You are on page 1of 13

165

2014,26(2):165-177
DOI: 10.1016/S1001-6058(14)60019-6

Direct numerical simulation of Open Von Krmn Swirling Flow*


XING Tao
Department of Mechanical Engineering, College of Engineering, University of Idaho, Idaho 83844-0902, USA,
E-mail: xing@uidaho.edu

(Received January 15, 2014, Revised April 10, 2014)

Abstract: Direct numerical simulations are used to investigate the Open Von Krmn Swirling Flow, a new type of unsteady
three-dimensional flow that is formed between two counter-rotating coaxial disks with an axial extraction enclosed by a cylinder
chamber. Solution verification shows that monotonic convergence is achieved on three systematically refined grids for average
pressure at the disk periphery with a small grid uncertainty at 3.5%. Effects of the rotational speeds and flow rates on the flow field
are examined. When the disks are rotating at the lowest speed, 100 RPM, only circular vortices are formed regardless of the flow
rates. When the disks are rotating at 300 RPM and 500 RPM, negative spiral vortex network is formed. The radial counterflow
concept for such spiral vortex network is verified by examining various horizontal cuts and radial velocity component, which show
radial outflows in two bands near the two disks and radial inflow in one band between them. Overall, the flow is similar to the
Stewartson type flow but with significant differences for all three velocity components due to the axial suction at the upper disk
center and gap between the disk periphery and chamber wall.
Key words: direct numerical simulation, Open Von Krmn flow, swirling, radial counterflow

Introduction
Fluid motion between two coaxial disks/plates
has been studied extensively for decades due to their
importance to industrial applications. Applications of
such flows in practice include heat and mass exchangers[1], disk reactor for intensified synthesis of biodiesel[2], open clutch system[3,4], lubrication[5], rotating
packed beds[6], and internal cooling-air systems of
most gas turbines[7], etc. The two disks/plates may corotate, counter-rotate, or one disk is stationary and the
other rotates (rotor-stator system), which creates dramatically different flow patterns.
Limited number of studies used analytical methods, likely due to the strong viscous effect within the
boundary layers near the disk surfaces and strong
three-dimensional features of the flow. Batchelor[8] solved the steady rotationally-symmetric viscous laminar flow between two infinite disks. When the two
disks are exactly counter-rotating, the distribution of
Biography: XING Tao (1973-), Male, Ph. D.,
Assistant Professor

tangential velocity is symmetrical about the mid-plane


and exhibits five distinct regions: two disk boundary
layers, a transition shear layer at mid-plane, and two
rotating cores on either side of the transition layer.
Stewartson[9] draw controversial conclusions on the
flow structure as he found that the flow is divided into
only three zones when the Reynolds number ReH =

H 2 / 100 , where H is the disk spacing, is


the rotational speed of the disk and is the fluid kinematic viscosity. The three regions are two boundary
layers on the two disk surfaces separated by a region
that has zero tangential velocity and uniform radial inflow. The work by Wilson and Schryer[10] numerically
solved the steady viscous flow between two coaxial
infinite disks with one stationary and the other rotating. The effects of applying a uniform suction through the rotating disk are determined. At large Reynolds
numbers, the equilibrium flow approaches an asymptotic state in which thin boundary layers exist near
both disks and an interior core rotates with nearly constant angular velocity. The flow field is assumed to be
axisymmetric. This study also demonstrates that more
than one steady (equilibrium) solution exist for the

166

time-dependent equations of motion. Witkowski et


al.[11] studied the first bifurcation in the axisymmetric
flow between two exactly counter-rotating disks with
very large aspect ratio R / H , where R is the
disk radius and 2 H is the inter-disk spacing. By neglecting the effect of curvature, they were able to reduce the order of Navier-Stokes equations and axisymmetric flow to parallel flow. They found that a centrifugal instability will always occur no matter how
large the local radius considered, which is different
from the plane Couette flow. As a result of the assumptions for axisymmetric flow, infinite disk sizes, and
very large aspect ratios, conclusions drawn by these
analytical studies may not be applicable for most industrial applications.
The first systematic experimental study on the
flow between two rotating co-axial disks at a relatively wide range of rotational conditions was conducted
by Soong et al.[12]. Three different modes of disk rotations, i.e., co-rotation, rotor-stator, and counter-rotation, were considered. When there is no shroud near
the disk rim, co-rotating disk flows are characterized
by an inboard core region of solid-body rotation, outboard vortical flow region, and Ekman layers over
disks with the presence of the vortex chains in gapview of the two-cell flow structure. Flow between
counter-rotating disks encounters large tangential
shear stemming from opposite tangential Coriolis
force near two disks, which enhance the fluid mixing
characteristics. The size of the disk gap plays an important role in formation of the flow structures. In general, smaller gap size reduces the size of the vortices,
weakens the external fluid ingestion in the gap region,
and suppresses the flow instability or turbulence.
Gauthier et al.[13] experimentally investigated the flow
between two rotating disks (aspect ratio of 20.9) enclosed by a cylinder in the cases of both co- and counter-rotation. It was found that the co-rotation case
and the weak counter-rotation cases are very similar to
the rotor-stator case in that the basic flow consists of
two boundary layers near each disk and the instability
patterns are the axisymmetric vortices and the positive
spirals. When the two disks are counter-rotating with a
higher rotation ratio, a new kind of instability pattern
is observed, called negative spirals. The recirculation
flow becomes organized into a two-cell structure with
the appearance of a stagnation circle on the slower
disk. Moisy et al.[14] conducted experimental and numerical study of the shear layer instability for the
same geometry but with various aspect ratios between
2 and 21. It was shown that the instability can be described in terms of a classical Kelvin-Helmholtz instability, where curvature has only a weak effect. The observed surrounding spiral arms result from the interaction of this unstable shear layer with the Ekman

boundary layers over the faster disk. Poncet et al.[15,16]


used computational fluid dynamics (CFD) to investigate the turbulent Von Krmn flow generated by two
counter-rotating smooth flat (viscous stirring) or bladed (inertial stirring) disks enclosed by a cylinder. For
viscous stirring, the flow close to the rotation axis is
of Stewartson type and shows three distinct regions:
two boundary layers and one shear layer at mid-plane.
For regions close to the periphery of the cavity, flow
is of Batchelor type with five distinct zones: two
boundary layers on the disks, a shear layer at midplane and two zones enclosed between the two.
Few studies investigated two-phase flows between two rotating coaxial disks. Yuan et al.[17] studied
aeration for disengaged wet clutches using a gas-liquid two-phase CFD model with experimental measurements of drag torque for validation. When the separator disk is stationary, air enters the clearance from
the outer periphery of it and oil flies off from the rotating disk, which reduce the drag torque. When the two
disks are counter-rotating, the two-phase flow pattern
depends on the difference between the two angular velocities. If the difference is large, air enters from the
low speed side of the plates. Otherwise, air enters
from the middle of the clearance at the outer radius
and both sides keep a thin oil film.
The objective of this study is to investigate a new
type of flow, Open Von Krmn Swirling Flow,
which features radial counterflow between two counterrotating disks enclosed in a chamber. It differs
from the well-known Von Krmn Flow as it has an
axial suction (outlet) at the center of the upper disk
and a gap between the two disks and chamber wall.
This serves as a simplified model of the McCutchen
Processors developed by Vorsana Inc. that centrifugally separates a fluid mixture using vortices created in
high shear between axially fed counter-rotating disk
impellers. The processor makes use of the radial counterflow concept. As the fluid mixture is spun at high
speed in the vortices, centrifugal force moves heavy
fractions toward the outside of the vortices and away
from the axis of rotation, while light fractions remain
inside the vortices, and are sucked inward by an axial
pump. This radial counterflow concept has not been
experimentally or numerically verified in previous
works. This study uses CFD to examine this concept
and other flow physics for Open Von Krmn
Swirling Flow. Parametric studies are performed to
elucidate the effect of flow rates and rotational speeds
on the formation of vortical structures. Quantitative
solution verification is performed on three systematically refined grids to estimate the grid uncertainties.
Validation of the CFD model is achieved by comparing with available experimental data for similar geometries and flow conditions.

167

1. Computational methods
The commercial CFD software, ANSYS/
FLUENT version 14.0[18] is used for all the simulations. FLUENT is a finite volume solver which provides a suite of numerical schemes and transition and
turbulence modeling options. For this study, transient
single-phase simulations are conducted using the pressure-based solver option, which is the typical predictor-corrector method with solution of pressure via a
pressure Poisson equation to enforce mass conservation. Pressure-velocity coupling is performed using
the SIMPLE scheme. The convective terms in the momentum equation are discretized using the third-order
MUSL scheme. Unsteady terms are discretized using
a second order implicit scheme. The time step is chosen to be 0.005 s for all simulations with large maximum iterations per time step to ensure that the minimum residuals are lower than 105 for the continuity
and three momentum equations for all simulations.
Due to the small disk radius and low rotational speed
of the disks, no turbulence model is applied. All simulations are conducted using ANSYS Academic Research CFD with high performance computing on a Dell
Precision T7500 that has 12 cores and 48 GB RAM.

1.1 Governing equations


Since direct numerical simulations are performed,
no turbulence models are used. For Cartesian coordinates, the incompressible continuity and momentum
equations are:
( V ) = 0

(1)

( V ) + ( V 2 ) = p + [ (V + V T )] + g
t
(2)
where V is the velocity vector, is the dynamic viscosity, is the density, p is the pressure, and g is
the gravitational acceleration.

enclosed by a stationary cylindrical chamber. The


upper disk rotates counter-clockwise viewed from the
top and the lower disk rotates in the opposite direction
at the same rotational speed . The disk diameter is
0.28 m. The gap size between the two disks is
0.0033 m. The chamber diameter is 0.32 m. The two
disks are away from the chamber wall by 0.0133 m in
the vertical direction ( Z ) . Fluid enters the chamber
through a circular hole with diameter 0.01965 m at the
bottom surface of the chamber. Two additional holes
with the same diameter are drilled on the upper disk
and top chamber wall that are connected by a short
circular pipe, which serves as the outlet of the fluid.
An O-type grid is created to model the flow. The fine
grid has a total of 798 675 points. For solution verification, additional two coarser grids are created systematically using a constant grid refinement ratio 2
in all three spatial directions.
1.3 Simulation design and flow parameters
A total of nine simulations are performed as summarized in Table 1. The simulations cover three flow
rates (48 GPM, 72 GPM and 96 GPM) and three rotational speeds (100 RPM, 300 RPM and 500 RPM).
Velocity inlet and pressure outlet are specified as
boundary conditions for the fluid. Rotational wall
boundaries are enforced using the prescribed rotational speeds. The fluid is water liquid with density
998.2 kg/m3 and kinematic viscosity 0.001003 kg/ms.
To facilitate the discussion and also generalize
the conclusions such that they are independent of specific geometry and flow properties, non-dimensional
parameters are used. The rotational Reynolds number
is Re R02 / , where R0 is the outer radius of the
disks and is the kinematic viscosity of the fluid.
The aspect ratio is defined as the ratio of disk spacing
S and R0 , G = S / R0 . For the cylindrical coordinates, the non-dimensional radial and axial coordinates
are defined as r = r / R0 and Z = Z / R0 , respectively. Thus, r = 0 is the rotational axis located at the
center of the disks and r = 1 is located at the disk
periphery. The upper disk is located at Z = 0 and the
lower disk is located at Z = 0.023. For the Cartesian
coordinates, Z is the same as in cylindrical coordinates and X and Y are non-dimensionalized using R0 .
Velocities are non-dimensionalized using V * = V /
( R0 ) , where V can be u , v , w in the Cartesian

Fig.1 Geometry and grid

1.2 Geometry and grid


The geometry is shown in Fig.1. Two co-axial
disks are counter-rotating at the same angular velocity

Coordinates or ur , u , and u z in the cylindrical coordinates. For simplicity, asterisks for all dimensionless
variables are dropped and units for dimensional variables are specified.

168

Table 1 Simulation matrix


Simulations

Q (GPM)

48

72

96

48

72

96

48

72

96

(RPM)

100

100

100

300

300

300

500

500

500

Re

3.25104

3.25104

3.25104

9.75104

9.75104

9.75104

1.63105

1.63105

1.63105

Fig.2 Gap-view flow structures using streamlines and pressure contour near the disk rim for Re = 7.9103

2. Validation of the CFD model


The CFD model is first validated against experimental data reported in previous literature for flows
between two counter-rotating disks without an axial
suction, either qualitatively for the flow pattern or
quantitatively for the disk drag torque.

2.1 Flow pattern between two counter-rotating disks


This validation case follows Soong et al.[12]. The
disk peripheries are open to atmosphere. A 100 (in
gap direction)600 (in radial direction) grid is used.
Simulation is conducted for Re = 7.9103. Two as-

pect ratios are examined. As shown by Fig.2, the fluid


structures inside the gap are very sensitive to the gap
spacing when it is reduced to a certain value. In this
study, G = 0.08 (Fig.2(a)) creates staggered vortex
chains and a wavy interface. The vortices close to the
upper disk are rotating clockwise whereas those close
to the lower disk are rotating counter-clockwise. This
is the same as what observed in experiment (Fig.2(b)).
When the gap size increases to 0.1, the vortex chain
suddenly disappears (Fig.2(c)) and further increase of
the rotational speed will not change the flow pattern.

169

Table 2 Validation of drag torque for a disengaged wet clutch pack


Friction plate speed (RPM)

Drag torque on fixed separator plate (Nm)


Experiment ( D )

CFD ( E % D)

Yuan et al.[17] ( E % D)

200

0.3295

0.3583 (8.7%)

0.3798 (15.3%)

316

0.5596

0.5634 (0.7%)

0.5682 (1.5%)

368

0.6402

0.6578 (2.7%)

0.6549 (2.3%)

2.2 Drag torque


Accurate prediction of the drag torque using the
CFD model is important to estimate the power consumption. The drag torque also indicates the accuracy of
the CFD model to predict the shear stresses near the
plate surfaces. The drag torque predicted by the current CFD model is validated for a disengaged wet clutch
pack. The simulation is compared to the experimental
data and CFD simulation by Yuan et al.[17]. To ensure
that the CFD results are independent of the grid resolution, a very fine grid (800 in radial100 in axial =
80 000 nodes) is used. This is about nine times finer
than the grid used by Yuan et al.[17] in their CFD
simulations, which had 300 in radial30 in axial =
9 000 nodes. The CFD model is built in axisymmetric
flow with swirl.
The experiment measures the drag torque on the
fixed separator plate when the friction plate is rotating
at different speeds. It was shown that at low rotational
speed, the drag torque increases almost linearly versus
speed to a peak value (Phase I). After the peak value
that corresponds to a critical friction plate speed, the
torque is reduced rapidly to nearly zero (Phase II). By
examining the flow field, it was found that Phase I
shows single-phase flow whereas Phase II shows twophase flow between the two plates. In other words, air
starts to enter the clearance at the critical speed and
the aeration causes the oil film to shrink. The greater
the speed, the more the air enters the clearance. Because the air viscosity is much smaller than oil viscosity, the drag torque rapidly decreases. Since this
study only investigates single-phase flow, validation is
conducted for three rotational speeds in Phase I and
results are summarized in Table 2. Compared to the
CFD by Yuan et al.[17], the current CFD has similar relative error for the 368 RPM but much lower relative
error for 200 RPM and 316 RPM. Since the CFD
solver used is the same, this improvement was likely
due to the much finer grid used in the current study.

1 850

1 881

2 303

R
0.07

P
2.51

3.1 Solution verification


Solution verification is a process for assessing
simulation numerical errors and associated uncertainties. In this study, the discretization error due to limited number of grid points is the main source of numerical errors. In this study, solution verification is performed for the average pressure at disk periphery on
three systematically refined grids that are generated
using a constant grid refinement ratio r = 2 in all
three spatial directions. The factor of safety method[19,20] is used to estimate the grid uncertainties and
results are summarized in Table 3. The fine grid
(mesh 1) has 770 788 grid points. Meshes 2 and 3 represent the medium and coarse grids, respectively.
Simulation 9 that has the highest disk rotational speed
and largest inlet flow rate is selected for solution verification. The solutions on the fine, medium, and coarse grids are S1 , S 2 and S3 , respectively. Solution
changes for medium-fine and coarse-medium solutions and the convergence ratio R are defined by

21 = S2 S1 , 32 = S3 S2 , R =

21
32

(3)

When 0 R 1 , monotonic convergence is achieved.


Then the three grid solutions can be used to compute
the estimated order of accuracy pRE , error RE , and
grid uncertainty U G (% S1 ) .

pRE


ln 32

= 21
ln(r )

RE =

21
pRE

(4)

(5)

U G (% S1 )

When solutions are in the asymptotic range, then


pRE = pth , however, in many circumstances, especially for industrial applications, solutions are far from
the asymptotic range such that pRE is greater or sma-

3.5%

ller than pth [21]. The ratio of pRE to pth is used as the
distance metric

Table 3 Solution verification for average pressure (Pa) at


disk periphery
Mesh number

3. Results and discussion

170

Fig.3 Three-dimensional vortical structures (Iso-surface of Q = 200 is colored by pressure in Pa) of single-phase water between the
two counter-rotating disks view from the top (vortices above the upper disk and below the lower disk have been blanked out):
(a)-(i) correspond to Simulations 1 to 9 in Table 1, respectively, and (j) averaged pressure of fluids at disk periphery for the
nine simulations

171

Fig.4 Streamlines and contour of the velocity component v in the slice at Y = 0 for Simulation 5 (length ratio of X and Z is
0.25)

P=

pRE
pth

(6)

and the grid uncertainty is estimated by


U G = (2.45 0.85 P ) RE , 0 P 1

(7a)

U G = (16.4 P 14.8 P ) RE , P 1

(7b)

As shown in Table 3, monotonic convergence is achieved with a low grid uncertainty of 3.5%S1 . This
suggests that the current fine grid resolution is sufficient and this fine grid is used for all simulations.
3.2 Flow physics
Three-dimensional top view of the vortical structures within the processor is shown in Fig.3 for the
nine simulations in Table 1. The vortical structures are
identified by the Q -criterion[22] and colored by pressure. To focus on the flow between the two disks, the
vortical structures above the upper disk and below the
lower disk have been blanked out.
For all the nine simulations, the highest and lowest pressures are located near the chamber wall and
axial suction (outlet), respectively. For the same rotational speed, the range of pressure values increases
with the increase of the inlet flow rate. When the disks
are rotating at the lowest speed, 100 RPM
(Figs.3(a)-3(c)), only circular vortices are formed regardless of the flow rates. With the increase of the inlet flow rate, more circular vortices move toward the
axial suction. For the two higher rotational speeds
( 300 RPM and 500 RPM ), negative spiral vortex
network is formed, which is similar to what was observed in the experiments by Gauthier et al.[13]. It also
shows for these two higher rotational speeds that increase of the flow rate creates larger size vortices but
the number of vortices decreases near the disk center.
To examine quantitatively the effect of flow rates

at the three different rotational speeds, average pressures of fluids at disk periphery are plot for the nine
simulations, as shown in Fig.3(j). Overall, the pressure
increases almost linearly with the increase of rotational speed for the two lower rotational speeds
100 RPM, 300 RPM and 300 RPM shows a larger
slope. For rotational speed 500 RPM, the pressure increases non-linearly (quadratically) as the increase of
flow rates.
Figure 4 shows streamlines and contour of the
velocity component v in the slice at Y = 0 for Simulation 5. In order to clearly show the flow field, the
length ratio of X and Z has been reduced from 1 to
0.25. Two vortex streets staggered to each other are
formed near the upper and lower disk surfaces, respectively, which is similar to the flow pattern shown in
Fig.2(a) for the study by Soong et al.[12]. The upper
vortices are rotating counter-clockwise whereas the
lower vortices are rotating clockwise. This results in a
shear layer between the two vortex streets where fluid
flows from the disk periphery to the center of the disk.
Fluids very close to the two disk surfaces are swept
out by the rotating of the disks, regardless of their rotating direction. However, due to the opposite rotational directions of the two disks, the upper and lower
vortex streets show negative and positive v velocities,
respectively, which is consistent with the rotational direction of the adjacent disk.
Figure 5 shows different views by examining
flows in various Z cuts for Simulation 5. All the subfigures in Fig.5 are colored by the Z velocity. By
comparing with Fig.3(e) and streamlines in Fig.4, the
interface between the positive and negative Z velocities in Fig.5 is corresponding to the local core of the
spiral vortices. For fluid inside the boundary layer of
the upper disk as shown in Fig.5(a), it has two velocity components. The first component is caused by the
local disk rotation and no-slip boundary condition enforced on the disk surface, r , where r is the radius
of the local point on the disk surface with respect to
the Z axis. The other component is the velocity in the

172

Fig.5 Streamlines and contour of Z velocity component in various Z cuts (Simulation 5): (a) inside the boundary layer of the
upper disk ( Z = 1 108 ) , (b) inside the boundary layer of the lower disk ( Z = 0.99) , (c) plane cross the upper vortex
street center ( Z = 0.22) , (d) plane cut through the lower vortex street center ( Z = 0.61) , and (e) plane cut through the
shear layer between the two vortex streets ( Z = 0.45)

radial direction caused by the centrifugal force. As a


result, fluid flows radially outward from the rotational
Z axis following negative spiral paths. However,
there is a small circular region with diameter 0.048 m
near the center where fluid flows toward the rotational
axis through a positive spiral. This is called a spiral
eye that is larger than the outlet diameter 0.01965 m,
which is caused by the strong suction at the outlet located at the center of the upper disk. Figure 5(b)

shows the flow inside the boundary layer of the lower


disk. Similar to the flow inside the boundary layer of
the upper disk, the fluid flows radially outward from
the rotational axis. But it follows positive spiral paths
as it rotates in the opposite direction to the upper disk.
There is also a spiral eye near the center with diameter r0 = 0.034 m, which is smaller than observed
inside the upper disk boundary layer. Inside the spiral
eye, fluid flows towards the Z axis through a nega-

173

tive spiral. For the plane cutting through the upper


vortex street center (Fig.5(c)), the streamlines are very
curvy with overall flow direction from the disk periphery to the center. The spiral vortex network is clearly shown and agrees well with the vortical structures
observed in Fig.3(e). The flow patterns in the plane
cutting through the lower vortex street center (Fig.5(d))
and plan across the shear layer between the two vortex
streets (Fig.5(e)) are similar to the flow pattern in the
plane cut across the upper vortex street center but with
a much smoother negative spiral.

Fig.6 Conversion of velocity vectors from Cartesian coordinate


to cylindrical polar coordinate

To better examine the radial counterflow concept,


various annular control surfaces are extracted within
the flow field. These annual control surfaces are crosssections of the flow between the disks at a constant radius from the rotational Z axis. To facilitate the analysis, the three Cartesian velocity components (u , v,
w) are converted to be the components in the cylindrical coordinates (ur , u , u z ) using
ur = u cos + v sin , u = u sin + v cos ,
u z = w , tan =

y
x

(8)

The correlation between these velocity components is shown in Fig.6. Figure 7 shows instantaneous
pressure (contour flood) and ur (contour line) in three
annular control surfaces for X 0 that are projected
to the vertical (Y , Z ) plane ( X 0 exhibits similar
features and thus not shown). There are two bands
with positive ur that are located inside boundary
layers of the two disks. This suggests that the net momentum for fluid close to the disk surface is radially
outward. Between the two bands, there is a shear layer
where ur is negative and the net momentum is radially inward. As the annual control surface move closer

to the rotational Z axis (smaller r ), the size of the


shear layer band increases and the band near the upper
disk is significantly suppressed. This is due to the
axial suction at the center of the upper disk. The largest negative radial velocity is located near the midplane between the two disks. Overall the magnitude of
the radial velocity in the shear layer band decreases
when r decreases, likely due to the increase of the
shear layer band size. The pressure variation in the
vertical Z direction is minor. Low and high pressures
are corresponding to the largest negative radial velocity regions and regions between them, respectively,
which can be explained using the Bernoulli effect.
The interfaces between the three bands can be
visualized using Iso-surface of the radial velocity
ur = 0, where the radial velocity changes direction.
Figures 8(a) and 8(b) show the instantaneous and time
averaged Iso-surface of ur = 0, respectively. Overall
the band near the lower disk is much thicker than the
band near the upper disk. The arrows show the flow
direction for each band. The time-averaging process
smooths the wavy interface observed for the instantaneous flow field. The averaged interface between
the two lower bands shows a smooth circle whereas
the averaged interface between the two upper bands
shows a shape similar to a volcano due to the axial
suction near the upper disk center.
Ravelet et al.[23] found that the structure of the
mean Von Krmn flow in the exact counter-rotating
regime can be decomposed into two poloidal recirculations in the (r , z ) plane. Similar flow pattern is observed in the current study as shown in Fig.9. The
fluid near the upper and lower disks are moved by two
opposite rotation speeds (u ) , and then swiped radially outward. As a result of mass conservation, a shear
layer develops between the two disks with radially inward velocity. However, unlike the Von Krmn flow
where the shear layer is located in the equatorial plane
(mid-plane between the two disks), the shear layer in
this study is closer to the upper disk and moves farther
away from the lower disk when flow approaches the
Z axis. The difference is caused by the axial suction
at the center of the upper disk. Axial profiles of the
tangential velocity component u of the mean flow at
four radial locations are show in Fig.10(a). The tangential velocity is small except inside the boundary layers of the upper and lower disks. When r increases,
the magnitude of the tangential velocity increases inside the two disk boundary layers but remains almost
constant near the shear layer at Z 0.0075. Similar
to the Stewartson flow structure observed by Poncet et
al.[16] for Von Krmn flow when r is small, three
zones are observed: an almost constant tangential velocity zone enclosed by two boundary layers on each
disk. However, the tangential velocity constant is

174

Fig.7 Instantaneous pressure (flood, Pa) and ur (line) in various annular control surfaces for X 0 that are projected to the
vertical (Y , Z ) plane (Simulation 5)

negative, not zero as observed for Von Krmn flow.


This indicates that most regions of the flow are impacted more by the lower disk than by the upper disk,
which is consistent with the much thicker lower bands
shown in Fig.8. The boundary layers of the disks are
also much thicker than those observed by Poncet et al.
who examined turbulent Von Krmn flows that have
much larger Re . While the Von Krmn flow shows
almost zero radial and axial velocity components, the
current open Von Krmn swirling flow exhibits significant magnitude of ur and weak u z (still non-zero),
as shown in Fig.10(b) and Fig.10(c), respectively. The
radial velocity component is about 25% of the magnitude of the tangential velocity component and reaches

maximum positive value and maximum negative


value inside the boundary layers of the disks and near
the shear layer between them, respectively. When r
increases, the magnitudes of both maximum positive
and maximum negative values also increase. The axial
velocity component is the smallest among the three
velocity components. It is zero on the two disk surfaces and in the region near the shear layer. It is positive
and negative in the regions between the upper disk
and the shear layer and between the shear layer and
lower disk, respectively.
4. Conclusions and future work
For the first time, unsteady three-dimensional

175

Fig.8 Iso-surface of ur = 0 that separates the three annular bands for Simulation 5 (length ratio of X and Z is 0.05)

Fig.9 Streamlines of the mean flow between the two disks colored by u for Simulation 5

direct numerical simulations are conducted to investigate flows between two counter-rotating coaxial disks
with an axial extraction enclosed by a cylinder chamber, which is called the Open Von Krmn
Swirling Flow. The CFD model is built on top of the
commercial CFD software, ANSYS FLUENT 14.0,
and validated by comparing against experimental data
published in previous literatures, either qualitatively
for the flow pattern or quantitatively for the drag torque. Quantitative solution verification is performed on
three systematically refined grids. Monotonic convergence is achieved for the average pressure at disk periphery with a small grid uncertainty at 3.5%. The fine
grid is then used for all the nine simulations that cover
three rotational speeds (100 RPM, 300 RPM, and 500
RPM) and three flow rates (48 GPM, 72 GPM, and 96
GPM).
This study reveals strong three-dimensional flow
structures, which undermines the use of axisymmetric
model with a two-dimensional grid to approximate the

flow field in most previous studies for similar geometry and flow conditions. The highest and lowest pressures are located near the chamber wall and axial suction, respectively. For the same rotational speed, the
range of pressure values increases with the increase of
the inlet flow rate. When the disks are rotating at the
lowest speed, 100 RPM , only circular vortices are
formed regardless of the flow rates. For the two higher
rotational speeds ( 300 RPM and 500 RPM ), negative spiral vortex network is formed. The slice cutting
through the spiral vortices at Y = 0 and X 0 shows
two staggered vortex streets that rotates counter-clockwise and clockwise near the upper and lower disks,
respectively.
The radial counterflow concept is verified by
examining various Z cuts and radial velocity component ur in the cylindrical coordinate. Two bands
with positive ur are located in regions very close to
the two disk surfaces where the net momentum of

176

fluid is radially outward. Between the upper and lower


bands, there is a shear layer where ur is negative and
the net momentum is radially inward. Overall the
lower band near the lower disk is much thicker than
the upper band near the upper disk. As the location
moves closer to the rotational Z axis, the size of the
shear layer band increases and the upper band is significantly suppressed. This is due to the axial suction at
the center of the upper disk. No significant change of
the lower band thickness is observed. Further analysis
of the two poloidal recirculations and the three velocity components in the (r , z ) plane show features similar to Stewartson flow but with significant differences
on the location of the shear layer and non-zero radial
and axial velocity components.

Future work includes extension of the current


geometry from model-scale to full-scale and validate
CFD simulations using full-scale experimental data
upon available. The smooth flat disks (viscous stirring)
may be replaced by bladed disks (inertial stirring) to
increase the efficiency of the disks in forcing the flow.
The current single phase simulations need to be extended to two- and multi-phase simulations to investigate the effect of the spiral vortex network on separation
of various phases. Preliminary results of the air-water
mixture flows show that the lighter-phase air tends to
be locked in the spiral vortex cores.
Acknowledgement
The author deeply appreciates the sponsorship
from Vorsana Inc. on this research.
References
[1]
[2]
[3]
[4]

[5]
[6]

[7]
[8]

[9]

[10]
[11]

[12]
Fig.10 Axial profiles of the three velocity components of the
mean flow in Y = 0 at four radial locations for Simulation 5

MARUOKA Y., BRAUER H. Fluid dynamics and mass


transfer in a multistage rotating-disk reactor[J]. International Chemical Engineering, 1989, 29(4): 577-615.
QIU Z., PETERA J. and WEATHERLEY L. R. Biodiesel synthesis in an intensified spinning disk reactor[J].
Chemical Engineering Journal, 2012, 210: 597-609.
APHALE C. R., CHO J. and SCHULTZ W. W. et al.
Modeling and parametric study of torque in open clutch
plates[J]. Journal of Tribology, 2006, 128(2): 422-430.
APHALE C. R., SCHULTZ W. W. and CECCIO S. L.
The influence of grooves on the fully wetted and aerated flow between open clutch plates[J]. Journal of Tribology, 2010, 132(1): 1-7.
HUGHES W. F. and ELCO R. A. Magnetohydrodynamic lubrication flow between parallel rotating disks[J].
Journal of Fluid Mechanics, 1962, 13(1): 21-32.
LLERENA-CHAVEZ H., LARACHI F. Analysis of
flow in rotating packed beds via CFD simulations-dry
pressure drop and gas flow maldistribution[J]. Chemical Engineering Science, 2009, 64(9): 2113-2126.
KILIC M., OWEN J. M. Computation of flow between
two disks rotating at different speeds[J]. Journal of
Turbomachinery, 2003, 125(2): 394-400.
BATCHELOR G. K. Note on a class of solutions of the
Navier-Stokes equations representing steady rotationally-symmetric flow[J]. Quarterly Journal of Mechanics and Applied Mathematics, 1951, 4(1): 29-41.
STEWARTSON K. On the flow between two rotating
coaxial disks[J]. Mathematical Proceedings of the
Cambridge Philosophical Society, 1953, 49(2): 333341.
WILSON L. O., SCHRYER N. L. Flow between a stationary and a rotating disk with suction[J]. Journal of
Fluid Mechanics, 1978, 85(3): 479-496.
WITKOWSKI L. M., DELBENDE I. and WALKER J.
S. et al. Axisymmetric stability of the flow between two
exactly counter-rotating disks with large aspect ratio[J].
Journal of Fluid Mechanics, 2006, 546: 193-202.
SOONG C. Y., WU C. C. and LIU T. P. Flow structure
between two co-axial disks rotating independently[J].
Experimental Thermal and Fluid Science, 2003,
27(3): 295-311.

177

[13]
[14]

[15]

[16]

[17]

GAUTHIER G., GONDRET P. and MOISY F. et al. Instabilities in the flow between co- and counter-rotating
disks[J]. Journal of Fluid Mechanics, 2002, 473: 1-21.
MOISY F., DOARE O. and PASUTTO T. et al. Experimental and numerical study of the shear layer instability
between two counter-rotating disks[J]. Journal of Fluid
Mechanics, 2004, 507: 175-202.
PONCET S., SCHIESTEL R. and MONCHAUX R.
Turbulent Von Krmn flow between two counter-rotating disks[C]. Proceedings of the 8th International
Symposium on Experimental and Computational
Aerothermodynamics of Internal Flows. Lyon,
France, 2007, 1-10.
PONCET S., SCHIESTEL R. and MONCHAUX R.
Turbulence modeling of the Von Krmn flow: Viscous
and inertial stirrings[J]. International Journal of Heat
and Fluid Flow, 2008, 29(1): 62-74.
YUAN S., GUO K. and HU J. et al. Study on aeration
for disengaged wet clutches using a two-phase flow
model[J]. Journal of Fluids Engineering, 2010,
132(11): 111304.

[18]
[19]
[20]

[21]

[22]

[23]

FLUENT USER GUIDE V14.0.0. ANSYS[S], 2011.


XING T., STERN F. Factors of safety for Richardson
extrapolation[J]. Journal of Fluids Engineering, 2010,
132(6): 061403.
XING T., STERN F. Closure to Discussion of Factors
of Safety for Richardson Extrapolation (2011, J.
Fluids Eng., 133, 115501)[J]. Journal of Fluids Engineering, 2011, 133(11): 115502.
XING T., CARRICA P. and STERN F. Computational
towing tank procedures for single run curves of resistance and propulsion[J]. Journal of Fluids Engineering,
2008, 130(10): 101102.
HUNT J. C. R., WRAY A. A. and MOIN P. Eddies, stream, and convergence zones in turbulent flows[R]. Report CTR-S88, Stanford NASA Center for Turbulence
Research, 1988.
RAVELET F., CHIFFAUDEL A. and DAVIAUD F. et
al. Toward an experimental Von Krmn dynamo: Numerical studies for an optimized design[J]. Physics of
Fluids, 2005, 17(11): 1-17.

You might also like