uids ow.

4.3.2.

Axial velocity

Axial velocity actually decides the overow and underow

assignment of the uid medium. Seen from Figs. 6(b) and 7(b),
axial velocity had the points of zero velocity in both sides of
the central axis in every section (Bhaskar et al., 2007). It was
obvious that these points were the turning point of the direction, and bounded by this point, axial velocity was positive in
internal part and increases gradually, which meant uid ow
upwards and this region was that of inner swirl. While axial
velocity was negative in external part, that means uid ow
downwards and this region was that of outer swirl. It was an
important characteristic of axial velocity, which was called the
locus of zero vertical velocity. Trough analyses of the gures,
it could be found that LZVV assumed cylindrical surface in
the cylinder of hydrocyclone, and conical surface in the cone.
So the conclusion could be drawn that LZVV in main separation space of axial-ow hydrocyclone is a surface combined
by cylindrical and conical surface.
In addition, with the sections going downward, the
maximum of positive values of axial velocity (in axis of hydrocyclone) decreased and on the contrary, the maximum of

Fig. 7 Pressure and velocities of three typical sections. (a) Static pressure; (b) axial velocity; (c) tangential velocity.

608

chemical engineering research and design 8 9 ( 2 0 1 1 ) 603610

Fig. 8 Sketch map of experiment unit in laser test.

Fig. 9 Contrast curve of simulated value and experimental value in Section 1.

negative values (near the wall) increased. Inevitably, these
phenomena caused the changes of the distribution of upward
and downward ow. With the uids getting closer to the spigot
outlet, upward ow decreased gradually, but the downward
ow was just the opposite. It could prove that the main separation spaces of hydrocyclone are the cylinder and cone.

4.3.3.

Tangential velocity

In terms of tangential velocity (as shown in Figs. 6(c) and 7(c)),

tangential velocity assumed a hump-shaped distribution
along the radial direction. Take the maximum as the boundary. Tangential velocity of internal quasi-forced vortex was
decreasing gradually with decreasing radius, and became
almost zero in the center. Tangential velocity of external
quasi-free vortex was decreasing with increasing radius, and
cut down to the minimum near the internal wall.
The tangential velocitys distributions in every section were
roughly similar. Along with uids downward move, tangential
velocity is gradually decreasing. And the structure of the cone
had expiatory effect on the loss of uids angular momentum.

5.

Experiment

The experimental system was mainly composed of axial-ow

hydrocyclone, the water supply system, LDV laser measuring
system, and ow and pressure measuring system, as shown in
Fig. 8. After stirring, trace particle mixed with the water evenly
in water tank, and then after centrifugal pumps compression
and acceleration, the mixture went though the entrance of the
hydrocyclone, and nally came back to the water tank through
overow and underow outlet. And a cycle of the mixtures

ow nished. In LDV laser measuring system, the ber probe

was controlled by the three-dimensional coordinate frame,
which provides the possibility to measure in different radius
of the different sections. In measuring, the emissive optical
axis of probe was in the same horizontal plane with the central axis of axis-ow hydrocyclone so that the laser could rip
symmetrically.
The operating and structural parameters used in the experiment were same to those used in simulation. Tangential
velocity and axial velocity obtained in the ow elds test were
compared with the simulation results in such three sections
as the following gures (Figs. 911).

5.1.

Section 154: the interface of cylinder and cone

The cylinder was the following part of the vane outlet, and
in view of the particularity of its location, the interface of
cylinder and cone was worth paying attention to. Seen from
Fig. 9, the simulation results matched well with the experimental results. For tangential velocity, the simulated values
were slightly bigger than the experimental values. For axial
velocity, the region of r < 3 was in the overow tube, the axial
velocity was upward. And because the diameter of the overow tube was small and the section was near the mouth of
overow tube, the upward region was smaller. Besides, axial
velocity near the boundary layer was downward, and this
region was of outer swirl.

5.2.

Section 450: middle section in cone

The cone was the main separated space which played an

important role in separating effect. Seen from Fig. 10, for tan-

chemical engineering research and design 8 9 ( 2 0 1 1 ) 603610

609

Fig. 10 Contrast curve of simulated value and experimental value in Section 2.

Fig. 11 Contrast curve of simulated value and experimental value in Section 3.

gential velocity, the simulation results matched better with
the experimental results in this section. Tangential velocity in
this space was composed of internal quasi-forced vortex and
external quasi-free vortex, which increased rst to the maximum and then decrease to 0 along with the radial distance.
For axial velocity, the distribution in Section 450 was similar to that in Section 154. But in this section, the difference
between the maximum and minimum of axial velocity was
smaller. Numerically, it could be clearly seen that axial velocity decreased obviously with the uids ow. On the one hand,
that was the manifestation of upward ows decrease. On the
other hand, this phenomenon could prolong the residence
time of uid which was conductive to separation.

5.3.

Section 1000: the section near the spigot outlet

The underow tube of axial-ow hydrocyclone was lengthened. The main role was to stabilize the uid entering from the
cone and cause it ow away steadily. Seen from Fig. 11, axial
velocity and tangential velocity were smaller and their simulated results and experimental results agreed well. For axial
velocity, the simulated values were slightly bigger than the
experimental values, and the upward ow was much smaller
than downward ow in this section. For tangential velocity,
the simulated values were more deected to the wall than
experimental values.
Generally speaking, through the comparisons in the cylinder, cone and underow tube, the simulated results matched

well with the experimental results in tendency and value. This

showed that using RSM model to simulate in hydrocyclone can
reect the internal ow eld accurately.

6.

Conclusions

In this paper, numerical simulation and LDV measurements

were done to study the internal ow eld of axial-ow hydrocyclone. The simulation results using RSM model matched
well with the experimental results. It has been proved that
guided vanes could divide the uid evenly and prevent the
eccentricity effectively. Through analyses of the whole ow
eld, the characteristics of the distribution of three important
ow parameters could be obtained: the distribution of static
pressure has the same distribution in every section, which
increased with the radius; the distribution of axial velocity
assumed upward and downward ow and the existence of
LZVV; Tangential velocity was composed of internal quasiforced vortex and external quasi-free vortex.

References
Barbosa, E.A., Vieira, L.G.M., Damasceno, J.J.R., Barrozo, M.A.S.,
2003. Differences of behavior between ltering hydrocyclone
with Bradley and Rietema geometry. Mater. Sci. Forum 416,
317322.
Bhaskar, K.U., Murthy, Y.R., Raju, M.R., Tiwari, S., Srivastava, J.K.,
Ramakrishnan, N., 2007. CFD simulation and experimental
validation studies on hydrocyclone. Miner. Eng. 20, 6071.

610

chemical engineering research and design 8 9 ( 2 0 1 1 ) 603610

Cullivan, J.C., Williams, R.A., Cross, C.R., 2003. Understanding the

hydrocyclone separator through computational uid
dynamics. Trans. IChemE 81, 455466.
Dai, G.Q., Chen, W.M., Li, J.M., Chu, L.Y., 1999. Experimental study
of solidliquid two-phase ow in a hydrocyclone. Chem. Eng.
J. 74, 211216.
Delgadillo, J.A., Rajamani, R.K., 2007. Exploration of hydrocyclone
designs using computational uid dynamics. Int. J. Miner.
Process. 84, 252261.
K.T. Hsieh, 1988. Phenomenological model of the hydrocyclone.
Ph.D. Thesis. The University of Utah, USA.
Hsieh, K.T., Rajamani, R.K., 1991. Mathematical model of the
hydrocyclone based on physics of uid ow. AIChE J. 37,
735746.
Lynch, A.J., Rao, T.C., 1975. Modeling and scale-up of
hydrocyclone classiers. In: Carta, M. (Ed.), Proceedings of the
11th International Mineral Processing Congress. Cagliari.
Aziende Tipograche Bardi, Rome, pp. 245269.
Mainza, A., Narasimha, M., Powell, M.S., Holtham, P.N., Brennan,
M., 2006. Study of ow behaviour in a three-product cyclone
using computational uid dynamics. Miner. Eng. 19,
10481058.
Narasimha, M., Sripriya, R., Banerjee, P.K., 2005. CFD modelling of
hydrocyclone prediction of cut size. Int. J. Miner. Process. 75,
5368.

Nowakowski, A.F., Cullivan, J.C., Williams, R.A., Dyakowski, T.,

2004. Application of CFD to modelling of the ow in
hydrocyclones. Miner. Eng. 17, 661669.
Rajamani, R.K., Milin, L., 1992. Fluid ow model of the
hydrocyclone for concentrated slurry classication. In:
Hydrocyclones: Analysis and Application: 4th Intentional
Conference, vol. 12, pp. 95108.
Souza, F.J., Vieira, L.G.M., Damasceno, J.J.R., Barrozo, M.A.S., 2000.
Analysis of the inuence of the ltering medium on the
behaviour of the ltering hydrocyclone. Powder Technol. 107,
259267.
Svarovsky, L., 1984. Hydrocyclone. Holt, Rinehart & Winston,
Eastbourne, UK.
Vieira, L.G.M., Barbosa, E.A., Damasceno, J.J.R., Barrozo, M.A.S.,
2005. Performance analysis design of ltering hydrocyclones.
Braz. J. Chem. Eng. 22, 143152.
Wang, B., Yu, A.B., 2008. Numerical study of the gasliquidsolid
ow in hydrocyclones with different conguration of vortex
nder. Chem. Eng. J. 135, 3342.
Xuming, D., Wangzhenbo, Youhai, J., 1995. An experimental
investigation on the performance comparison of two kinds of
inlets of hydrocyclones. Chem. Eng. Mach. (China) 32 (2),
6971.