Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/257378406

Numerical investigation of cone angle effect on the flow field and separation
efficiency of deoiling hydrocyclones

Article  in  Heat and Mass Transfer · February 2012


DOI: 10.1007/s00231-012-1085-8

CITATIONS READS

21 1,349

3 authors, including:

Maysam Saidi Bijan Farhanieh


Razi University Sharif University of Technology
40 PUBLICATIONS   173 CITATIONS    102 PUBLICATIONS   916 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Respiratory drug delivery View project

wrap around fin configurations View project

All content following this page was uploaded by Reza Maddahian on 18 April 2016.

The user has requested enhancement of the downloaded file.


Heat Mass Transfer
DOI 10.1007/s00231-012-1085-8

ORIGINAL

Numerical investigation of cone angle effect on the flow field


and separation efficiency of deoiling hydrocyclones
Maysam Saidi • Reza Maddahian • Bijan Farhanieh

Received: 13 December 2011 / Accepted: 8 October 2012


Ó Springer-Verlag Berlin Heidelberg 2012

Abstract In this study, the effect of cone angle on the flow Qi Inlet volume flow rate (m3/h)
field and separation efficiency of deoiling hydrocyclones is R Flow split (Qoverflow/Qinlet) (%)
investigated taking advantage of large eddy simulation. The Re Reynolds number (–)
_ __
dynamic Smagorinsky is employed to determine the residual Lij Lij ¼ ui uj  ui uj (m2/s2)
stress tensor of the continuous phase. The method of  2 _ _ 
Mij _    _
Lagrangian particle tracking with an optimized search algo-   2 
Mij ¼ D  S  Sij D S Sij (m2/s2)
rithm (closest cell) is applied to evaluate the separation effi-
PDR Pressure differential ratio (–)
ciency of deoiling hydrocyclone. Simulations are performed
Sij Strain tensor (1/s)
on a 35-mm deoiling hydrocyclone with the three different
Ud Droplet velocity (m/s)
cone angles of 6, 10 and 20 degree. The numerical results
Vd Droplet diameter (m3)
revealed that the changes in the cone angle would affect the
Xd Droplet coordinate (m)
velocity and pressure distribution inside hydrocyclone, and
Z Axial distance from the top wall (mm)
lead to changes in the separation efficiency. However, the
large cone angle increases the tangential velocity and pressure Greek letters
gradient inside the hydrocyclone, but reduces the separation h Cone angle (deg)
efficiency. The reasons behind the decrease in the separation l Viscosity (kg/ms)
efficiency are the flow structure and reduction of oil droplets m Kinematic viscosity (m2/s2)
residence time in hydrocyclones with large cone angles. q Density (kg/m3)
s Stress tensor (N/m2)
List of symbols D Filter width (m)
CD Drag coefficient (–)
Subscripts
CS Smagorinsky constant (–)
d Droplet
D Hydrocyclone diameter (m)
f Fluid phase (water)
d Droplet diameter (m)
i, j, k, l Coordination index
FD Drag force (kg/m2 s2)
FP Pressure gradient force (kg/m2 s2) Superscripts
FV Virtual mass force (kg/m2 s2) 0
Time fluctuation quantity
k Density ratio (k = qd/qf) (–)  Filtered quantity
p Static pressure (kg/ms2) _ Test filtered quantity

M. Saidi  R. Maddahian (&)  B. Farhanieh 1 Introduction


Center of Excellence in Energy Conversion (CEEC),
School of Mechanical Engineering, Sharif University
of Technology, P.O. Box: 11365-9567, Tehran, Iran Although hydrocyclones have been utilized in separation
e-mail: r_maddahian@mech.sharif.edu of materials for over a 100 years, application of deoiling

123
Heat Mass Transfer

hydrocyclones in offshore platforms has been developed in The oil–water flow in deoiling hydrocyclones was simu-
the recent years. The need for a high-efficiency compact lated by Hargreaves and Silvester [14] taking advantage of
oil–water separator, especially in platforms where there is Eulerian–Lagrangian method. They employed Algebraic
space limitation, is one of the main reasons for develop- Stress Model with a 2-D cylindrical coordinate system. The
ment of deoiling hydrocyclones. effects of particle–particle interaction, slip and droplet
The swirling flow inside hydrocyclones induces a cen- coalescence are ignored. The obtained results were in
trifugal force and leads to separation owing to the density acceptable agreement with the experimental data. The idea
difference between phases. The major differences between of using Lagrangian Particle Trajectory (LPT) method for
the separation processes occurring in deoiling and desander estimation of efficiency was also implemented by Wolbert
hydrocyclones were reported by Thew [1] and Caldenty et al. [15]. The flow field, velocity distribution and sepa-
[2]. The density difference in solid–liquid mixture is ration efficiency of a 10-mm deoiling hydrocyclone was
greater than liquid–liquid types. Therefore, separating solid obtained by Grady et al. [16] using Algebraic Slip Mixture
from liquid is easier than liquid from liquid. The solid (ASM) and Reynolds Stress Model (RSM) multiphase
particles migrate to the wall region in desander hydro- model. Petty and Park [17] employed direct numerical
cyclones while oil droplets move to center in deoiling simulation to simulate a miniature hydrocyclone. Their
hydrocyclones. Thus, the near wall region is of particular results showed that a 3g centrifugal acceleration is created
significance in desander hydrocyclones, whereas the center in the 5 mm miniature hydrocyclones for the estimated
flow features become the predominant area of study in the pressure drop and flow rates of 1 l/s and 1 kPa, respectively.
deoiling types. The liquid droplets may be broken up to the Huang [18] simulated the three dimensional turbulent flow
smaller ones due to the shear rate increasing to a critical in deoiling hydrocyclones using Eulerian–Eulerian
level. In addition to the mentioned differences, the flow approach and Reynolds Stress Model. The obtained results
splits of desander and deoiler hydrocyclones are distinct, so showed accumulation of oil near the axis. The separation
the flow feature of the continuous phase is not the same. efficiency was also estimated based on phase concentra-
The discussed differences have led to difficulties in the tion. The separation curve for Colman type hydrocyclones
design of deoiling hydrocyclones. was in good agreement with the measured ones. The effect
Although the application of common hydrocyclones in of various inlet types and inlet chamber body profiles on
oil–water separation was suggested by Simkin and Olney [3] the separation efficiency of deoiling hydrocyclones was
and Sheng et al. [4], fundamental studies on deoiling studied by Noroozi and Hashemabadi [19, 20]. The sepa-
hydrocyclones were started by Colman and Thew [5]. The ration efficiency was improved by 10 and 8 % with
experimental investigations of Colman et al. [6] and Colman application of a helical inlet and an exponential body
and Thew [7] indicated the independency of separation profile. Kharoua et al. [21] simulated the Colman type
efficiency from the flow split if it is in the range 0.5 to 10 %. hydrocyclone making use of the slip mixture model. They
Moreover for constant droplet size distribution in the inlet, revealed that Reynolds Stress Model (RSM) can better
the size distribution in the outlet is independent of the flow predict the separation efficiency in comparison with
split. The migration probability curves are also independent Renormalization Group (RNG) k - e. The velocity field
of the flow split. A new design of hydrocyclone for water and the effective parameters on turbulence inside deoiling
treatment was proposed by Thew [1]. High swirl flow, hydrocyclones, such as oil droplet diameter, flow rate and
smaller size preventing large pressure drop and minimum inlet oil concentration, were investigated. The complete
instability, and turbulence near the axis were the specific review of hydrocyclones for the deoiling purpose can be
features of the new design. The operation curves, principle of found in [22]. The first application of LES turbulence
operations and the first field study of hydrocyclones were model for simulation of the continuous velocity field inside
obtained by Meldrum [8]. Young et al. [9] tried to optimize deoiling hydrocyclones was introduced by Saidi et al. [23].
the 35-mm hydrocyclone designed by Colman and Thew [5]. They illustrated the capability and potential of large eddy
They studied the effects of operational and geometrical simulation in prediction of deoiling hydrocyclones’ flow
parameters, such as inlet size, cylindrical diameter, cone field providing a comparison between LES, LRR and
angle, straight section length, flow rate, and droplet diameter standard k - e models. They also calculated the separation
on the separation efficiency. Based on their experimental efficiency with different drag correlations.
results, a new geometry was proposed for hydrocyclones. The literature review discloses the lack of information
The recent investigations on hydrocyclones focus on oper- about the effect of geometrical parameters such as cone
ational parameters [10, 11], velocity field [12] and distribu- angle, underflow diameter and etc. on the velocity field and
tion of oil droplets [13] in deoiling hydrocyclones. separation efficiency of deoiling hydrocyclones. In this
Among the numerous investigations on deoiling hydro- regard, the authors tried to shed some light on this subject
cyclones, only few numerical studies have been conducted. using the advantages of Large Eddy Simulation (LES) in

123
Heat Mass Transfer

simulation of swirl flows [23]. The recent investigations ssgs


ij ¼ ui uj  u
i uj ð4Þ
show the applicability of LES approach not only in pre-
diction of velocity field, but also in evaluation of the sep- Using Eq. (1), the residual stress tensor can be written
aration efficiency associated with desander and deoiling as:
hydrocyclones [23–28]. ssgs 0 0 i u0j þ u0i uj þ ui uj  ui uj
ij ¼ ui uj þ u ð5Þ
It is worth to mention that the previous researches on
deoiling hydrocyclones have been performed using Rey- Equation (5) is another representation of the residual
nolds Averaged Navier–Stokes (RANS) turbulence models, stress tensor incorporated in the dynamic Smagorinsky
whereas the present research is one of the primary inves- model.
tigations on geometrical parameters of deoiling hydrocy- One of the most typical methods of modeling residual
clones using LES approach. stress tensor employs the eddy viscosity approach defined
The most important geometric factor which influences as below:
the separation efficiency is cone angle. Consequently, three dij sgs
different designs with the cone angles of 6, 10 and 20 ssgs
ij  s ¼ 2mt Sij ð6Þ
3 kk
degree are considered in this research. The continuous
velocity field is obtained taking advantage of LES where Sij is the strain tensor defined by Eq. (7).
 
approach. Besides, a FORTRAN code is developed to track 1 oui o uj
the oil droplets inside the deoiling hydrocyclones and Sij ¼ þ ð7Þ
2 oxj oxi
estimate the separation efficiency. The results of tangential
and axial velocities, pressure distribution and separation Smagorinsky [29] used the following expression to
efficiency of the different designs are presented. calculate the turbulent kinematic viscosity:
 
mt ¼ ðCS DÞ2 S ð8Þ
  pffiffiffiffiffiffiffiffiffiffiffiffi
where S ¼ 2Sij Sij and D is the filter width. Smagorinsky
2 Governing equations
constant (CS) is found to vary in the range of 0.065 [30] to
0.25 [31] depending upon flow and geometry. Assuming a
2.1 Continous phase
constant value for CS is one of the flaws of Smagorinsky
subgrid model. To overcome this problem, we applied
Navier–Stokes (continuity and momentum) equations
dynamic Smagorinsky model which introduces some ideas
assuming Newtonian incompressible flow with negligible
to adopt this constant dynamically and calculate it. Germano
volume forces are considered in this research. A general
et al. [32] proposed a dynamic SGS model in which CS is
low-pass filter is applied to the Navier–Stokes equations to
calculated and is not an arbitrarily selected value. The main
decompose the velocity into the resolved and residual
idea behind dynamic Smagorinsky model consists of
components. The large scales affected by the flow geom-
introducing a test filter (D) with a larger width than the
etry specify the properties of the turbulent flow such as heat
original one. In the numerical simulations, the grid size is
and mass transfer, and therefore, should be resolved. The
often assumed as the primary filter, and named as the implicit
small scales only dissipate energy and could be modeled
numerical filter or the grid filter, while the test filter is applied
using an appropriate subgrid turbulence model.
explicitly. This two-level filtering technique acts as an
The decomposed velocity (resolved and residual) com-
experimental tool of investigation and helps to reach better
ponents can be written as:
results physically [33]. Applying the test filter to implicit
ui ¼ ui þ u0i ð1Þ filtered Navier–Stokes equation leads to:
Applying the decomposed velocity into mass and 1 Lij Mij
CS ¼  ð9Þ
momentum equations and performing the filtration 2 Mkl Mkl
process results in the following equations: To enhance the stability of the model, local averaging is
o
ui employed to calculate dynamic Smagorinsky constant. In
¼0 ð2Þ addition, negative values of the effective viscosity are
oxi
removed by clipping to zero.
sgs 
o ui uj Þ
ui oð 1 o
p o2 ui osij
þ ¼ þm  ð3Þ 1 Lij Mij
ot oxj q oxi oxj oxj oxj CS ¼  ð10Þ
2 hMkl Mkl i  2 _ _ 
_   _
where ssgs
ij is the residual stress tensor which describes the
_ __
where Lij ¼ ui uj  ui uj and Mij ¼ D  S  Sij D2 SSij .
unresolved scales and can be written as:

123
Heat Mass Transfer

Further details of the turbulence model can be found in where Red is the droplet Reynolds number and defined as
Germano et al. [32] and Lilly’s [34] studies. follows:
 
qf Uf  Ud d
2.2 Dispersed phase Red ¼ ð14Þ
lf
The oil droplets are considered as the dispersed phase in
the deoiling hydrocyclone. In order to estimate the sepa- 3 Geometry of the problem
ration efficiency of the deoiling hydrocyclone, the
Lagrangian Particle Tracking (LPT) method is applied with The simulations are performed on 35-mm hydrocyclones
the following assumptions [2, 15]: with two symmetrical inlets (5 9 10 mm) entering tan-
– The particle–particle interaction, coalescence or break gentially from top of its cylindrical section. The hydrocy-
up is ignored. clone has two outlets, one at its top and the other at its end,
– The concentration of oil droplets is assumed to be named overflow and underflow, respectively. Overflow is
sufficiently dilute so that Lagrangian approach can be continued to the inside of the hydrocyclone with the length
applied. of LO and called vortex finder. The three different hydro-
– Droplets are considered rigid, and their deformation cyclone designs are labeled (A), (B) and (C) having cone
due to shear forces is ignored [35]. angles of 6, 10 and 20 degrees, respectively. The geomet-
rical parameters of designs are displayed in Fig. 1.
The density difference between the continuous and dis- Flow split is defined as the ratio of the volume flow from
persed phases is small. Therefore in addition to the drag overflow to that of the inlet. The flow split varies in the
force, the pressure gradient and virtual mass forces should be range of 0.04–0.06 in the present work. Outlet pressures are
taken into account as well [36]. Table 1 summarizes the changed to reach the flow split value. Table 2 presents the
forces acting upon oil droplets in the deoiling hydrocyclones. operational parameters of the deoiling hydrocyclone. This
In the Lagrangian reference frame for unsteady flows, hydrocyclone is designed for oily waste water refinement
the force balance on oil droplets results in:
~d X
dU
qd V d ¼ ~
F ð11Þ
dt
By substituting forces from Table 1 in Eq. (11) we have:
    
1 dU ~d 3 1  
1þ ¼ CD ðU ~ d ÞU
~f  U ~d
~f  U
2k dt 4Dd k
! !
1 DU ~f 1 DU ~f
þ þ ð12Þ
k Dt 2k Dt

where CD is drag coefficient and k is density ratio (k = qd/qf).


Drag coefficient for liquid droplets is different from solid
particles. The internal circulation generated due to the slip
flow between the oil droplet and water leads to decrease in
the drag force. The correlation suggested by Hadamard [37]
and Rybczynski [38] for creeping flow (Red \1) is used. The
validity of the mentioned correlation is approved by recent
studies [39]. For Reynolds numbers between 1 and 500, the
drag coefficient introduced by Rivkind and Ryskin [40] is
applied. For higher Reynolds numbers, the drag coefficient
has the constant value of 0.44.
All the correlations used in calculation of the drag
coefficient are presented in Eq. (13).
8
8 3kþ2
>
< Red ð kþ1 Þ Red 1
CD ¼ 1þk 1 24 1=3 0:78
> ½kðRe þ 4Red Þ þ 14:9Red  1\Red 500
: d
0:44 Red [500
ð13Þ Fig. 1 Hydrocyclone geometry dimensions

123
Heat Mass Transfer

Table 1 Important forces acting upon oil droplets in deoiling


hydrocyclones
Force Correlation

Drag force 2
 
F D ¼ ðpr 2Þqf CD ðU
~ ~ d ÞU
~f  U ~d 
~f  U
~
Pressure gradient force ~ DU
F P ¼ qf Vd Dt f
~
Virtual mass force ~ qV DU ~d
F V ¼ f2 d Dt f  DU
Dt

having oil concentrations less than 0.3 vol.% in the inlet


[12].

4 Mesh generation and boundary conditions

The generated mesh for simulation of the flow inside


deoiling hydrocyclones is illustrated in Fig. 2. The mesh
Fig. 3 Tangential velocity at Z/D = 2.29 and 6 degree of cone angle
becomes finer in the regions with higher gradient, espe- with different mesh sizes
cially near the wall and the core of the hydrocyclone. Three
different grids are utilized to show the grid independency
of the results. The tangential velocity of design (A) for Table 2 Operational parameters of hydrocyclone
three different grid sizes, coarse (434,128 cells), medium Qi R qwater qoil lwater loil
(670,404 cells) and fine (1,025,912 cells) is depicted in (m3/h) (%) (kg/m3) (kg/m3) (kg/m-s) (kg/m-s)
Fig. 3. It can be observed that the maximum differences
1.5 4–6 1,000 850 0.001 0.0032
between the coarse and medium, and between the medium
and fine grids are 23 and 10 %, respectively. The nodal
distributions of the final grids considered for simulations
Table 3 Number of grids in each direction and overall number of
are provided in Table 3. volume elements
There are three types of boundaries considered as fol-
lows in all the simulations: Design Cone 1st cylinder Cone 2nd cylinder Total
angle cells cells cells

A 6 248,820 495,322 281,770 1,025,912


B 10 248,820 296,600 281,770 827,190
C 20 248,820 163,130 281,770 693,720

– Inlet
The uniform inlet velocity of 4.167 m/s, equivalent to
1.5 m3/h flow rate (Reinlet ¼ 27;777), with 5 % of turbulent
intensity is implemented for both inlets.
– Wall
No-slip condition is assumed on the walls. The grid is
fine enough near the lateral wall for the non-dimensional
wall distance (y?) of the first node to be below 1. There-
fore, no wall function is used for the lateral wall.
– Outlet
The gauge static pressure, determined based on the
desired split ratio, is assumed to be fixed at the overflow and
underflow. The implemented pressures at the boundaries for
Fig. 2 Generated mesh for hydrocyclone simulation the three different designs are reported in Table 4. The

123
Heat Mass Transfer

Table 4 Boundary pressure and PDR for three designs Table 5 Stopping distance, initial Reynolds number and time to
travel 95 % of the stopping distance for standard density sphere
Design Overflow Underflow Inlets PDR
Pmean (kPa) Pmean (kPa) Pmean (kPa) Particle Re0 Stopping Time to travel
diameter distance, 95 % of
A 0 0 150 1.000 (lm) V0 = 10 m/s stopping
B -50 0 220 1.227 (mm) distance (s)
C -163 0 327 1.498
1 0.66 0.035 1.1 9 10-5
10 6.6 2.3 8.5 9 10-4
100 66 127 0.065
reference pressure is located at underflow and the pressures
at the other boundaries are measured from underflow. Reference Hinds [43]
The split ratio of the deoiling hydrocyclone is controlled
by Pressure Differential Ratio (PDR) defined as follows:
Pinlet  Poverflow
PDR ¼ ð15Þ
Pinlet  Punderflow
The values of PDR for the three different designs are also
reported in Table 4.

5 Numerical method

An open source C?? code, OpenFOAM is used to solve


the filtered Navier–Stokes equations. The code is devel-
oped based upon finite volume solver with collocated grid.
The Pressure-Implicit Split Operator (PISO) algorithm
handles the linkage between the velocities and pressure
[41]. Preconditioned conjugate gradient solver for pressure
and preconditioned bi-conjugate gradient solver for the

Fig. 5 Comparison of tangential velocity between present LES


results and experimental measurement [12] at Z/D = 2.29 and 6
degree of cone angle

other parameters are used. Unbounded second order linear


central differencing method is employed for convection
and diffusion, and backward scheme is applied for time—
which has second order accuracy.
Simulation is started with k - e turbulence model and
switched to LRR (Launder-Reece-Rodi) Reynolds Stress
Transport Model. After preliminary convergence, the
solution is considered as an initial condition of LES model.
The considered solution procedure helps the convergence
sequence and decreases the computation time. The time
steps are set in the range of 10-6 to 3 9 10-6 in order to
keep the upper limit of the courant number smaller than 0.3
during the convergence process. All the simulations are
performed on two paralleled PCs with Core-i7 processors.
Data sampling begins after 0.5 s of the start of LES sim-
ulations, and continues for 2 s, when the mean values of
the results become constant. The computation time for
reaching a constant mean value in the largest grid number
Fig. 4 An oil droplet path to overflow is around 30 days.

123
Heat Mass Transfer

Fig. 6 Tangential velocity


distribution of continuous phase
inside deoiling hydrocyclone in
design a A, b B and c C

123
Heat Mass Transfer

After prediction of the continuous phase flow field,


Lagrangian particle trajectory is used to calculate the
separation efficiency in dilute oil–water hydrocyclone.
A FORTRAN code is developed to solve an integrated
equation of the force balance on the oil droplets.
The code reads the geometry and the obtained velocity
field of continuous phase, and solves the ordinary differ-
ential transport equations of dispersed phase using 4th
order Runge–Kutta method. The searching process to find
the location of the particles is a highly time consuming
task, especially in large eddy simulation where the number
of cells is significantly high. Therefore, the optimized
Closest Cell (CC) algorithm introduced by Sani and Saidi
[42] is applied as the searching algorithm in order to reduce
the computation time of particle tracking procedures. The
cell which contains oil droplet in each time step is named
host cell, and the flow properties affecting the droplet
trajectory are taken from the host cell. By assuming the old
host cell from the last time step as a potential host cell and
calculating the distance between its cell center and its
neighbors with the target, the new potential host cell can be
selected. This procedure continues till the distance between
the potential host cell and the droplet coordination is less
than its neighbors. Figure 4 presents the path of a sample
particle exiting the hydrocyclone’s overflow in design (A).
Oil droplets are injected into the hydrocyclone from the
both inlets. A subroutine is used to set the location of
injection from inlet faces randomly. The injection and the
tracking process continue for each diameter until a constant
value of efficiency is achieved independent of the particle
numbers or injection position.
Some injected droplets do not leave the hydrocyclone
from overflow or underflow, and are trapped in the hy-
drocyclone. The trapped droplets are not considered for
efficiency estimation. The efficiency is defined as the ratio
of droplets leaving the hydrocyclone from the overflow to
the sum of the droplets exiting either overflow or under-
flow. The efficiency is defined for a particular size of oil
droplets.
Noverflow
Efficiency ¼  100 % ð16Þ
Noverflow þ Nunderflow
Time step of Lagrangian method is less than particle
relaxation time and its value is sufficient to maintain the oil
droplet more than three iterations in each cell.

6 Results and discussion

6.1 Results validation

The obtained tangential velocity is compared with experi- Fig. 7 Tangential velocity profile versus radial position in a top,
mental measurements of Bai et al. [12] and is presented in b middle and c bottom of cone section

123
Heat Mass Transfer

Fig. 8 Axial velocity


distribution of continuous phase
inside deoiling hydrocyclone in
design a A, b B and c C

123
Heat Mass Transfer

Fig. 9 Schematic recirculation


zones in design a A, b B and c C

Fig. 5. There is a good agreement between experimental and bottom of hydrocyclone cone. As outlined before, the
and computed numerical results. The general trend of the higher tangential velocities are associated with design (C).
curve can be predicted in numerical simulations. The The radial variation of axial velocity at different dis-
Rankine vortex can be distinguished with a sharp forced tances from the top wall of hydrocyclone is depicted in
vortex tangential velocity profile in the hydrocyclone’s Fig. 8. The swirling flow inside the deoiling hydrocyclone
core and the free vortex one in the outer region. It is is composed of two vortices. The outer vortex near the wall
notable that due to the anisotropic behavior of the flow, of hydrocyclone moves downward to the underflow while
selecting an appropriate turbulence model is crucial to the inner vortex moves upward toward the overflow. The
come by accurate results. upward velocity near the axis and downward velocity near
There is not any experimental or numerical result the walls are due to the secondary vortex which occurs in
available for the separation efficiency of the studied hy- hydrocyclone. The secondary vortexes arising as a result of
drocyclone, therefore in order to validate the results of radial pressure gradient can be characterized by positive
separation efficiency, the developed Lagrangian Particle upward velocity.
Tracking (LPT) code is verified through injection of par- The maximum axial distance which has positive upward
ticles with different diameters in a stationary fluid and velocity occurs at the positions of 200, 100 and 70 mm
calculating the stopping distances. The results were com- from the top wall of design (A), (B) and (C), respectively.
pared with Hinds [43] and provided in Table 5. The This position is in agreement with the experimental mea-
agreement is observed between the computed results and surements of Bai et al. [12] for design (A). It can be
the reported data. concluded that the recirculation zone is greater in design
(A) in comparison with the other designs. By precise
6.2 Velocity distribution observation into Fig. 8, we notice that there is a second
recirculation zone at the end of the cone for design (C).
Tangential velocity distributions for the three designs are This region is created due to severe pressure gradient at the
presented in Fig. 6. The general Rankine tangential cone end. The schematic of recirculation zones inside
velocity profile can be observed for the three designs. deoiling hydrocyclones is illustrated in Fig. 9 for the three
Although the inlet flow rate is the same for different cases, designs.
design (C) demonstrates the higher tangential velocity. The The detailed comparison between the axial velocities for
peak of velocity is in the order of 8, 12 and 18 (m/s) for 6, the top, middle and bottom of the cone is presented in
10 and 20 degree cone angles, respectively. The radial Fig. 10. It can be noticed in Fig. 10c that the second
position of the peak which specifies the ratio of forced to recirculation zone affects the distribution of the axial
free vortex tangential velocity profile remains nearly con- velocity in the bottom of the cone.
stant in the first cylindrical and the following conical sec-
tions, but increases along the axis in lower cylindrical 6.3 Pressure distribution
section of hydrocyclone.
The detailed comparison of tangential velocities can be The swirling motion creates a pressure gradient across the
seen in Fig. 7. The velocities are reported for top, middle radial position which causes migration of lighter phase

123
Heat Mass Transfer

toward the center. The migration velocity is a function of


density difference between the dispersed and the continu-
ous phase, radial pressure gradient, relaxation time of
droplets and the residence time of flow [15]. The reason
behind the secondary vortex in the hydrocyclones is the
radial pressure variation. The greater is the radial pressure
gradient, the higher separation efficiency would be.
The time averaged static pressure distribution inside
deoiling hydrocyclone is displayed in Fig. 10. The mini-
mum pressure occurs at the hydrocyclone axis. The pres-
sure gradient decreases along the hydrocyclone axis. The
pressure gradient of 290, 360 and 640 (kPa) which corre-
sponds to design (A), (B) and (C), respectively can be seen
in Fig. 11.
The comparison of pressure variation for different radial
line in top, middle and bottom of the cone for three dif-
ferent designs is illustrated in Fig. 12. Larger cone angles
lead to more severe pressure gradients.
As provided in Table 4, the value of PDR for design
(A) is smaller than designs (B) and (C). Therefore, the
pressure drop inside the deoiling hydrocyclones measured
from the inlet to the underflow in design (A) is less than
others. It should be mentioned that the pressure of overflow
should be low enough to remove the oily waste water from
the central core of hydrocyclone.

6.4 Separation efficiency

Since the density of oil droplets is lower than water, they


tend to move toward the center due to the centrifugal force,
while being carried toward the underflow by the continuous
water phase. If droplets reach the recirculation zone before
leaving hydrocyclone, they will be captured and removed
from water. The length of recirculation zone in design
(A) is greater than the two other designs. As a result, it
would be predictable that design (A) bears higher effi-
ciency compared to design (B) and (C).
Figure 13 shows the results of separation efficiency for
the three designs (A), (B) and (C). It can be observed that
the design (A) with a 6-degree cone angle has higher
efficiency than designs (B) and (C). Although the reduction
of cross-sectional area results in higher tangential velocity
and pressure gradient in design (C), and accelerates the
separation process, it reduces the particle residence time in
the hydrocyclone by increasing the axial velocity. Increase
in the tangential velocity is proportional to the first power
of radius reduction, i.e., conservation of angular momen-
tum, while the axial velocity is proportional to the second
power of radius reduction, i.e., conservation of mass.
Therefore, it should be a relation between the desired
Fig. 10 Axial velocity profile versus radial position in a top, b middle tangential velocity and particle residence time to obtain the
and c bottom of cone section maximum separation efficiency.

123
Heat Mass Transfer

Fig. 11 Static pressure


contours in design a A, b B
and c C

7 Conclusion zones, and affects the separation efficiency of the hydro-


cyclone. The recirculation zone in design (A) is larger than
The parametric study of cone section in deoiling hydro- the others. The numerical results reveal that increasing
cyclones is conducted taking advantage of LES approach. cone angle increases the tangential velocity and the pres-
The continuous phase velocity field and pressure distribu- sure gradient in hydrocyclones, yet reduces the oil resi-
tion is calculated for three different cone angles. Good dence time. As a result, the separation efficiency is
agreement is observed between the obtained velocity and decreased in deoiling hydrocyclones with large cone
the available experimental data. The flow structures are angles. Design (A) with a 6-degree cone angle has the
presented for three different designs. The highly swirling maximum separation efficiency in comparison with the
flow in design (C) creates two separated recirculation other cone angles.

123
Heat Mass Transfer

Fig. 13 Separation efficiency of hydrocyclone in different designs

References

1. Thew MT (1986) Hydrocyclone redesign for liquid–liquid sepa-


ration. Chem Eng 427:17–23
2. Caldentey JC (2000) A mechanistic model for liquid hydrocy-
clones (LHC). Dissertation, The University of Tulsa
3. Simkin DJ, Olney RB (1956) Phase separation and mass transfer
in a liquid–liquid cyclone. AIChE 2:545–551
4. Sheng HP, Welker JR, Sliepcevich CM (1974) Liquid–liquid
separation in a conventional hydrocyclone. Can J Chem Eng
52:487–491
5. Colman D, Thew MT (1980) Hydrocyclone to give a highly con-
centrated sample of a lighter dispersed phase. In: International
conference on hydrocyclones, BHRA, Cambridge, UK, pp 209–223
6. Colman D, Thew MT, Corney D (1980) Hydrocyclones for oil/
water separation. In: International conference on hydrocyclones,
BHRA, Cambridge, UK, pp 143–165
7. Colman D, Thew MT (1983) Correlation of separation results
from light dispersion hydrocyclones. Chem Eng Res Des 61:233–
240
8. Meldrum N (1987) Hydrocyclones: a solution to produced water
treatment. In: 19th annual conference on offshore technology,
Houston, SPE 16642
9. Young GAB, Wakley WD, Taggart DL, Andrews SL, Worrell JR
(1994) Oil–water separation using hydrocyclones: an experi-
mental search for optimum dimensions. J Petrol Sci Eng 11:37–
50
10. Belaidi A, Thew MT (2003) The effect of oil and gas content on
the controllability and separation in a de-oiling hydrocyclone.
Chem Eng Res Des 81(3):305–314
11. Husveg T, Rambeau O, Drengstig T, Bilstad T (2007) Perfor-
mance of a deoiling hydrocyclone during variable flow rates.
Miner Eng 20:368–379
12. Bai Z, Wang H, Tu S (2009) Experimental study of flow patterns
in deoiling hydrocyclone. Miner Eng 22(4):319–323
13. Zhou N, Gao Y, An W, Yang M (2010) Investigation of velocity
Fig. 12 Pressure gradient versus radial position in a top, b middle field and oil distribution in an oil–water hydrocyclone using a
and c bottom of cone section particle dynamics analyzer. Chem Eng J 157(1):73–79

123
Heat Mass Transfer

14. Hargreaves JH, Silvester RS (1990) Computational fluid 28. Delgadillo JA, Rajamani RK (2007) Large Eddy simulation
dynamics applied to the analysis of deoiling hydrocyclone per- (LES) of large hydrocyclones. Particul Sci Technol 25:227–245
formance. Chem Eng Res Des 68(4):365–383 29. Smagorinsky J (1963) General circulation experiments with the
15. Wolbert D, Ma BF, Aurelle Y, Seureau J (1995) Efficiency primitive equations. Mon Weather Rev 91:99–165
estimation of liquid–liquid hydrocyclones using trajectory anal- 30. Moin P, Kim J (1982) Numerical investigation of turbulent
ysis. AIChE 41(6):1395–1402 channel flow. J Fluid Mech 118:341–377
16. Grady SA, Wesson GD, Abdullah M, Kalu EE (2003) Prediction 31. Jones W, Wille M (1995) Large eddy simulation of a jet in a cross
of 10-mm hydrocyclone separation efficiency using computa- flow. In: 10th symposium on turbulent shear flows, The Penn
tional fluid dynamics. Filtr Sep 40(9):41–46 State Univ, pp 41–46
17. Petty CA, Parks SM (2004) Flow structures within miniature 32. Germano M, Piomelli U, Moin P, Cabot WH (1991) A dynamic
hydrocyclones. Miner Eng 17(5):615–624 subgrid-scale eddy viscosity model. Phys Fluids A 3:31760–
18. Huang S (2005) Numerical simulation of oil–water hydrocyclone 31765
using Reynolds-stress model for Eulerian multiphase flows. Can J 33. Germano M (1992) Turbulence: the Filtering approach. J Fluid
Chem Eng 83(5):829–834 Mech 238:325–336
19. Noroozi S, Hashemabadi SH (2009) CFD simulation of inlet 34. Lilly DK (1992) A proposed modification of the Germano sub-
design effect on deoiling hydrocyclone separation efficiency. grid scale closure method. Phys Fluids A 4:633–635
Chem Eng Technol 32(12):1885–1893 35. Loth E (2008) Quasi-steady shape and drag of deformable bub-
20. Noroozi S, Hashemabadi SH (2011) CFD analysis of inlet bles and drops. Int J Multiphase Flow 34(6):523–546
chamber body profile effects on de-oiling hydrocyclone effi- 36. Maddahian R (2012) Investigation of two-fluid flow contains oil
ciency. Chem Eng Res Des 89(7):968–977 and water in hydrocyclones. Dissertation, Sharif University of
21. Kharoua N, Khezzar L, Nemouchi Z (2010) Computational fluid Technology
dynamics study of the parameters affecting oil–water hydrocy- 37. Hadamard J (1911) Mouvement Permanent lent d’une Sphere
clone performance. Proc Int Mech Eng E-J Pro 224:119–128 Liquide et Visqueuse dans un Liquide Visqueux. C R Acad Sci
22. Kharoua N, Khezzar L, Nemouchi Z (2010) Hydrocyclones for 152:1735–1743
de-oiling applications—a review. Petrol Sci Technol 28(7):738– 38. Rybczynski W (1911) Über die fortschreitende Bewegung einer
755 flüssigen Kugel in einem zähen Medium. Bull Acad Sci Cracovi
23. Saidi M, Maddahian R, Farhanieh B, Afshin H (2012) Modeling A:40–46
of flow field and separation efficiency of a deoiling hydrocyclone 39. Saboni A, Alexandrova S (2002) Numerical study of the drag on
using large eddy simulation. Int J Miner Process 112–113:84–93 a fluid sphere. AIChE 48(12):2992–2994
24. Slack MD, Prasad RO, Bakker A, Boysan F (2000) Advances in 40. Rivkind VY, Ryskin GM (1976) Flow structure in motion of
cyclone modeling using unstructured grids. Trans Inst Chem Eng spherical drop in a fluid medium at intermediate Reynolds
78A, 1098–1104 numbers. Fluid Dyn 11(1):5–12
25. Delgadillo JA, Rajamani RK (2005) A comparative study of three 41. Issa RI (1986) Solution of the implicitly discretised fluid flow
turbulence closure models for the hydrocyclone problem. Int J equations by operator-splitting. J Comp Phys 62:40–65
Miner Process 77(4):217–230 42. Sani M, Saidi MS (2009) A set of particle locating algorithms not
26. Delgadillo JA, Rajamani RK (2005) Hydrocyclone modeling: requiring face belonging to cell connectivity data. J Comp Phys
large eddy simulation CFD approach. Miner Metall Proc 22(4): 228:7357–7367
225–232 43. Hinds WC (1999) Aerosol technology: properties, behavior and
27. Delgadillo JA, Rajamani RK (2007) Exploration of hydrocyclone measurement of airborne particles. Wiley, New York
designs using computational fluid dynamics. Int J Miner Process
84:252–261

123

View publication stats

You might also like