Download as pdf or txt
Download as pdf or txt
You are on page 1of 29

Computers & Fluids 32 (2003) 1405–1433

www.elsevier.com/locate/compfluid

Surface effects on transient three-dimensional flows


around rotating spheres at moderate Reynolds numbers
H. Niazmand, M. Renksizbulut *

Department of Mechanical Engineering, University of Waterloo, Waterloo, Ont., Canada N2L 3G1
Received 27 December 2001; accepted 13 November 2002

Abstract
Transient wake flow patterns and dynamic forces acting on a rotating spherical particle with non-uni-
form surface blowing are studied numerically for Reynolds numbers up to 300 and dimensionless angular
velocities up to X ¼ 1. This range of Reynolds numbers includes the three distinct wake regimes i.e., the
steady axisymmetric, the steady non-symmetrical and the unsteady with vortex shedding. The Navier–
Stokes equations for an incompressible viscous flow are solved by a finite volume method in a three-
dimensional, time accurate manner. An interesting feature associated with particle rotation and surface
blowing is that they can affect the near wake structure in such a way that unsteady three-dimensional wake
flow with vortex shedding develops at lower Reynolds numbers as compared to flow over a solid sphere in
the absence of these effects and thus, vortex shedding occurs even at Re ¼ 200. Global properties, such as
the lift and drag coefficients, and the Strouhal number are also significantly affected. It is shown that the
present data for the average lift and drag coefficients correlate well with:
CL =ð1 þ XÞ3:6 ¼ 0:11
CD ð1 þ 20VS Þ0:2 =ð1 þ XÞRe=1000 ¼ 24ð1 þ Re2=3 =6Þ=Re
where VS is the average surface blowing velocity normalized by the free stream velocity.
Ó 2003 Elsevier Science Ltd. All rights reserved.

Keywords: Rotating sphere; Unsteady wake; Vortex shedding; Surface blowing

1. Introduction

The three-dimensional flow around an isolated spherical particle moving steadily in an


incompressible flow can be considered as a simplified case of the larger family of immersed

*
Corresponding author. Tel.: +519-888-4567x3977; fax: +519-888-6197.
E-mail address: metin@uwaterloo.ca (M. Renksizbulut).

0045-7930/03/$ - see front matter Ó 2003 Elsevier Science Ltd. All rights reserved.
doi:10.1016/S0045-7930(02)00115-9
1406 H. Niazmand, M. Renksizbulut / Computers & Fluids 32 (2003) 1405–1433

bluff-body flows encountered in many engineering applications. Despite the symmetry of the
body, instabilities are known to generate a complex three-dimensional unsteady flow at suffi-
ciently large Reynolds numbers.
The most complex feature of the flow around a sphere is its wake and one can identify three
distinct wake regimes for particle Reynolds numbers of up to 300. It is well known that flow past a
solid sphere separates at a Reynolds number of Re ffi 20 (based on diameter) and forms a steady
axisymmetric wake behind the body. With increasing Re, this attached recirculating zone grows in
size but remains essentially steady and axisymmetric up to Re ffi 212. At higher Reynolds num-
bers, the wake becomes unstable to infinitesimal perturbations and becomes non-symmetrical, yet
remains steady. In contrast to flow over a circular cylinder, which becomes unsteady upon the loss
of symmetry, this non-symmetrical wake structure remains steady and attached to the sphere.
At Reynolds numbers beyond 270, the flow behind the sphere undergoes yet another transition
and develops into a fully three-dimensional, time dependent wake with periodic vortex shed-
ding.
The nature of these transitions and corresponding wake regimes can have important conse-
quences in a number of practical applications and has been studied by a number of researchers
using three different methodologies: Flow visualization techniques (e.g., [1–10]), numerical in-
stability analysis using perturbation techniques (e.g., [11,12]), and computational approaches that
solve the full Navier–Stokes equations and have the capability to mimic the physical instabilities
and capture the transition between the different wake regimes (e.g., [13–17]). Since it is known that
instabilities are responsible for the generation of different wake regimes, it is expected that effects
such as surface blowing and particle rotation would significantly change the flow structure around
a sphere and affect the aforementioned transitional Reynolds numbers.
A review of the relevant literature reveals that, despite the fact that the different wake regimes
and the associated transitional Reynolds numbers were clearly identified for flow over a solid
sphere at Reynolds numbers up to 300, the effects of particle rotation and surface blowing on
these wake regimes have not been examined in detail. Studies concerned with the effects of surface
blowing are limited to axisymmetric cases; see Chuchottaworn [18] and Cliffe and Lever [19]
among others. On the other hand, specific problems involving rotating spheres in uniform flow
have been the subject of several investigations as will be discussed later. However, they are either
related to the lower range of Reynolds numbers in the axisymmetric wake regime, or to high
Reynolds numbers beyond the range of our interest. Few studies, that cover the same range of
Reynolds numbers as the present study, are basically concerned with the overall average flow
parameters such as lift and drag and thus the transient effects of particle rotation on the wake flow
have been ignored. Furthermore, none of these studies have addressed the effects of surface
blowing in combination with particle rotation.
The purpose of the present work is to investigate computationally the effects of surface blowing
on the three-dimensional flow around a rotating spherical particle translating in an incompressible
viscous fluid. The Navier–Stokes equations are solved using a finite volume formulation based on
a pressure correction procedure in a generalized curvilinear coordinate system and in a time ac-
curate manner. The transient effects of non-uniform surface blowing (simulating droplet vapor-
ization) on the wake structure, and on the lift and drag will be investigated for different angular
velocities at Reynolds numbers up to 300. In particular, the effects of surface blowing and particle
rotation on the transitional Reynolds numbers between the different wake regimes will be studied
H. Niazmand, M. Renksizbulut / Computers & Fluids 32 (2003) 1405–1433 1407

and it will be shown that these effects can lead to the development of a time dependent three-
dimensional periodic wake with vortex shedding at particle Reynolds number as low as 200.
Particle rotation in uniform flow has been investigated both numerically and experimentally as
well as theoretically but only in the limit of vanishing Reynolds numbers. A good review of the
theoretical studies is given by Clift et al. [20]. Rubinow and Keller [21] by means of matched
asymptotic expansions showed that, in the limit of Re ! 0, drag is unchanged by rotation,
whereas, the lift coefficient is found to be linearly proportional to rotational speed X, such that
CL ¼ 2X. Non-dimensional rotational speed is defined as X ¼ Rx=U1 , where R is the sphere
radius and x is the angular velocity. On the other hand, a number of experimental results con-
cerning the forces acting on a rotating sphere at high Reynolds numbers (Re > 10000) have been
reported, (e.g., [22–24]). These experiments showed that the lift coefficient is basically independent
of Re in this range of Reynolds numbers. Barkla and Auchterlonie [25] have conducted experi-
mental measurements of the lift and drag for a rotating sphere in the range of 1500 < Re < 3000.
They found that the lift coefficient is proportional to X; CL ¼ ð0:16  0:04ÞX for the rotational
speeds in the range of 2 < X < 4. Tsuji et al. [26] estimated the lift force on a spinning sphere by
studying the trajectories of spheres bouncing on an inclined plate. Again they found a linear
relation between the lift coefficient and rotational speed; CL ¼ ð0:4  0:10ÞX for 550 < Re < 1600
and X < 0:7.
In the range of moderate Reynolds numbers (1 < Re < 300), much less information is available.
The only known works are the experimental studies of Oesterle and Dinh [27], Best [28] and the
numerical work of Chegroun and Oesterle [29]. Oesterle and Dinh [27] conducted experiments to
measure the lift force experienced by a spinning sphere with constant rotational speed (1 < X < 6)
and Reynolds numbers in the range of 1 6 Re 6 140. They also established a correlation to esti-
mate the lift coefficient and found that CL decreases with decreasing rotational speed or increasing
Reynolds number. Moreover, their experimental results seem to indicate that the influence of X
vanishes for Re beyond about 100. At such values of Re, the lift coefficient is very slightly de-
creasing with increasing Re.
Best [28] studied the influence of the particle rotation on the stability of isolated spherical and
spheroidal particle wakes experimentally for particle Reynolds numbers less than 300 and rota-
tional speed up to 10 rev/s. Unfortunately he has not studied the rotational effects on drag and lift
coefficients. His results showed that particle rotation reduces the size of the wake and at suffi-
ciently high spinning rates the wake is destroyed and absent. He also concluded that the shear
layer, which is formed on the underside of the rotating sphere at high spinning rates, might have a
contribution to the turbulence enhancement of the free stream.
Chegroun and Oesterle [29] performed a numerical investigation of a rotating sphere in the
range of 0 6 Re 6 40. Their results showed that the lift coefficient decreases with increasing Re at a
given rotational speed. Furthermore, at low Reynolds numbers (Re < 5) the numerical prediction
of lift coefficients were approximately proportional to the rotational speed. However, for higher
Reynolds numbers the ratio of CL =X was found to decrease with increasing Re considerably.
The problem of a spinning spherical particle in a linear shear flow at moderate Reynolds
numbers has also received attention in literature. Salem and Oesterle [30] carried out numerical
computations for Re < 40 and non-dimensional shear rates between )0.3 and 0.3; however, their
report did not include numerical results for the lift forces except for very low Re . They considered
rotational speeds in the range of 2 6 X 6 2 and proposed an expression for the calculation of the
1408 H. Niazmand, M. Renksizbulut / Computers & Fluids 32 (2003) 1405–1433

torque exerted on a spinning particle in a shear flow. Kurose and Komori [31] have conducted a
similar numerical study for a wide range of Reynolds numbers between 1 and 500 and different
shear rates and rotational speeds in the range of 0 6 X 6 0:25. They found that the drag coefficient
is increased slightly with increasing rotational speed at a fixed Re number. Furthermore, the lift
coefficient appears to increase by increasing rotational speed. However, it tends to approach a
constant value for Re P 200, for a given rotational speed. Although they have extended their
study to higher Reynolds numbers which includes the steady asymmetric and unsteady wake
regimes, the effects of flow shearing and particle rotation on the transitional Reynolds numbers
and the resulting wake structure were not considered.
This brief review clarifies that there is still a lack of information regarding the effects of particle
rotation in the range of intermediate Reynolds numbers, which corresponds to the frequently
encountered case of conveying particles in a gas or liquid flow. In this type of two-phase flows,
particles attain rotational speeds due to particle collisions as well as by the mean fluid shear. In
addition in many saltating particle motions in geophysical flows (see for example [32]), particle
rotation may lift the particle far higher into the flow than if rotation were absent, due to the
generation of a lift force known as the Magnus effect [21]. Best [28] have documented the range
of particle rotation rates in geophysical flows from 10 to 40 rev/s in water flows to estimates of
between 100 and 1000 rev/s in air from past experimental studies. Considering the range of
particle sizes that have been used in these experiments, the rotational speeds encountered lie in the
range of 0 < X < 0:5. Furthermore, in many combustors, evaporating droplets may acquire an-
gular velocity about their own axis because of the atomizer design and/or because of the turbulent
ambient flow field. Thus, the investigation of the behavior of spherical particle with general ro-
tation in space along with non-uniform surface blowing (considered as a simplified model for a
vaporizing droplet), makes a useful contribution to the understanding of fundamental transport
phenomena associated with spherical particles.

2. Problem statement and formulation

Consider a time dependent, three-dimensional incompressible flow over a rotating spherical


particle with surface blowing. The particle is suddenly placed in a uniform laminar flow and since
the velocity of the free stream is constant, the unsteadiness of the flow field is only related to the
wake behavior. The flow direction is the positive x coordinate parallel to the x–y plane, which is
the plane of symmetry as shown in Fig. 1. Two coordinate systems are used in the formulation of
the problem: the Cartesian coordinates ðx; y; zÞ and the non-orthogonal generalized coordinates
ðn; g; fÞ. Using a spherical grid, Cartesian coordinates are given as:
x¼ r cos h; y ¼ r sin h cos u; z ¼ r sin h sin u ð1Þ
In the present work, the generalized coordinates ðn; g; fÞ lie along the spherical coordinates
ðh; u; rÞ, i.e., n is the angular, g is the azimuthal, and f is the radial coordinate. In the angular
direction h, and in the azimuthal direction u, the grid points are linearly distributed, however, in
the radial direction the grids are stretched toward the sphere, with a stretching factor less than 1.1.
The amount of stretching was chosen to place about 10 grid points within pffiffiffiffiffiffi the boundary layer
thickness near the front stagnation point which, can be estimated as 1= Re. The domain of the
H. Niazmand, M. Renksizbulut / Computers & Fluids 32 (2003) 1405–1433 1409

Fig. 1. Flow geometry and coordinates.

Fig. 2. Definition of the over lapping grid configuration in azimuthal direction.

flow is bounded by ð1; 1; 1Þ < n, g, f < ðnmax ; gmax ; fmax Þ where, n ¼ 1 and nmax correspond, re-
spectively, to the negative x-axis in upstream of the flow field and the positive x-axis in the
downstream of the flow field, and f ¼ 1 and fmax denote, respectively the sphere surface and the
far field boundary surrounding the sphere (r ¼ R1 ). However, in g coordinate or azimuthal di-
rection u, there are two grid points overlapping, as shown in Fig. 2, to allow for the imple-
mentation of periodic boundary conditions in this direction. Therefore, g ¼ 1 refers to x–y plane
in the positive z-direction i.e., u ¼ 0 and g ¼ gmax corresponds to u ¼ 2p þ Du. Uniform spacing
(Dn ¼ Dg ¼ Df ¼ 1) is used in the generalized coordinates for simplicity.

3. Governing equations and boundary conditions

The laminar, incompressible, viscous flow around a spherical particle is governed by the full
Navier–Stokes equations. The governing equations are cast in integral form, which is more
suitable for the control-volume numerical method.
1410 H. Niazmand, M. Renksizbulut / Computers & Fluids 32 (2003) 1405–1433
Z
qV~ ~
n dA ¼ 0 ð2Þ
A
Z Z Z
o $
qV~ d8 þ qV~V~ ~
n dA ¼ ð p~
nþ s ~
nÞ dA ð3Þ
ot 8 A A

where q is density, 8 is volume, A is area, p is pressure, V~ is velocity vector, ~


n is the normal unit
$
vector, and s is the viscous stress tensor. Inflow boundary condition corresponds to a uniform
flow such that u ¼ U1 , v ¼ w ¼ 0. For outflow, zero gradients along streamlines are applied to all
three velocity components.
~ f ¼ 0;
V~ r f ¼ u; v; w ð4Þ
Since at f ¼ fmax , or at the outer boundary (r ¼ R1 ), the grid surface is a combination of the
inflow and outflow, the outflow boundary conditions are applied for h P 120°. It was found that
the solution is not very sensitive to the choice of this angle as long as it is greater than 90°. The
n ¼ 1 and nmax boundaries denote the grid points along the x-axis of the Cartesian coordinates
where first order azimuthal averages were assigned. In the periodic boundary or g-direction, as
mentioned earlier, there are two grid points overlapping and therefore grid points at gmax are
equivalent to points at g ¼ 2, and similarly grid points at gmax 1 to points at g ¼ 1, as shown in
Fig. 2. Periodic boundary conditions were imposed in this direction, which for velocity component
in x-direction reads as:
ug¼1 ¼ ugmax 1 ; ugmax ¼ ug¼2 ð5Þ
On the surface of a solid sphere, the flow satisfies the no-slip boundary condition:
u¼v¼w¼0 ð6Þ
In the case of flow over a porous sphere with mass injection, the surface velocity distribution, Vsurf ,
is specified as:
Vsurf =U1 ¼ Cð1 þ K cos hÞ ð7Þ
where h is measured form the front stagnation point. Coefficient C is a measure of the strength of
the surface blowing and it is chosen in the range 0:01 6 C 6 0:04, which corresponds to conditions
encountered in many droplet vaporization processes. Coefficient K ¼ 0 refers to the case of
uniform surface blowing. For non-uniform surface blowing, K is set to 1, which leads to higher
surface blowing in the front half of the sphere, thus simulating droplet vaporization in a con-
vective environment. On the surface of a rotating sphere with an angular speed of xz the velocity
components are:
Rxz cos u sin h Rxz cos u cos h
u¼ ; v¼ ; w¼0 ð8Þ
U1 U1
where R is the radius of the sphere.
Initial conditions correspond to a uniform free stream in the positive x-direction and a pre-
scribed surface velocity distribution and/or sphere rotational speed. The initial pressure is spec-
ified as zero over the whole computational domain. The non-dimensional parameters that appear
in the equations and initial and boundary conditions are the Reynolds numbers based on the
H. Niazmand, M. Renksizbulut / Computers & Fluids 32 (2003) 1405–1433 1411

diameter of the sphere, X the rotational angular speed defined as X ¼ Rx=U1 , and coefficients C
and K as defined in Eq. (7).
The drag, FD , and lift, FL , forces are the components of the fluid forces acting on the particle in
the x and y directions respectively, and are given as the sum of the pressure and viscous stress
contributions:
Z Z
$
FD ¼ FD;p þ FD;f ¼ ex ~
p~ n dA þ ~ ex dA
n s ~ ð9Þ
A A
Z Z
$
FL ¼ FL;p þ FL;f ¼ n dA þ
ey ~
p~ ey dA
n s ~
~ ð10Þ
A A

ex and ~
where ~ ey are the unit vectors in the x and y directions respectively. The drag and lift co-
efficients are defined, respectively, as:
FD FL
CD ¼ 2 pR2
; CL ¼ 2 pR2
ð11Þ
0:5qU1 0:5qU1
It should be noted that, in this analysis, the spherical particle does not accelerate or rotate owing
to aerodynamic forces acting on it from the flow field.

4. Numerical solution

The control-volume formulation was used to develop the finite difference equations from the
governing equations. All discretizations were performed in the generalized coordinate system
ðn; g; fÞ. The transient terms were discretized using a second order accurate three point backward
scheme. Central differencing was used for discretizing the diffusion terms. However, for convective
terms a central differencing scheme with deferred correction was employed after linearizing them
using the best available estimates of velocity components from previous iteration. Because of
instability problems associated with central differencing at high Reynolds numbers, simple upwind
differencing was used in generating the active finite difference coefficients, with a correction ap-
plied to the source term so that the net result is that of central differencing [33]. It is well known
that the convection terms may be the source of numerical instabilities when the mesh Reynolds
number becomes too large. In order to avoid such a problem, which may occur far from the
sphere where the internodal spacing in the radial direction becomes greater, a hybrid scheme was
used for the discretization of convective terms near the outer boundary as proposed in Ref. [33].
The numerical method employed, has been developed by Dwyer [34] and improved further to
properly capture the vortex shedding process at moderate Reynolds numbers. The method con-
sists of two steps. The velocity components are first calculated from the momentum equations
using an alternating-direction-predictor-corrector scheme. Since the pressure field is not known at
this stage, pressure field at the previous time step is used. Then the pressure correction is calcu-
lated from the pressure correction equation designed to satisfy continuity equation directly. The
pressure correction equation is of Poisson type and it is solved by the successive-line-over-
relaxation method. This process continues and new estimates for pressure and velocities are
obtained, until the solution converges at each time step.
1412 H. Niazmand, M. Renksizbulut / Computers & Fluids 32 (2003) 1405–1433

5. Computational parameters and numerical accuracy

The accuracy of the developed finite difference algorithm was tested by predicting the axi-
symmetric flow over a solid sphere at Re ¼ 50, 100, 200 and by examining the effects of grid
resolution on the lift and drag coefficients for flow over a rotating sphere at Re ¼ 200 and
X ¼ 0:16. Although the flow is axisymmetric in the first case and there is a plane of symmetry in
the second case, the complete computational domain with periodic boundary conditions is con-
sidered, and the three-dimensional solution scheme is fully exercised in a time accurate manner for
all cases.
The calculations for the axisymmetric case were performed for a grid of ðnmax ; gmax ; fmax Þ ¼
ð71; 50; 71Þ in the computational domain with the outer boundary located at 20 sphere radii from
the sphere center. Table 1 lists the drag coefficient as a function of Re and compares them with
experimental results of Roos and Willmarth [6] and numerical data presented in [20]. The table
also includes separation angles, measured from the front stagnation point, which are in good
agreement with the results from Clift et al. [20]. Fig. 3 shows a comparison of the non-dimensional
wake length (based on the diameter of the sphere), measured from the surface of the sphere to the
end of the stationary, recirculating zone. The comparison is made with the numerical results of
Johnson and Patel [15] and Shirayama [13] and experimental results of Karla and Uhlherr [35].

Table 1
Comparison of the drag coefficient and separation angle with other results
Re Drag coefficient Separation angle (h)a
Present work [6]b [20] Present work [20]
50 1.588 1.620 1.57 139.3 139.3
100 1.095 1.090 1.096 126.8 126.5
200 0.776 0.778 0.772 116.4 116.4
a
Values of separation angle measured from the front stagnation point.
b
Interpolated values are used.

Fig. 3. Comparison of the wake length for flow over a solid sphere.
H. Niazmand, M. Renksizbulut / Computers & Fluids 32 (2003) 1405–1433 1413

Table 2
Drag and lift coefficients as a function of grid density at Re ¼ 200 and X ¼ 0:16
Case nmax  gmax  fmax R1 =Ra Exp.b CD CL
1 51  50  51 20 1.1 0.8122 0.2031
2 51  50  61 20 1.09 0.8122 0.2025
3 61  50  61 20 1.09 0.8105 0.2017
4 71  50  71 30 1.09 0.8091 0.2011
5 71  50  71 20 1.09 0.8103 0.2011
6 81  50  81 40 1.09 0.8101 0.2007
a
R1 , is the outer boundary location, R is the radius of the sphere.
b
Exp., is the expansion ration between the grid points in r-direction.

The effects of grid resolution were examined for flow over a rotating sphere at Re ¼ 200 and
X ¼ 0:16. Table 2 shows the drag and lift coefficients for different grid densities. The extent of grid
expansion in some cases was adjusted, based on the number of grid points and the location of the
outer boundary to achieve a comparable grid density near the surface of the sphere. The maxi-
mum difference in drag and lift coefficients between all cases is about 1.2% and 0.4% respectively.
It is also clear that changing the location of the outer boundary from 20 to 30 radii and to 40 radii
does not introduce an appreciable change to the drag and lift coefficients. For most cases con-
sidered in this study, the grid points of ð71; 50; 71Þ with the location of the far field boundary at 20
radii from the center of the sphere were used. A dimensionless time step of Dsdif ¼ 0:0001 based
on the diffusion time scale (sdif ¼ mt=D2 , where m is kinematic viscosity and t is time) were used to
initiate the calculation. However, this time step was increased by a factor of 1.02 to a maximum
value in the range of 0:0001 6 Dsdif;max 6 0:005 depending on the Reynolds number, surface
blowing strength and rotational speed. These small time steps at early times are required to resolve
the high gradients associated with the impulsive start-up in a time accurate fashion. Reducing
the maximum time step by half for case 5 in Table 2 affected the lift and drag coefficients by less
than 1%.

6. Results and discussion

Calculations were performed for Reynolds numbers up to 300 with non-dimensional rotational
speeds up to X ¼ 1. This range of Reynolds numbers covers the three different wake regimes
discussed earlier. Re ¼ 100 and 200 in the steady axisymmetric wake regime, Re ¼ 250 in steady
non-symmetrical wake regime, and Re ¼ 300 in unsteady wake regime will be considered as
representatives in each wake regime. The effects of rotational speed and surface blowing on the
wake structure, the transitional Reynolds numbers, and the lift and drag coefficients will be
presented.

6.1. Steady axisymmetric wake regime

First, the changes induced in the flow structure by particle rotation in the absence of surface
blowing will be examined. The typical values considered here are Re ¼ 100 and X ¼ 0:25. Fig. 4(a)
1414 H. Niazmand, M. Renksizbulut / Computers & Fluids 32 (2003) 1405–1433

Fig. 4. Streamlines for flow over a rotating sphere at Re ¼ 100 and X ¼ 0:25, (a) x–y plane, (b) x–z plane.

and (b) shows the streamlines in the symmetry x–y plane (which is orthogonal to the rotational
axis) and the x–z plane. Considering the streamlines in Fig. 4(a), it is clear that the low velocity
region in the wake is displaced to the underside of the particle and its size is not constant in the
azimuthal direction due to the difference in the top and the bottom of the vortex ring. However, in
x–z plane the flow is symmetric as expected, and the separation point has moved slightly backward
(h ¼ 129) as compared to the flow over solid sphere in the absence of the rotation (h ¼ 126:8). The
recirculating zone in this plane is smaller when compared to the case of zero rotation and becomes
smaller as the rotational speed increases. The recirculating zone can be totally absent in both
planes for X P 0:5, as is seen in Fig. 5(a) and (b) in which rotational speed has increased to
X ¼ 0:5. The experimental work of Best [28] has also indicated that the entrainment of the flow
around the particle increases as the rotational speed increase and the recirculating region ceases to
exit for X P 0:5. It is expected that flow parameters such as the lift and drag coefficients are also
affected by the modifications brought about by rotation. Fig. 6 shows the time history (based on
the convective time scale, s ¼ tU1 =D, where t is time) of drag coefficient for Re ¼ 100 and different
angular velocities. There is a continuous increase in the drag coefficient with increasing X, with a
maximum difference of about 14% for the case of X ¼ 1:0 as compared to the case of zero ro-
tation. It is also evident that even at a high rotational speed of X ¼ 1:0, the wake flow approaches
a steady state pattern at this Reynolds number. The same behavior can be seen in the time history
of the lift coefficient in Fig. 7. The lift coefficient for all rotational speeds increases sharply at early
times and approaches its final values asymptotically. The final values of the lift coefficients are
much below the values predicted by the theoretical expression CL ¼ 2X derived by Rubinow and
Keller [21] for creeping flows. However, the steady state CL values in Fig. 7 tend to increase almost
linearly with rotational speed for X < 0:5.
As mentioned earlier Oesterle and Dinh [27] proposed a correlation based on their experimental
results to estimate the lift coefficient in the range 10 < Re < 140 and 1 6 X 6 6. For comparison,
the case of X ¼ 2:0 for Re ¼ 100 has also been examined. Present numerical calculations predict
H. Niazmand, M. Renksizbulut / Computers & Fluids 32 (2003) 1405–1433 1415

Fig. 5. Streamlines for flow over a rotating sphere at Re ¼ 100 and X ¼ 0:5, (a) x–y plane, (b) x–z plane.

Fig. 6. Time history of drag coefficients for Re ¼ 100.

Fig. 7. Time history of lift coefficients for Re ¼ 100.


1416 H. Niazmand, M. Renksizbulut / Computers & Fluids 32 (2003) 1405–1433

CL ¼ 0:581 and 0.647 for X ¼ 1:0 and 2.0 respectively, whereas the experimental values are
CL ¼ 0:685 and 0.745 for the same conditions. The discrepancy can be justified based on the
accuracy of the measured values for lift coefficients, which were reported to be about 20% [27].
As stated earlier, the steady axisymmetric wake regime for a solid sphere ends at a Reynolds
number of about 212. Here the case of Re ¼ 200 with particle rotation is examined next. The wake
flow pattern for Re ¼ 200 and X ¼ 0:16 (not shown) is similar to Re ¼ 100 and X ¼ 0:25 shown in
Fig. 4(a) and (b). A shear layer forms on the underside of the sphere where fluid being dragged
around meets flow underneath the particle. This unstable shear layer can roll up and create a
vortex eye as Reynolds number increases. The generation of this vortex eye and its subsequent
growth in time, leads to eddy shedding, which is the case for Re ¼ 200 and X ¼ 0:25 (not shown).
Wake flow patterns for this rotational speed give rise to vortex shedding in early times which
manifest itself in an oscillatory behavior in the drag and lift coefficients. However, the amplitude
of these oscillations is continuously decreasing and the wake tends to approach a steady state
structure (Figs. 9 and 10). But, the periodic nature of the wake pattern is preserved in time when
the rotational speed is increased to X ¼ 0:5 as can be seen in Fig. 8. This figure shows the in-
stantaneous streamlines in the x–y plane for two time levels in one cycle for Re ¼ 200 and X ¼ 0:5.
The rolling up of the shear layer and the formation of a small recirculating region is evident in the
top panel. However, this vortex loses its identity in the next stage and leaves behind a shear layer
that will roll up and form another vortex. Although these vortices form relatively far from the
sphere, almost one diameter away from the center of the sphere in the lower part of the wake, they
still have a noticeable effect on the drag and lift coefficients which results in an oscillatory behavior
in both drag and lift coefficients as can be seen in Figs. 9 and 10. In these figures the time evolution
of drag and lift coefficients for Re ¼ 200 and four rotational speeds have been presented. The
amplitudes of the oscillations for X ¼ 0:5 are about 2  10 3 and 7  10 3 for the drag and lift
coefficients respectively. For comparison, the amplitude of the drag and lift coefficient oscilla-
tions are about 2:4  10 3 and 1:6  10 2 at Re ¼ 300 for a non-rotating solid sphere. By in-

Fig. 8. Streamlines at two time levels in one cycle for flow over a rotating sphere at Re ¼ 200 and X ¼ 0:5.
H. Niazmand, M. Renksizbulut / Computers & Fluids 32 (2003) 1405–1433 1417

Fig. 9. Time history of drag coefficients for Re ¼ 200.

Fig. 10. Time history of lift coefficients for Re ¼ 200.

creasing the rotational speed, the vortex shedding process plays a stronger role in the drag and lift
fluctuations as in the case of X ¼ 1:0 in Figs. 9 and 10. Furthermore, Fig. 10 indicates that the
Strouhal number, (fD=U1 , where f is the frequency of the vortex shedding) associated with the
unsteadiness in wake increase with increasing X. It is interesting to note that in Fig. 9, there is
about 17% and 30% increase in the average drag coefficients at respective rotational speeds of
X ¼ 0:5 and X ¼ 1:0, as compared to the case of a non-rotating sphere.
Next the effects of surface blowing for Re ¼ 200 and X ¼ 0 are examined. Fig. 11(a) and (b)
shows the x–y plane and x–z plane streamlines with non-uniform surface blowing such that
C ¼ 0:04 and K ¼ 1 in Eq. (7). It is interesting to note that the wake pattern is not symmetrical
anymore and features associated with the non-symmetrical wake regime of 212 < Re < 270 are
present, as described in [15,16]. Some information regarding this change can be obtained from the
time histories of the drag and lift coefficients shown in Fig. 12. Time history of the drag coefficient
for flow over solid sphere is also included for reference. Because of symmetrical geometry and
initial conditions, the flow field at first develops into an axisymmetric wake structure. However,
this structure seems not to be a stable solution at this flow conditions and numerical noise similar
to natural disturbances eventually transfer the wake flow into a more stable structure at about
s ¼ 90, which turns out to be a non-symmetrical steady pattern. It is also worth noting that
1418 H. Niazmand, M. Renksizbulut / Computers & Fluids 32 (2003) 1405–1433

Fig. 11. Streamlines for flow over a non-rotating spherical particle with non-uniform surface blowing: C ¼ 0:04 at
Re ¼ 200; (a) x–y plane, (b) x–z plane.

Fig. 12. Time evolution of drag and lift coefficients for Re ¼ 200 with non-uniform surface blowing, C ¼ 0:04, K ¼ 1.

despite the considerable changes in the wake flow pattern, there is only a slight increase in drag
coefficient as a result of this transition. Compared to solid sphere case there is about 10%
reduction in total drag due to surface blowing. The obvious reduction in friction drag due to
the surface blowing is essentially offset by the increase in pressure drag.
More severe wake unsteadiness can be observed when both particle rotation and surface
blowing are simultaneously applied. Fig. 13 shows the time history of drag coefficient for different
surface blowing velocities at X ¼ 0:25. While surface blowing with the strength C ¼ 0:04 decreases
the total drag (see Fig. 12), particle rotation causes an increase in total drag (see Fig. 9). It appears
that the net effect is an oscillatory behavior in the time history of the drag coefficient, with an
average value slightly below the drag coefficient of a solid sphere. Uniform surface blowing results
in a higher pressure drag contribution to the total drag as compared to non-uniform surface
blowing and thus, accounts for the higher drag coefficient for that case in Fig. 13. Reducing the
H. Niazmand, M. Renksizbulut / Computers & Fluids 32 (2003) 1405–1433 1419

Fig. 13. Time history of drag coefficients for a rotating sphere at Re ¼ 200 and different surface blowing.

strength of surface blowing to C ¼ 0:01 while the rotational speed is held constant results in
displacing the drag coefficient to above its value corresponding to flow over a solid sphere. Even
this low surface blowing displays an oscillatory behavior in the time history of drag coefficient,
thus indicating that vortex shedding is present.
Considering the time histories of the lift coefficient in Fig. 14, for the same velocity profiles as in
Fig. 13, the effects of non-uniform surface blowing on the oscillation amplitude of CL are ap-
parent. Comparing the average CL of these cases with the average CL for the case with the same
rotational speed but in the absence of surface blowing, i.e., CL ¼ 0:275 (Fig. 10), indicates that
non-uniform surface blowing increase CL to about CL ¼ 0:285 for C ¼ 0:04 and CL ¼ 0:282 for
C ¼ 0:01. Whereas, uniform surface blowing case with CL ¼ 0:273 is essentially unaffected.

6.2. Steady non-symmetrical wake regime

Computations performed for flow over a solid sphere at Re ¼ 250 indicate a non-symmetrical
wake regime. Although the wake is no longer symmetric, the flow contains a plane of symmetry
and the wake is steady and attached to the sphere. The wake flow pattern and its pressure

Fig. 14. Time history of lift coefficients for flow over a rotating sphere with X ¼ 0:25 at Re ¼ 200, and different surface
blowing velocities.
1420 H. Niazmand, M. Renksizbulut / Computers & Fluids 32 (2003) 1405–1433

Fig. 15. Time history of drag and lift coefficients for flow over a solid sphere at Re ¼ 250, X ¼ 0.

distribution for this case have been discussed previously by Niazmand and Renksizbulut [16]. Here
we examine the time history of drag and lift coefficients plotted in Fig. 15 is examined. Similar to
Fig. 12, it is evident that the wake flow initially tends to develop into a symmetrical pattern,
however, at later times (s ffi 70), it transforms into to a non-symmetrical structure as its steady
configuration. This is consistent with the experimental observations of Magarvey and Bishop [2]
for the motion of a liquid drop falling is a surrounding immiscible liquid. Experimental obser-
vations of the wake behind a free falling solid sphere by Goldburg and Florsheim [4] also indicate
a non-symmetrical wake pattern in the range 212 < Re < 270. Computed values of CD and CL for
this Reynolds number are 0.709 and 0.062, respectively. The same value for CL with a negative
sign was found in the numerical calculations of Johnson and Patel [15] (note that the y-axis
orientation is arbitrary and the lift force can be positive or negative). Unfortunately, they have not
reported a value for CD . The experimental correlation of Morsi and Alexander [36] indicates a
value of 0.71 for drag coefficient. A noteworthy feature of Fig. 15 is the slight increase in drag
coefficient despite the transformation of the wake pattern from a 2-D axisymmetric pattern to a
three-dimensional non-symmetrical structure, which builds up a secondary flow inside the core of
the wake in azimuthal direction (see [15,16] for details of the wake pattern). This fact, as pointed
out by Roos and Willmarth [6], is in harmony with the observation that the standard sphere CD –
Re data give no indication of the Re at which non-symmetrical structure or unsteadiness com-
mences, i.e., the data do not show any jump or bend to mark the onset of these wake transitions.
At this Reynolds number, transition to an unsteady wake is much more likely to occur when
particle rotation and/or surface blowing are present. Previously [16], it was shown that the wake
flow at Re ¼ 250 develops into an unsteady vortex shedding pattern if surface blowing is intro-
duced into the flow. The changes that particle rotation brings about into the wake flow will be
examined here with the help of Fig. 16, where the instantaneous streamlines in the x–y plane are
plotted in a sequence. The region of the interest is the near wake region extending to about four
diameters away from the center of the sphere. Four time levels in one period of the unsteady wake
flow are shown and the time levels have been chosen in a way that the general features of the
formation, growth and shedding of the vortices can be clearly observed. In the first panel a new
vortex eye is forming in the lower part of the wake region, which entrains fluid from the upper side
of the sphere. The low velocity region downstream from the particle is displaced to the underside
of the particle due to its rotation. Second panel shows that the recirculating zone has grown
considerably by entraining a large amount of the fluid from both, the lower side and the upper
H. Niazmand, M. Renksizbulut / Computers & Fluids 32 (2003) 1405–1433 1421

Fig. 16. Streamlines at four time levels in one cycle for flow over a rotating particle with X ¼ 0:25 at Re ¼ 250, in x–y
plane.

side of the sphere. Despite the rapid growth, the center of the recirculating zone, while spiraling
inward, has remained in line with the eye of the previous panel. In the third stage, it is clear that
the eye of the recirculating zone, while its spiraling direction has changed, has moved down-
stream. A large amount of fluid coming from the upper side of the particle encircles the recir-
culating zone and joins downstream. At the final stage, the recirculating zone has been convected
farther downstream and its spiraling appearance is not present any more. In other words, it has
been shed, and what is left behind is a strong shear layer, which will roll up and will form a new
recirculating eye near the sphere, as in the first panel. This unsteady wake, in contrast to the
unsteady wake for Re ¼ 300 as discussed in [15,16] contains only one recirculating zone. It is also
interesting to note that the wake flow pattern in panel II in which the circulating zone is about to
reach its maximum size corresponds to the minimum drag and lift coefficients in the cycle, while
the wake flow pattern in panel IV corresponds to the maximum lift and drag in the cycle.
1422 H. Niazmand, M. Renksizbulut / Computers & Fluids 32 (2003) 1405–1433

The same flow pattern can be observed at lower rotational speed of X ¼ 0:16, except for an
additional small recirculating zone that forms above the vortex eye discussed for X ¼ 0:25 during
the period of its formation and growth in the panels I and II.
Fig. 17 shows the x–z plane instantaneous streamlines corresponding to their x–y plane coun-
terparts. The recirculating zones in this plane develop different flow patterns in a cycle and spe-
cially the spiraling direction of the recirculating zone changes with time. In the first and the second
panel streamlines are spiraling outwards, whereas in the third and the last panel, the streamlines
are spiraling inwards toward the eye of the recirculating zone. It should be noted that the above
streamlines do not represent true two-dimensional streamlines due to the existence of the out of
plane velocity components in x–z plane. They neither correspond to the particle paths since the
flow is unsteady. Nevertheless, these streamlines provide some useful information about the ki-

Fig. 17. Streamlines at four time levels in one cycle for flow over a rotating particle with X ¼ 0:25 at Re ¼ 250, in x–z
plane.
H. Niazmand, M. Renksizbulut / Computers & Fluids 32 (2003) 1405–1433 1423

Fig. 18. Time history of drag coefficients for flow over a rotating sphere at Re ¼ 250 and different rotational speeds.

Fig. 19. Time history of lift coefficients for flow over a rotating sphere at Re ¼ 250 and different rotational speeds.

nematics of the flow field. However, in symmetry plane, i.e., x–y plane, there are no out of plane
velocity components and therefore, the traces in Fig. 16 do represent true streamlines.
In Figs. 18 and 19 the time dependent drag and lift coefficients for Re ¼ 250 at four rotational
speeds are presented. Both figures show a clear periodic wake for rotational speeds higher than
X > 0:05, although this behavior is more pronounced for CL . The trend of variation of both CL
and CD with increasing rotational speed is similar to the Re ¼ 200 case shown in Figs. 9 and 10,
except for the amplitude of the oscillations for X ¼ 0:5 which shows a decrease as compared to
X ¼ 0:25. Furthermore, the time variation of the CD at this rotational speed is not of sinusoidal
form anymore. This type of behavior can be attributed to the increasing entrainment of the fluid
around the particle with increasing rotational speed, which causes the displacement of the wake
further to the lower part of the sphere and also leads to the formation of the eddies farther from
the particle and therefore reduce their influence on the oscillations of the CD and CL .

6.3. Unsteady wake regime

Present calculations indicate an unsteady wake flow for Re ¼ 300 over a solid sphere. At the
same time, this Reynolds number is close enough to the steady flow regime, so as to provide some
1424 H. Niazmand, M. Renksizbulut / Computers & Fluids 32 (2003) 1405–1433

insight into the transition mechanism. As the Reynolds number increases beyond about 270 not
every naturally small disturbances of the flow, random or not, can be damped and this leads to a
non-symmetrical eddy shedding pattern. Such disturbances always occur during any physical
experiment. Conceptually, the origin of the destruction of the symmetrical pattern can be ex-
plained by the presence of multiple perturbation sources in the physical model such as non-uni-
form inlet conditions, surface irregularities of the boundaries, and vibrations in the system [37]. In
the case of numerical simulation of the unsteady flow past a solid sphere all these destabilizing
effects are absent. As the geometry of the flow and the initial and boundary conditions are
symmetric, the Navier–Stokes equations lead to a symmetric solution even for values of Reynolds
numbers beyond the critical limits as indicated by Anderson et al. [38], Shirayama [13] and Braza
et al. [37]. However, the truncation and round-off errors, as well as those due to the numerical
scheme and the sweep direction in the solution procedure are among the numerical perturbation
factors that could eventually transforms the symmetrical wake solution into a vortex shedding
pattern. All calculations in this study were carried out without imposing any perturbations on the
flow field and all periodic properties of the wake flow were allowed to arise naturally. An ex-
panded plot of the time evolution of the drag and lift coefficients for Re ¼ 300 are shown in
Fig. 20, which covers slightly more than one cycle of the oscillation for clarity, where CD appears
to have a small phase lead relative to CL . This phase lead is due to the pressure drag component as
can be seen in Fig. 21, which shows the contribution of the each drag component to the total drag

Fig. 20. Expanded view of time variation of drag and lift coefficients for flow over a solid sphere at Re ¼ 300, X ¼ 0.

Fig. 21. Expanded view of time variation of drag components for flow over a solid sphere at Re ¼ 300, X ¼ 0.
H. Niazmand, M. Renksizbulut / Computers & Fluids 32 (2003) 1405–1433 1425

and it is evident that pressure drag is the dominant component. The average pressure drag is
almost 1.5 times the average friction drag. Assuming the x–y plane as symmetry plane, the lateral
force along the y-axis is the lift force. As stated earlier, due to symmetrical geometry, the y-axis
orientation is arbitrary and therefore, the lift force can be positive or negative. For present cal-
culation this force is negative. From Fig. 20 for Re ¼ 300, it is found that the average drag and lift
coefficients are about 0.654 and )0.069 with respective oscillation amplitudes of about 2:4  10 3
and 1:4  10 2 . Johnson and Patel [15] calculated CD ¼ 0:656 and CL ¼ 0:069 with oscillation
amplitudes of about 3:5  10 3 and 1:6  10 2 , respectively. The experimental data of Morsi and
Alexander [36] gives a value of CD ¼ 0:65 for this Reynolds number. In the present work, a
Strouhal number of 0.126 is obtained based on the frequency spectra for both drag and lift co-
efficients, which agrees well with the experimental result of Sakamoto and Haniu [10] and with
numerical calculations of Kurose and Komori [31].
A more general view of the flow for this Reynolds number may be obtained with the help of
Fig. 22, which shows a sequence of the instantaneous streamlines for the symmetry plane. Similar
to Fig. 16, the near wake region in the x–y plane is considered and four time levels in one cycle are
presented sequentially. The features of the formation, growth and shedding of the vortices are in
close resemblance to the numerical predictions of Johnson and Patel [15] for the same conditions.
It is also interesting to note that the vortices are all shed form the same orientation. This fact is
supported by the experimental observations within a small range of Reynolds numbers beyond the
onset of vortex shedding [5]. In addition, the same fact accounts for the non-zero average value for
the lift coefficient for flow over a solid sphere despite the symmetry of the geometry (Fig. 20),
which is in contrast to flow over a circular cylinder that gives a zero mean lift coefficient due to the
alternating nature of the vortex shedding process.
In Fig. 23, the time evolution of the drag coefficient at Re ¼ 300 for three rotational speeds is
plotted. The periodic properties of the flow are obviously affected by particle rotation. As com-
pared to the average drag coefficient with zero rotational speed (CD ¼ 0:654), the present results
indicate an increase of about 4%, 8% and 12% in the average drag coefficients for rotational
speeds of X ¼ 0:05, 0.16 and 0.25, respectively. Similar to other wake regimes, increasing the
rotational speed leads to an increase in the oscillation amplitudes and the Strouhal number of
vortex shedding. For X ¼ 0:16 and 0.25, the amplitude of the oscillations are two and three times
larger than the corresponding solid sphere case, respectively. Moreover, the establishment time for
a regular periodic behavior of the drag coefficient is increasing at lower rotational speeds. In fact
present calculation shows that flow over a non-rotating solid sphere does not reach a regular
periodic state within the investigation time of Fig. 23 in the absence of any artificial perturbation
sources introduced into the flow field. Fig. 24 shows the time dependent viscous and pressure
components of the drag coefficient. It is evident that the pressure contributions are dominant.
Furthermore, particle rotation has a major effect on the pressure drag components while the
friction drag components are only slightly influenced by rotational speeds.
Fig. 25 shows the time evolution of the lift coefficients for the same conditions as in Fig. 23.
Particle rotation increases the average lift coefficients and the oscillation amplitudes considerably,
as compared to the flow over a solid sphere. However, the average values of the lift coefficients are
very close to the average values for Re ¼ 200 and 250 at the same rotational speeds. Similar
behavior has been reported in the experimental results of Oesterle and Dinh [27] for Reynolds
numbers beyond 100 and rotational speeds higher than 1.
1426 H. Niazmand, M. Renksizbulut / Computers & Fluids 32 (2003) 1405–1433

Fig. 22. Streamlines at four time levels in one cycle for flow over solid sphere at Re ¼ 300, x–y plane.

Fig. 23. Time evolution of drag coefficients for flow over a rotating sphere at Re ¼ 300 and different rotational speeds.
H. Niazmand, M. Renksizbulut / Computers & Fluids 32 (2003) 1405–1433 1427

Fig. 24. Time evolution of drag components for flow over a rotating sphere at Re ¼ 300 and different rotational speeds.

Fig. 25. Time evolution of lift coefficients for flow over a rotating sphere at Re ¼ 300 and different rotational speeds.

The ratio of the average lift to average drag force also increases with increasing rotational
speed. This ratio is about 0.1 for flow over a non-rotating solid sphere at Re ¼ 300 and increases
to 0.17, 0.30, 0.37 and 0.5 for rotational speeds of X ¼ 0:05, 0.16, 0.25 and 0.5 respectively. It is
evident that at higher rotational speeds the lateral force is comparable to the axial force, which
can modify the trajectories of the individual particles significantly as compared to a non-rotating
particle.
In Figs. 26 and 27 the time variation of the drag and lift coefficients for a rotating sphere
at Re ¼ 300 are plotted when surface blowing is simultaneously applied. Non-uniform surface
blowing with strength C ¼ 0:02 and two rotational speeds X ¼ 0:16 and 0.25 are considered.
Comparison with Figs. 23 and 25 indicates that the average drag coefficients have slightly de-
creased for both rotational speeds although the oscillation amplitudes have increased. Further-
more, both drag and lift coefficients for X ¼ 0:16 have not reached a quasi-steady state even after
a long period of time (s ¼ 100). It appears that there is a low frequency pattern (with a frequency
of 1=3 of the vortex shedding frequency), consisting of three successive oscillations with unequal
amplitudes, however the difference in the amplitudes vanishes very slowly in time.
The time-averaged behavior of the lift and drag coefficients discussed so far will be studied next.
Fig. 28 shows the variation of CD as a function of particle Reynolds number. The results are only
1428 H. Niazmand, M. Renksizbulut / Computers & Fluids 32 (2003) 1405–1433

Fig. 26. Time evolution of drag coefficients over a rotating particle with non-uniform surface blowing, C ¼ 0:02 and
K ¼ 1 at Re ¼ 300.

Fig. 27. Time evolution of lift coefficients over a rotating particle with non-uniform surface blowing, C ¼ 0:02 and
K ¼ 1 at Re ¼ 300.

Fig. 28. Comparison of drag coefficients for flow over a rotating sphere.
H. Niazmand, M. Renksizbulut / Computers & Fluids 32 (2003) 1405–1433 1429

Fig. 29. Comparison of lift coefficients for flow over a rotating sphere.

presented for two rotational speeds for comparison with the numerical calculations of Kurose and
Komori [31]. In addition to the good agreement, it can be observed that the effect of particle
rotation on the drag coefficient is more significant at higher Reynolds numbers. In Fig. 29, the
variation of the lift coefficients with respect to Reynolds numbers for several rotational speeds are
plotted. For comparison, numerical results of Ref. [31] for two rotational speeds (X ¼ 0:16 and
0.25) are also included, which show good agreement. Lift coefficients in Fig. 29 reach a minimum
at about Re ¼ 20 for X > 0:05 and then increase smoothly with increasing Reynolds numbers.
However, lift coefficients approaches constant values asymptotically for Reynolds numbers ex-
ceeding about 200 at a fixed rotational speed. It is also clear that these asymptotic values are
increasing with increasing rotational speed. Furthermore, these values are close to the measured
values of lift coefficients by Tsuji et al. [26]. They found CL in the range of 0.05–0.3 for
550 6 Re 6 1600 at rotational speeds between 0.1 and 0.5. As mentioned earlier, experimental
results of Oesterle and Dinh [27] also indicate that the influence of X almost vanishes for Reynolds
numbers beyond about 100 and X > 1. It is also noticeable that the average values of the lift
coefficient for flow over a non-rotating solid sphere (X ¼ 0) at Re ¼ 250 and 300 are almost equal.

6.4. Correlations for lift and drag

The smooth variation of lift coefficients for Re > 100 in Fig. 29 suggests that the curves cor-
responding to different rotational speeds might collapse into a single curve. As shown in Fig. 30,
all numerical data for 0:16 6 X 6 0:5 and 100 6 Re 6 500 correlate within 20% with the following
equation:
CL
3:6
¼ 0:11 ð12Þ
ð1 þ XÞ
which includes data in the range 300 < Re 6 500 taken from the numerical results of [31].
In Fig. 31 the time-averaged variation of drag coefficient with respect to rotational speed X for
different Reynolds numbers is plotted. A steady increase in drag coefficients with increasing ro-
tational speed at a given Reynolds number is evident. Furthermore, the rate of increase in CD with
increasing X is somewhat higher for higher Reynolds numbers. It is also evident that for Re 6 50
1430 H. Niazmand, M. Renksizbulut / Computers & Fluids 32 (2003) 1405–1433

Fig. 30. Lift coefficient, Eq. (12).

Fig. 31. Variation of drag coefficients with rotational speeds at different Re.

drag coefficients are almost independent of the rotational speeds. Therefore, particles do not
experience higher drag as compared to the solid sphere due to their rotation. For the same reason
and clarity of the figure, results related to Reynolds numbers 10 and 20 are not shown.
Finally in searching for an engineering correlation that can provide accurate information re-
garding drag coefficient when particles are rotating or when there is surface blowing or when both
effects are simultaneously applied, it is found that all numerical data in the present study can be
correlated within 10% with the following equation (see Fig. 32):
CD ð1 þ 20VS Þ0:2
¼ CDS ð13Þ
ð1 þ XÞRe=1000
where VS is the average surface blowing velocity normalized by U1 , and CDS is the standard drag
coefficient for flow over a solid sphere, which are defined, respectively, as:
Z
1
VS ¼ Vsurf dA ð14Þ
U1 ð4pR2 Þ A
H. Niazmand, M. Renksizbulut / Computers & Fluids 32 (2003) 1405–1433 1431

Fig. 32. Drag coefficient correlation, Eq. (13).

 
24 Re2=3
CDS ¼ 1þ 1 6 Re 6 1000 ð15Þ
Re 6
Eq. (13) is valid for 0 6 VS 6 0:04, 0 6 X 6 0:5 and 10 6 Re 6 300. However, it is found that for
cases with particle rotation only, Eq. (13) is valid for a more extended range of Reynolds numbers,
that is, 10 6 Re 6 500. In Fig. 32, data in the range 300 < Re 6 500 and 0 < X 6 0:25 were taken
from Kurose and Komori [31].

7. Concluding remarks

A numerical investigation has been conducted to examine the effects of particle rotation and
surface blowing on the flow past a sphere at moderate Reynolds numbers. The parametric ranges
considered in this study are 10 6 Re 6 300, rotational speed, 0 < X 6 1:0 and surface blowing
strength, 0 6 C 6 0:04. Numerical calculations predict a steady axisymmetric wake regime for
Re 6 212 and a steady non-symmetrical wake up to Re ¼ 270, while the separation occurs at
Re ¼ 20. However, an unsteady wake flow was found at Re ¼ 300. Numerical prediction of wake
flow patterns and lift and drag coefficients have been compared with existing experimental and
numerical data for flow over a solid sphere, and good agreement has been found. Calculations for
different rotational speeds and various surface blowing velocities, in each wake regime, were
performed and it was found that both effects can decrease the transitional Reynolds numbers
between the wake regimes such that a steady non-symmetrical wake pattern develops due to
surface blowing at Re ¼ 200. Furthermore, vortex shedding can also occur at Re ¼ 200 if surface
blowing and particle rotation are applied simultaneously. This type of unsteady three-dimensional
wake flow has not been reported by previous investigators. However, Sakamoto and Haniu [10]
have also observed that linear shear flow promotes the generation of eddy shedding at Reynolds
numbers as low as 230.
Time evolution of the drag and lift coefficients with particle rotation and surface blowing have
been presented. It is also shown that global properties, such as average lift and drag coefficients,
1432 H. Niazmand, M. Renksizbulut / Computers & Fluids 32 (2003) 1405–1433

and the Strouhal number are also affected considerably by the changes introduced by both effects.
Correlations for calculating the average drag and lift coefficients are proposed based on the
numerical data of the present study.

Acknowledgement

The financial support of the Natural Sciences and Engineering Research Council of Canada
(NSERC) is gratefully acknowledged.

References

[1] Taneda S. Experimental investigation of the wake behind a sphere at low Reynolds numbers. J Phys Soc Jpn
1956;11:1104–8.
[2] Magarvey RH, Bishop RL. Transition ranges for three-dimensional wakes. Can J Phys 1961;39:1418–22.
[3] Magarvey RH, Maclatchy CS. Vortices in sphere wakes. Can J Phys 1965;43:1649.
[4] Goldburg A, Florsheim BH. Transition and Strouhal number for incompressible wake of various bodies. Phys
Fluids 1966;9:45.
[5] Achenbach E. Vortex shedding from spheres. J Fluid Mech 1974;62:209–21.
[6] Roos FW, Willmarth WW. Some experimental results on sphere and disk drag. AIAA J 1971;9(2):285–91.
[7] Nakamura I. Steady wake behind a sphere. Phys Fluids 1976;19:5–8.
[8] Kim HJ, Durbin PA. Observations of the frequencies in a sphere wake and of drag increase by acoustic excitation.
Phys Fluids 1988;19:3260–5.
[9] Sakamoto H, Haniu H. A study of vortex shedding from spheres in a uniform flow. Trans ASME: J Fluids Eng
1990;112:386–92.
[10] Sakamoto H, Haniu H. The formation mechanism and shedding frequency of vortices from a sphere in uniform
shear flow. J Fluid Mech 1995;287:151–71.
[11] Kim I, Pearlstein A. Stability of the flow past a sphere. J Fluid Mech 1990;211:73–93.
[12] Natarajan R, Acrivos A. The instability of the steady flow past spheres and disks. J Fluid Mech 1993;254:323–44.
[13] Shirayama S. Flow past a sphere: topological transitions of the vorticity field. AIAA J 1992;30(2):349–58.
[14] Tomboulides AG, Steven AO. Numerical investigation of transitional and weak turbulent flow past a sphere.
J Fluid Mech 2000;416:45–73.
[15] Johnson A, Patel VC. Flow past spheres up to Reynolds number of 300. J Fluid Mech 1999;378:19–70.
[16] Niazmand H, Renksizbulut M. 3-D effects of surface blowing on vortex shedding from spheres at moderate
Reynolds numbers. In: Proceedings of the Eight Annual Conference of the Computational Fluid Dynamics Society
of Canada, 11–13 June 2000, Montreal, vol. 2. p. 723–32.
[17] Lee S. A numerical study of the unsteady wake behind a sphere in a uniform flow at moderate Reynolds numbers.
Comput Fluids 2000;29:639–67.
[18] Chuchottaworn P, Fujinami A, Asano K. Numerical analysis of the effect of mass injection or suction on drag
coefficients of a sphere. J Chem Eng Jpn 1983;16:18–24.
[19] Cliffe KA, Lever DA. Isothermal flow past a blowing sphere. TP.1042, Theoretical Physics Division, AERE,
Harwell, Oxford shire, May 1984.
[20] Clift R, Grace JR, Weber ME. Bubbles, drops and particles. New York: Academic Press; 1970.
[21] Rubinow SI, Keller JB. The transverse force on a spinning sphere moving in a viscous fluid. J Fluid Mech
1961;11:447–59.
[22] Maccoll JH. Aerodynamics of a spinning sphere. J R Aeronaut Soc 1928;32:777–98.
[23] Davies JM. The aerodynamics of golf balls. J Appl Phys 1949;20:821–8.
[24] Tani I. BaseballÕs curved balls. Science 1950;20:405–9.
H. Niazmand, M. Renksizbulut / Computers & Fluids 32 (2003) 1405–1433 1433

[25] Barkla HM, Auchterlonie LJ. The Magnus or Robins effect on rotating spheres. J Fluid Mech 1971;47:437–47.
[26] Tsuji Y, Morikawa Y, Mizuno O. Experimental measurements of the Magnus force on a rotating sphere at low
Reynolds numbers. Trans ASME: J Fluids Eng 1985;107:484–8.
[27] Oesterle B, Dinh B. Experiments on the lift of a spinning sphere in a range of intermediate Reynolds numbers. Exp
Fluids 1998;25:16–22.
[28] Best JL. The influence of particle rotation on wake stability at particle Reynolds numbers, ReP < 300-implications
for turbulence modulation in two-phase flows. Int J Multiphase Flow 1998;24:693–720.
[29] Chegroun N, Oesterle B. Etude numerique de la trainee, de la portance et du couple sur une sphere en translation et
en rotation. In: Actes 11eme Congres Francais de Mecanique, Lille-Villeneuve dÕAscq, France 1993, vol. 3. p. 81–4.
[30] Salem MB, Oesterle B. A shear flow around a spinning sphere: Numerical study at moderate Reynolds numbers.
Int J Multiphase Flow 1998;24:563–85.
[31] Kurose R, Komori S. Drag and lift forces on a rotating sphere in a linear shear flow. J Fluid Mech 1999;384:183–
206.
[32] Lee H-Y, Hsu I-S. Particle spinning motion during saltating process. J Hydraul Eng 1996;122:587–90.
[33] Ferziger JH, Peric M. Computational methods for fluid dynamics. 2nd ed. Springer; 1998.
[34] Dwyer HA. Calculations of droplet dynamics in high temperature environments. Prog Energy Combust Sci
1989;15:131–58.
[35] Karla TR, Uhlherr PHT. Properties of bluff body wakes. In: 4th Australian Conference of Hydraulic and Fluid
Mechanics Melbourne, Australia: Monash University, 1971. p. 8–12.
[36] Morsi SA, Alexander AJ. An investigation of particle trajectories in two-phase flow systems. J Fluid Mech
1972;55(part 2):183–208.
[37] Braza M, Chassaing P, Minh HH. Numerical study and physical analysis of the pressure and velocity fields in the
near wake of a circular cylinder. J Fluid Mech 1986;165:79–130.
[38] Anderson CR, Greengard C, Greengard L, Rokhlin V. On the accurate calculation of vortex shedding. Phys Fluids
1990;2(6):883–5.

You might also like