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

net/publication/258927553

Effect of Axis Ratio on Fluid Flow Around an Elliptic Cylinder—A Numerical

Study

Article in Journal of Fluids Engineering · July 2013

CITATIONS READS

4 1,312

1 author:

Kalyana Raman
Molex
5 PUBLICATIONS 65 CITATIONS

SEE PROFILE

All content following this page was uploaded by Kalyana Raman on 11 February 2014.

The user has requested enhancement of the downloaded file.

 Effect of Axis Ratio on Fluid
Flow Around an Elliptic
S. Kalyana Raman
M.S. Research Scholar
Cylinder—A Numerical Study
e-mail: kalyan_aero@yahoo.com The bluff body simulations over canonical forms like circular and square cylinders are
1 very well studied and the correlations for bulk parameters like mean drag coefficient and
K. Arul Prakash Strouhal numbers for the same are reported widely. In the case of elliptic cylinder, the lit-
Assistant Professor erature is very sparse, especially for moderate Reynolds number (Re). Hence, in this
e-mail: arulk@iitm.ac.in work, a detailed study about fluid flow characteristics over an elliptic cylinder placed in
a free stream is performed. Simulations are carried out for different Re ranging from 50
S. Vengadesan to 500 with axis ratio (AR) varied between 0.1 to 1.0 in steps of 0.1. Immersed boundary
Associate Professor method is used for the solid boundary condition implementation which avoids the grid
e-mail: vengades@iitm.ac.in generation for each AR and a single Cartesian grid is used for all the simulations. The
effect of AR for various Reynolds numbers is also focused on using the in-house code.
Department of Applied Mechanics, The influence of AR is phenomenal for all the Re and the values of wake length, drag
Indian Institute of Technology Madras, coefficient, and Strouhal number decrease with decreasing AR for a particular Re. The
Chennai 627006, India critical ARs, for vortex shedding and wake formation, are identified for various Re.
Detailed correlations for wake length, critical ARs for vortex shedding and wake forma-
tion, mean drag coefficient and Strouhal number, in terms of AR, are reported in this
work. [DOI: 10.1115/1.4024862]

Keywords: elliptic cylinder, axis ratio, immersed boundary method, wake length, drag
coefficient, Strouhal number

1 Introduction A cylinder with an elliptic cross section is considered as a basic

form since it can be altered into a circular cylinder or a flat plate
Simulation of fluid flow over bluff bodies is a basic and
by varying AR (b/a). In addition, elliptic cylinders are sometimes
foremost step for many industrial applications, such as design of
preferred over circular cylinders in special applications because
automotive and aircraft components, skyscrapers, heat exchang-
the wake region behind them and drag coefficient are small for an
ers, bridge constructions, electronic devices, and many other
elliptic cylinder when compared to that for a circular cylinder. For
engineering applications. In most of the bluff body flows, vortex
a particular Reynolds number (Re), where Re is based on hydrau-
shedding is a common phenomenon and it plays an important role
lic diameter (dH), the wake length decreases with decreasing AR.
in aerodynamic forces, heat transfer, mixing, and structural
It is possible to find a critical AR at which wake length becomes
vibration. The initial observation of vortex shedding was done by
zero for a particular Re, and that AR value is termed as critical AR
Leonardo da Vinci in the sixteenth century. He portrayed the
for wake length (ARcrw). At Re  47, vortices are shed from either
vortex patterns behind the obstacles through drawings. The over-
side of the cylinder in case of circular cylinder (i.e. AR ¼ 1.0).
whelming development of engineering applications induced the
When the value of AR is decreased, at some critical value, vortex
exploration of the studies on vortex shedding phenomenon behind
shedding is suppressed and only standing wake is observed behind
various bluff bodies till date. Many researchers investigated the
the cylinder for a particular Re. The critical AR at which vortex
flow behind bluff bodies experimentally and numerically. Exten-
shedding is suppressed is known as ARcrs. The fluid flow charac-
sive studies on fluid flow over bluff bodies were reported in
teristics vary with changes in the wake region behind the cylinder.
Refs. [1,2].
Hence, it is very important to know the value of the ARcrw and
In general, the bluff bodies are classified into two major shapes,
ARcrs to predict the fluid flow characteristics.
namely circular type and square type. The extensive reviews
Unlike canonical problems like fluid flow behind a circular
about the flow over circular and square cylinders were given in
cylinder, the literature on the fluid flow over elliptical cylinder
Refs. [2–5]. Many engineering applications require the prediction
is very limited, specifically for low Re flows. The performance
of bulk parameters like mean drag coefficient (Cd_av) and Strouhal
of an elliptical geometry surpasses that of circular geometry in
number (St). The fluid flow over bluff bodies are investigated till
many industrial applications associated with limited space. Ear-
date in order to obtain an in depth understanding of the flow
lier studies on elliptic cylinder were mainly focused on cylinders
characteristics as well as finding better correlations for the bulk
oriented at a certain angle with respect to the free stream flow.
parameters. Williamson and Brown [6] reviewed the various
Lugst and Haussling [8] studied the laminar flow past an
St–Re relationship for circular cylinder and Henderson [7]
abruptly accelerated elliptic cylinder at an incidence angle of
reported the critical Re, for vortex shedding and the three-
45 deg and they concluded that steady state solutions were
dimensional effect for circular cylinder and also studied the effect
obtained for this problem till Re ¼ 30 and the famous von
of vortex shedding on Cd_av.
Karman shedding was noticed when Re ¼ 200. Tandea [9] car-
ried many experiments in order to provide clear photographs of
streaklines and streamlines for the flow over impulsively started
1
Corresponding author. elliptic cylinder oriented at various angle to the free stream with
Contributed by the Fluids Engineering Division of ASME for publication in the
JOURNAL OF FLUIDS ENGINEERING. Manuscript received November 8, 2012; final
Re ranging from 40 to 400. A semianalytic study about flow
manuscript received June 4, 2013; published online August 7, 2013. Assoc. Editor: over impulsively started elliptic cylinder at various angles of
Zhongquan Charlie Zheng. incidence with respect to free stream was done by Patel [10] and

Journal of Fluids Engineering Copyright V

C 2013 by ASME NOVEMBER 2013, Vol. 135 / 111201-1

Downloaded From: http://fluidsengineering.asmedigitalcollection.asme.org/ on 02/10/2014 Terms of Use: http://asme.org/terms

 his work concluded that the vanishing of wall shear does not 2 Governing Equations and Numerical Methodology
indicate the separation. Ota and Nishayama [11] experimentally
The governing equations for unsteady, incompressible fluid
investigated the flow over elliptic cylinders in tandem arrange-
flow with constant properties are
ment. Ota et al. [12] investigated the flow around an elliptic cyl-
inder experimentally and reported the dependence of critical Re ru¼0 (1)
with respect to angle of attack.
Jackson [13] carried out extensive studies on elliptic cylinders @u 1 2
þ ðu  rÞu ¼ rp þ r uþf (2)
of different ARs for various Re. The critical Re for vortex shed- @s Re
ding for different ARs was identified in his work. The critical Re
for vortex shedding for various incidence angle of the cylinder Equations (1) and (2) are conservation of mass and momentum,
with respect to the free stream when AR ¼ 0.5 was also reported in respectively.
his work. Park et al. [14] investigated the effect of incidence on The dimensionless variables in the above equations are defined
the unsteady laminar flow past an impulsively started, slender as
elliptic cylinder. In their work, the flow regimes were classified
into five zones for Re ranging from 25 to 600. The steady flow s ¼ tu1 =dH ; x ¼ x=dH ; u ¼ u=u1 ; p ¼ p=qu21
regime was divided into two zones, first one was fully attached
flow and the other one was flow with rotating bubbles. The where t, x, u, and p are dimensional time, length, velocity, and
unsteady flow regime was divided into three zones characterized pressure, respectively. In addition, q, u1, and dH represents den-
as incipient, intermediate and pure von Karman vortex zones. sity, free stream velocity, and hydraulic diameter, respectively.
There were few studies in which oscillating elliptic cylinders were The hydraulic diameter is denoted as dH ¼ 4 A/P, where A is the
considered. Badr [15] simulated the flow over oscillating elliptic area of the cylinder and P is the perimeter of the cylinder.
cylinder inclined to the free stream flow for Re ranging from 102 The nondimensionalised number used in the above equation is,
to 103 and D’Alessio and Kocabiyik [16] studied the flow over Re ¼ qu1 dH/l. For spatial differencing of momentum equations,
transversely oscillating inclined elliptic cylinder of AR ¼ 0.5 at second order central differencing is used for diffusion terms. Sec-
Re ¼ 102. Kocabiyik and D’Alessio [17] investigated the inline ond order central differencing in conservative form is used for
oscillation of inclined elliptic cylinder of AR ¼ 0.5 at Re ¼ 102. convective terms. First order Euler integration is applied for
Recently Faruquee et al. [18] reported the effect of AR on fluid unsteady term. The velocity forcing term is represented as f.
flow over an elliptic cylinder at Re ¼ 40. The value of ARcrw for The IBM method proposed by Su et al. [22] is used in the pres-
Re ¼ 40 was also found in his work. ent study. Further two important changes are made. First, time
The Immersed Boundary Method (IBM) has emerged in the integration of the Navier–Stokes equations has been carried out
last few years as a suitable numerical procedure capable of by means of a second order projection method developed by
handling these kinds of geometries very effectively. In IBM, the Brown et al. [23] which provides second order convergence for
solid boundaries are represented by discrete Lagrangian points. pressure. Second, the Eulerian and Lagrangian variables’ interac-
The mathematical formulation is based on the use of Eulerian tion have been carried out by means of 4-point regularized d func-
variables, to describe the dynamics of the fluid, and of Lagrangian tion [24], which provides a wider stability region than any other d
variables along the moving or stationary structure. In IBM, the function. The algorithm and the implementation of no-slip bound-
effects of the boundary are accounted by introducing a forcing ary condition on the solid body for one time step, is explained
term, in the governing equations, localized at the boundary of the below
body. The forcing term is obtained by applying virtual forces on
Lagrangian points, which represent the solid boundary. The (1) The aforesaid momentum equation is solved without con-
virtual forces are defined by the required boundary conditions. sidering the immersed boundary and the intermediate
The force exerted by the structure on the fluid is taken into velocities u*(x) at the Eulerian grid points are calculated as
account by means of a Dirac-d function. One of the main advan- follows
tages of this approach, for moving boundary problems, over u  un 3 1
unstructured mesh and body fitted methodologies, is that grid gen- ¼  rh ðuuÞn þ rh ðuuÞn1
eration in every time step is not required. In addition, the whole Ds 2 2
n 1
domain is solved on simple orthogonal Cartesian grid irrespective  rh ðpÞ þ r2 ðu þ un Þ (3)
of the shape of the body. Presently, numerous variant approaches 2Re h
of IBM exist, which are discussed in detail in Refs. [19–22]. The
same simple orthogonal Cartesian grid can be used for simulating The convective and diffusive terms in the Navier–Stokes
flow over elliptic cylinder, even though its shape changes for equation are discretized using Adams–Bashforth and Cran-
various ARs. k–Nicolson schemes, respectively, which provides overall
In all the above literature, studies were carried out for a small second order accuracy.
range of Re, incidence with free stream and for a particular AR. (2) The velocities ul ðxl Þ at the Lagrangian points are interpo-
There were very little systematic and comprehensive studies car- lated using nearby Eulerian points
ried out for understanding the flow structure completely. General X
correlating equations for properties like St and Cd_av for fluid flow ul ðxl Þ ¼ u ðxÞdh ðx  xl Þh2 (4)
over an elliptic cylinder was not reported for a wide range of AR x
and low Re. Therefore in the present study, fluid flow over an
elliptic cylinder, with various ARs, is simulated and analyzed (3) The Lagrangian forces f l ðxl Þ are calculated from the pre-
using the in-house code. A robust Navier–Stokes solver on a stag- scribed boundary velocity unþ1 l at each time step and the
gered grid using finite–difference discretization is used for the interpolated velocity calculated in step 2
simulations and IBM is used for implementing the boundary con- !
XM X  
ditions over the elliptic cylinder. Among many variants of existing
dh ðx  xl Þdh x  ðxl Þj f l ðxl Þ
IBM, the method proposed by Su et al. [22] is followed in this x
j¼1
work. In this method, Lagrangian forces are calculated implicitly,
which eliminates time step restriction in the Lagrangian forces’ unþ1
l  ul ðxl Þ
ðxl Þ
¼ (5)
calculation. General correlation equations for bulk parameters like Ds
Cd_av, St for fluid flow over an elliptic cylinder are obtained from
this study. where M is the total number of Lagrangian points.

111201-2 / Vol. 135, NOVEMBER 2013 Transactions of the ASME

Downloaded From: http://fluidsengineering.asmedigitalcollection.asme.org/ on 02/10/2014 Terms of Use: http://asme.org/terms

 Fig. 1 Computational domain with boundary conditions

(4) The Lagrangian forces f l ðxl Þ are distributed to nearby Although the location of boundary is smeared to several Eulerian
Eulerian points, f*(x) grid points, the velocity field obtained using Eq. (7) contains dis-
continuity in the velocity gradient across the immersed boundary.
X
M   This discontinuity may also exist in the divergence free field.
f  ðxÞ ¼ f l ððxl Þk Þdh x  ðxl Þk Ds (6)
k¼1
However, Guy and Hartenstine [26] showed that the solution con-
verges to the original solution if the velocity field obtained after
where Ds is the distance between adjacent Lagrangian Eq. (7) is close to divergence free. Therefore, steps 6 and 7 in the
points. above algorithm are repeated until the divergence of error norm is
(5) Intermediate velocities u*(x) are corrected using the distrib- less than or equal to 107 for each time step. The calculation of
uted force to get another intermediate velocity, u** the forcing term implicitly does not influence the stability of the
time integration. In this method, the summation of the Lagrangian
u  u
¼ f (7) forces of all the Lagrangian points representing the solid body
Ds will give the aerodynamic forces
(6) Solve the pressure Poisson equation as shown in Eq. (8)
rh  u X
M X
M
r2h p ¼ (8) Lift ¼  fly ðxk Þ; Drag ¼  flx ðxk Þ (13)
Ds k¼1 k¼1

The algebraic system of equations resulting from discretiza-

tion of pressure Poisson equation is solved using BICGSTAB where flx and fly are Cartesian components of fl. The dimensionless
(2) [25], one of the most efficient elliptic equation solvers force coefficients are calculated as follows
(7) Correct the pressure and velocities [23] as shown below
Lift Drag
1 CL ¼ ; CD ¼ (14)
pnþ1 ¼ pn þ p  rh  u (9) 0:5qu21 0:5qu21
2Re
unþ1  un The time averaged drag coefficient (Cd_av) is defined as
¼ rh p (10)
Ds ð s2
1
Cd av ¼ CD ds (15)
In the above formulation, d is the Dirac d function, which is s2  s1 s1
employed to transfer the quantities between Eulerian and Lagran-
gian domains effectively where CD is the instantaneous drag coefficient calculated using
Eq. (14). The time interval (s2  s1) is chosen in such a way that it
1 x y is sufficiently large and also an integer multiple of vortex shed-
d¼ 2/ / (11)
h h h ding period.

where h is the Eulerian mesh width, x and y are Cartesian compo-

3 Problem Definition and Validation
nents of x. / is the hat function. In this study, a 4-point d function
[24] is used which is given by Eq. (12) shown below The problem that is considered here is the fluid flow character-
8  istics over unconfined elliptic cylinders of different AR. The com-
1 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi putational domain with boundary conditions for fluid flow
>
> 3  2jrj þ 1 þ 4jrj  4r 2 if 0  jrj  1
>
>
<8 analysis is shown in Fig. 1.
/ðrÞ ¼ 1  pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi In this problem, hydraulic diameter is taken as the characteristic
>
> 5  2jrj þ 7 þ 12jrj  4r 2 if 1  jrj  2 length and free stream velocity u1 is taken as the characteristic
>
> 8
: velocity. Simulations are carried out for fluid flow over elliptic
0 otherwise
cylinders with AR ranging from 0.1 to 1.0 in steps of 0.1 for vari-
(12) ous Re varied between 50 to 500.

Journal of Fluids Engineering NOVEMBER 2013, Vol. 135 / 111201-3

Downloaded From: http://fluidsengineering.asmedigitalcollection.asme.org/ on 02/10/2014 Terms of Use: http://asme.org/terms

 Table 1 Comparison of Cd_av value obtained using different
Ds/h ratio

Ds/h Cd_av (Re ¼ 100, h ¼ 0.01)

2.00 1.22
1.67 1.25
0.99 1.29
0.86 1.28

Table 2 Comparison of Cd_av value obtained using different

mesh sizes h

Cd_av, AR ¼ 0.1 Cd_av, AR ¼ 1.0

h Re ¼ 50 Re ¼ 100 Re ¼ 50 Re ¼ 100

0.04 0.671 0.456 1.438 1.25

0.025 0.662 0.443 1.430 1.262
0.01 0.660 0.440 1.470 1.291
Fig. 2 Comparison of Cd_av (Filled symbols and dashed line 0.005 0.661 0.441 1.471 1.297
correspond to Cd_av versus AR plot at Re 5 40)

respectively. Table 1 shows the comparison of time averaged drag

coefficient (Cd_av) with various Ds/h ratio at Re ¼ 100 with mesh
size h ¼ 0.01 and it is found that the optimum value of Ds/h ratio
is nearly equal to 0.99.
For grid independence study, different meshes with h ¼ 0.04,
0.025, 0.01, and 0.005, by maintaining Ds/h ¼ 0.99 are generated
and flow over elliptic cylinders of AR ¼ 0.1 and 1.0 are simulated
at Re ¼ 50 and 100. The values of time averaged drag coefficient
(Cd_av) for each case is compared in Table 2. From Table 2, one
can observe that there is not much variation in the value Cd_av af-
ter h ¼ 0.01. Hence, for all subsequent simulations, the value of h
is chosen as 0.01 and the corresponding number of Lagrangian
points as 315.

4 Results and Discussions

4.1 Onset of Vortex Shedding. As the fluid passes the cylin-
der, along the rear surface, the flow separates and vortices will
start shedding from the cylinder. For circular cylinder (i.e.
AR ¼ 1.0) the value of critical Reynolds number when the vortices
begin to shed is 47 [2]. Therefore there is a vortex shedding
behind circular cylinder (AR ¼ 1) for Re  50, and in the case of
Fig. 3 Comparison of rear axis velocity behind the circular cyl- elliptic cylinder, for a particular Re, shedding can be observed
inder (AR 5 1.0) at Re 5 40 only if AR > ARcrs (Critical AR for vortex shedding) [32].
The critical AR can be identified using the time history of the
coefficient of lift (CL) which is shown in Fig. 4. From Fig. 4(a), it
3.1 Validation. The code developed based on the above is evident that the value of critical AR, when the vortices begin to
scheme was validated for fluid flow characteristics by Sudhakar shed at Re ¼ 50 is 0.9, whereas for Re ¼ 100 (Fig. 4(b)), it is 0.6.
and Vengadesan [27,28]. The buoyancy driven flow was also vali- Likewise, the ARcrs value is found for all the Re, and it is plotted
dated in Raman et al. [29]. Further in this work, fluid flow over in Fig. 5. The power fit of the ARcrs value against Re from the
circular cylinder (AR ¼ 1.0) is simulated at Re ¼ 20, 40, 50, 100, data shown in Fig. 5 is given in Eq. (16)
and 150 and the value of time averaged drag coefficient (Cd_av) is 3
compared with values available in literature [1,7,30,31]. The flow ARcrs ¼ 17:5Re4 (16)
over elliptic cylinder is also tested for Re ¼ 40 and the value of
Cd_av is compared with Faruquee et al. [18]. Figure 2 depicts the Figure 6 shows the vorticity contours for various ARs at
comparison of Cd_av obtained using present simulations and litera- AR ¼ ARcrs and AR > ARcrs. From these plots, one can observe
ture values. A good agreement between the values is observed. that there is no vortex shedding for AR  ARcrs, but for
Figure 3 shows the validation of rear axis velocity at the wake AR > ARcrs, vortex shedding is noticed. Thus, using Fig. 5 and Eq.
center line for circular cylinder at Re ¼ 40. (16), one can predict the value of ARcrs for a particular Re.

3.2 Grid Independence Study. The optimum number of 4.2 Standing Wake. When AR < ARcrs, a standing wake is
Lagrangian points representing the solid surface is found by simu- formed behind the cylinder. The standing wake consists of a pair
lating the flow over circular cylinder (AR ¼ 1.0) with various of counter rotating vortices and it is characterized by the wake
number of Lagrangian points viz. 160, 190, 315, and 365, length (Lw). As AR is decreased from ARcrs, the wake length also
which correspond to the Ds/h ratio of 2.0, 1.67, 0.99, and 0.86 decreases and at some point Lw reaches zero. At that point, the

111201-4 / Vol. 135, NOVEMBER 2013 Transactions of the ASME

Downloaded From: http://fluidsengineering.asmedigitalcollection.asme.org/ on 02/10/2014 Terms of Use: http://asme.org/terms

 Fig. 4 Time history of CL (a) Re 5 50; (b) Re 5 100

4.3 Rear Axis Velocity. Figure 9 shows the rear axis velocity
behind the elliptic cylinders of AR  ARcrs. The central axis line
in the downstream of the cylinder separates the two counter rotat-
ing vortices formed above and below the axis. Therefore the axial
velocities along the central line provide significant data about the
wake. Usually the flow inside the wake is towards the cylinder
and outside the wake is away from the cylinder. Hence the flow
inside the wake changes their direction and returns towards the
cylinder and it will have maximum returning velocity at some
point in the central line. The point where the velocity of the cen-
tral line equals to zero is known as the wake stagnation point.
Figures 9(a) and 9(b) depict that all the cylinders with different
AR that produce wake have a similar rear axis profile, but the vari-
ation of streamwise velocity depends strongly on the AR. Initially,
the direction of velocity is towards the cylinder surface (flow
reversal) and its magnitude increases until it reaches maximum
returning velocity point. After that, the velocity starts increasing
towards the downstream direction and reaches zero at wake stag-
nation point. Further downstream, the velocity reaches the free-
stream velocity for all AR slowly at a sufficiently long distance.
The magnitude of the returning velocity decreases with decreas-
ing AR and eventually it becomes zero at AR ¼ 0.2 for Re ¼ 50
Fig. 5 Effect of Re on ARcrs (Fig. 9(a)), whereas at AR ¼ 0.1 for Re ¼ 100 (Fig. 9(b)). This sig-
nifies that there are no vortices behind the cylinders beyond these
AR. Hence for Re ¼ 50, wake is observed only if AR  0.3,
flow is attached at the rear surface of the cylinder and the AR at
whereas for Re ¼ 100 the value of AR is found as 0.2, which again
which Lw becomes zero is known as the critical AR for wake
emphasizes the results obtained through Eqs. (17) and (18). The
length, (ARcrw). Figure 7 depicts the different flow regime behind
distance between the wake stagnation point and surface of the
the elliptic cylinder. The wake behind the cylinder is observed
cylinder decreases with decreasing AR. On the other hand, the
even for AR ¼ 0.1 when Re > 100, hence Re ¼ 50 and 100 cases
maximum returning velocity and distance between wake stagna-
are considered for the ARcrw calculation. Figure 8 shows the plot
tion point and cylinder surface increases with increasing AR.
between Lw / dH (nondimensionalised Lw) and AR for Re ¼ 50 and
100. The quadratic curve fit obtained from the data shown in
Fig. 8 for Re ¼ 50 and 100 is given by Eqs. (17) and (18), 4.4 Coefficient of Drag. The drag forces acting on the cylin-
respectively der is due to the skin friction and pressure distribution acting on
the surface of the cylinder. For AR < ARcrw, the drag force
Lw =dH ¼ 2:604AR2 þ 0:847AR  0:471 for Re ¼ 50 (17) depends only on the skin friction and for AR > ARcrw, drag coeffi-
cient depends on both skin friction and pressure force. Figure 10
Lw =dH ¼ 10:615AR2  3:171AR þ 0:228 for Re ¼ 100 (18) shows the mean drag coefficient (Cd_av) values for different AR
and Reynolds number. It can be observed from Fig. 10 that
From these equations, the value of ARcrw for Re ¼ 50 and 100 is increasing the AR increases the value of Cd_av. However, there is a
found to be 0.29 and 0.18, respectively. For Re ¼ 100, the quad- sudden jump in the value of Cd_av at AR ¼ ARcrw, which may be
ratic equation (Eq. (18)) will not give the correct value of Lw/dH due to the fact that when AR > ARcrw, the pressure drag created by
for AR > 0.6, because vortex shedding is observed behind the cyl- the wake plays a significant role. The value of Cd_av decreases
inder of AR > 0.6. with increasing Re till AR ¼ 0.8, because increasing the Re

Journal of Fluids Engineering NOVEMBER 2013, Vol. 135 / 111201-5

Downloaded From: http://fluidsengineering.asmedigitalcollection.asme.org/ on 02/10/2014 Terms of Use: http://asme.org/terms

 Fig. 6 Instantaneous vorticity contours at (left side AR 5 ARcrs, right side AR > ARcrs,) (a) Re 5 100, AR 5 0.6 (b) Re 5 100.
AR 5 0.7 (c) Re 5 200, AR 5 0.4 (d) Re 5 200. AR 5 0.5 (e) Re 5 400, AR 5 0.2 (f) Re 5 400. AR 5 0.3.

decreases skin friction as well as the wake size. However, the Cd av ¼ 0:950Re0:067 for AR¼ 1:0 (20)
value of Cd_av increases with increasing Re when AR > 0.8. This
may be because at AR > 0.8, the bluff body nature of the cylinder
becomes prominent causing an increase in Cd_av value with The value of Cd_av is normalized using Cdo, where Cdo is mean
increasing Re. drag coefficient corresponding to cylinder of AR ¼ 1.0, for
A functional form of Cd_av can be obtained by normalizing AR  ARcrs and for AR  ARcrs, it is AR ¼ 0.1. The value of Cdo is
the value of Cd_av using a suitable function of Re ði:e:Þ f ðReÞ. calculated using Eqs. (20) and (19) for AR  ARcrs and
Figure 11 depicts the graph between Cd_av and Re for AR ¼ 0.1 AR  ARcrs, respectively. The plot between the normalized value
and 1.0 and from the graph one can observe that the values fit into of the Cd_av using Cdo against AR is given in Fig. 12.
a power fit profile as given below From Fig. 12, one can see that two groups of regimes based on
Cd_av/Cdo are formed, one for AR  ARcrs and another one for
AR  ARcrs. An approximate function for the two groups are given
Cd av ¼ 6:164Re0:57 for AR¼ 0:1 (19) in Eqs. (21) and (22).

111201-6 / Vol. 135, NOVEMBER 2013 Transactions of the ASME

Downloaded From: http://fluidsengineering.asmedigitalcollection.asme.org/ on 02/10/2014 Terms of Use: http://asme.org/terms

 Fig. 7 Representation of different flow regime

Fig. 8 Effect of AR on wake length

Fig. 9 Effect of AR on rear axis velocity (a) Re 5 50 (b)

Cd av =Cdo ¼ 0:73 þ 2:34AR þ 0:57AR2  1:55AR3 Re 5 100
for AR  ARcrs ðCdo is Eq: ð19ÞÞ (21)

¼ 0:017 þ 0:61AR þ 0:74AR2  0:345AR3 the vortex shedding is characterized by using Strouhal number
Cd av =Cdo
(St)
for AR  ARcrs ðCdo is Eq: ð20ÞÞ (22)
f  dH
St ¼ (24)
The Cd_av value for flow over flat plate can be obtained by insert- u1
ing AR ¼ 0.0 (b ¼ 0.0) in Eq. (21) and Table 3 compares Cd_av for
AR ¼ 0.0 with the value obtained by the correlation given in Williamson and Brown [6] reported various St–Re relationship for
Eq. (23) reported by Dennis and Dunwoody [33] circular pcylinder
ffiffiffiffiffiffi (AR ¼ 1.0) wake and also depicted that three
term 1= Re series yields excellent precision
pffiffiffiffiffiin
ffi representing the
1:328 4:12 data. The Eq. (25) shows the three term 1= Re series for circular
Cd av ¼ pffiffiffiffiffiffi þ (23) cylinder reported by Williamson and Brown [6]
Re Re
1:1129 0:4821
The wetted area considered in Eq. (23) is only one side of the St ¼ 0:2731  pffiffiffiffiffiffi þ (25)
plate. From Table 3 we can observe a reasonable agreement with Re Re
the literature value.
Figure 13 shows the comparison of St of present results at
AR ¼ 1.0 with Eq. (25) for various Re and it shows a good agree-
4.5 Strouhal Number. Section 4.1 illustrates that, when ment with the Eq. (25). The plot between St number and AR for
AR > ARcrs, there will be vortex shedding behind the cylinder and different Re is depicted in Fig. 14. From this figure, it is evident

Journal of Fluids Engineering NOVEMBER 2013, Vol. 135 / 111201-7

Downloaded From: http://fluidsengineering.asmedigitalcollection.asme.org/ on 02/10/2014 Terms of Use: http://asme.org/terms

 Table 3 Comparison of Cd_av value obtained for flat plate
(AR 5 0.0)

Re Dennis and Dunwoody [33] (Eq. (23)) Present case (Eq. (21))

100 0.3480 0.3212

200 0.2290 0.2200
300 0.1808 0.1705
400 0.1534 0.1501
500 0.1352 0.1314

Fig. 10 Effect of AR and Re on mean drag coefficient

Fig. 13 Comparison of St for circular cylinder (AR 5 1.0) for

various Re

Fig. 11 Cd_av values for different Re at AR 5 0.1 and 1.0

Fig. 14 Effect of AR and Re on Strouhal number

that the value of Strouhal number decreases with increase in AR

for any particular Re and increases with increase in Re for any
particular AR. A functional relationship for Strouhal number can
be obtained by plotting the value of St  AR=St0 against AR, where
St0 is the value of Strouhal number at AR ¼ 1.0 calculated using
Eq. (25). Figure 15 shows the plot between St  AR=St0 against
AR. From the figure, it is observed that all the values fall into a
single curve and functional form of St  AR=St0 value in terms of
Fig. 12 Cd_av/Cdo values for different AR AR is obtained as given in Eq. (26).

111201-8 / Vol. 135, NOVEMBER 2013 Transactions of the ASME

Downloaded From: http://fluidsengineering.asmedigitalcollection.asme.org/ on 02/10/2014 Terms of Use: http://asme.org/terms

 Nomenclature
a¼ semimajor axis, m
b¼ semiminor axis, m
AR ¼ axis ratio (AR ¼ b/a)
ARcrs ¼ critical AR for vortex shedding
ARcrw ¼ critical AR for wake length
Cd_av ¼ mean drag coefficient
dH ¼ hydraulic diameter, m
f¼ frequency of the vortex shedding, s1
f¼ velocity forcing term at Eulerian points
h¼ Eulerian mesh width, m
Lw ¼ wake length, m
M¼ number of Lagrangian points
p¼ dimensionless pressure
Re ¼ Reynolds number (Re ¼ q U1 dH/l)
St ¼ Strouhal number (St ¼ f  dH =u1 )
t¼ dimensional time, s
u¼ dimensionless velocity vector
x¼ dimensionless Cartesian coordinates
Fig. 15 St  AR=St0 values for various AR
Greek Symbols
Table 4 Comparison of Strouhal number Dh ¼ discretized quantity
Ds ¼ distance between two successive Lagrangian points, m
Mittal and Kim and Present d¼ Dirac d function
Jackson Balachandar Sengupta (Eq. (26))
l¼ absolute viscosity of fluid, kgm1 s1
Re, AR [13] [34] [35]
¼ kinematic viscosity of fluid, m2 s1
AR ¼ 0.5, Re ¼ 141.44 0.25580 – – 0.260 q¼ density of fluid, kgm3
AR ¼ 0.5, Re ¼ 525 – 0.21 – 0.208 s¼ dimensionless time
AR ¼ 0.6, Re ¼ 200 – – 0.256 0.254 /¼ hat function
AR ¼ 0.6, Re ¼ 400 – – 0.290 0.285
AR ¼ 0.8, Re ¼ 200 – – 0.215 0.210
AR ¼ 0.8, Re ¼ 400 – – 0.240 0.239 Subscripts
l ¼ variables at Lagrangian point
Note: References cited in the table are [13,34,35] v ¼ variables at the virtual point
1 ¼ variables at the ambient condition

St  AR=St0 ¼ 2:365AR3  4:618AR2 þ 3:382AR  0:105 (26) Superscripts

Table 4 compares the value of Strouhal number obtained using n1 ¼ variables at old time level
Eq. (26) and the values from the literature. We can observe a good n¼ variables at the current time level
agreement with the literature value. nþ1 ¼ variables at the next time level
*,** ¼ intermediate values
5 Conclusions
In the present paper, an attempt is made to study the effect of References
AR on fluid flow characteristics over an elliptic cylinder. An IBM [1] Tritton, D. J., 1959, “Experiments on the Flow Past a Circular Cylinder at Low
based Navier–Stokes solver is used for the simulations, which Reynolds Numbers,” J. Fluid Mech., 6, pp. 547–567.
reduces the complexities that arise due to grid generation. There- [2] Zdravkovich, M. M., 1997, Flow Around Circular Cylinders, Vol. 1: Funda-
mentals, Oxford University Press, New York.
fore, a single Cartesian grid is used effectively in all the cases and [3] Williamson, C. H. K., 1996, “Vortex Dynamics in the Cylinder Wake,” Ann.
the following conclusions are gleaned: Rev. Fluid Mech., 28, pp. 477–539.
[4] Okajima, A., 1982, “Strouhal Numbers of Rectangular Cylinders,” J. Fluid
(1) the onset of vortex shedding for different Re is studied and Mech., 123, pp. 379–398.
critical AR for vortex shedding (ARcrs) for different Re is [5] Sharma, A., and Eswaran, V., 2004, “Heat and Fluid Flow Across a Square
found. Cylinder in the Two-Dimensional Laminar Flow Regime,” Numer. Heat Trans-
fer, Part A, 45, pp. 247–269. pffiffiffiffiffiffi
(2) the critical AR for wake to form behind the cylinder (ARcrw) [6] Williamson, C. H. K., and Brown, G. L., 1998, “A Series in 1= Re to Repre-
for Re ¼ 50 and 100 is found as 0.29 and 0.18, respectively. sent the Strouhal–Reynolds Number Relationship of the Cylinder Wake,”
(3) the value of Cd_av increases with increasing AR, but J. Fluids Struct., 12(8), pp. 1073–1085.
decreases with increasing Reynolds number. [7] Henderson, R. D., 1995, “Details of the Drag Curve Near the Onset of Vortex
Shedding,” Phys. Fluids, 7(9), pp. 2102–2104.
(4) the strouhal number value decreases with increasing AR, [8] Lugst, H. J., and Haussling, H. J., 1974, “Laminar Flow Past an Abruptly
but increases with increasing Reynolds number. Accelerated Elliptic Cylinder at 45 deg Incidence,” J. Fluid Mech., 65, pp.
(5) a functional relationship for ARcrs, ARcrw, Cd_av and St are 711–734.
found for flow over an elliptic cylinder of various AR and at [9] Tandea, S., 1977, “Unsteady Separated Flow Around Bodies,” Prog. Aerosp.
Sci., 17, pp. 287–348.
different Re. [10] Patel, K. A., 1981, “Flow Around the Impulsively Started Elliptic Cylinder at
Various Angles of Attack,” Comput. Fluids, 9, pp. 167–175.
[11] Ota, T., and Nishiyama, H., 1986, “Flow Around Two Elliptic Cylinders in
Acknowledgment Tandem Arrangement,” ASME J. Fluids Eng., 108, pp. 98–103.
We acknowledge the financial support extended by Aerody- [12] Ota, T., Nishiyama, H., and Taoka, Y., 1987, “Flow Around an Elliptic Cylin-
der in the Critical Reynolds Number Regime,” ASME J. Fluids Eng., 109, pp.
namics panel of Aeronautics Research and Development Board, 149–155.
Government of India, Grant No. DARO/2008/1031516/M/I dated [13] Jackson, C. P., 1987, “A Finite Element Study of the Onset of Vortex Shedding
29/07/09. in Flow Past Variously Shaped Bodies,” J. Fluid Mech., 182, pp. 23–45.

Journal of Fluids Engineering NOVEMBER 2013, Vol. 135 / 111201-9

Downloaded From: http://fluidsengineering.asmedigitalcollection.asme.org/ on 02/10/2014 Terms of Use: http://asme.org/terms

 [14] Park, J. K., Park, S. O., and Hyun, J. M., 1989, “Flow Regimes of Unsteady [25] Van der Vorst, H. A., 2000, “Iterative Methods for Large Linear Systems,”
Laminar Flow Past a Slender Elliptic Cylinder at Incidence,” Int. J. Heat Fluid Tech. Report, Mathematical Institute, Utrecht University, Utrecht, The
Flow, 10, pp. 311–317. Netherlands.
[15] Badr, H., 1994, “Oscillating Viscous Flow Over an Inclined Elliptic Cylinder,” [26] Guy, R. D., and Hartenstine, D. A., 2010, “On the Accuracy of Direct Forcing
Ocean Eng., 21, pp. 401–426. Immersed Boundary Methods with Projection Methods,” J. Comput. Phys., 229,
[16] D’Alessio, S. J. D., and Kocabiyik, S., 2001, “Numerical Simulation of the pp. 2479–2496.
Flow Induced by a Transversely Oscillating Inclined Elliptic Cylinder,” J. Flu- [27] Sudhakar, Y., and Vengadesan, S., 2010, “Flight Force Production by Flapping
ids Struct., 15, pp. 691–715. Insect Wings in Inclined Stroke Plane Kinematics,” Comput. Fluids, 39, pp.
[17] Kocabiyik, S., and D’Alessio, S. J. D., 2004, “Numerical Study of Flow Around 683–695.
an Inclined Elliptic Cylinder Oscillating in Line With an Incident Uniform [28] Sudhakar, Y., and Vengadesan, S., 2012, “Vortex Shedding Characteristics of a
Flow,” Eru. J. Mech. B/Fluids, 23, pp. 279–302. Circular Cylinder With an Oscillating Wake Splitter Plate,” Comput. Fluids,
[18] Faruquee, Z., Ting, D. S.-K., Fartaj, A., Barron, R. M., and Carriveau, R., 2007, 53, pp. 40–52.
“The Effects of Axis Ratio on Laminar Fluid Flow Around an Elliptical Cylin- [29] Raman, S. K., Prakash, K. A., and Vengadesan, S., 2012, “Natural Convection
der,” Int. J. Heat Fluid Flow, 28, pp. 1178–1189. from a Heated Elliptic Cylinder With Different Axis Ration in a Square
[19] Mittal, R., and Iaccarino, G., 2005, “Immersed Boundary Methods,” Ann. Rev. Enclosure,” Numer. Heat Transfer Part A, 62, pp. 639–658.
Fluid Mech., 37, pp. 239–261. [30] Lange, C., Durst, F., and Breuer, M., 1998, “Momentum and Heat Transfer
[20] Kim, J., Kim, D., and Choi, H., 2001, “An Immersed-Boundary Finite-Volume from Cylinders in Laminar Crossflow at 104  Re  200,” Int. J. Heat Mass
Method for Simulations of Flow in Complex Geometries,” J. Comput. Phys., Transfer, 41(22), pp. 3409–3430.
171, pp. 132–150. [31] Dennis, S. C. R., and Chang, G.-Z., 1970, “Numerical Solutions for Steady
[21] Pacheco, J. R., Pacheco-Vega, A., Rodic, T., and Peck, R. E., 2005, “Numerical Flow Past a Circular Cylinder at Reynolds Numbers up to 100,” J. Fluid Mech.,
Simulation of Heat Transfer and Fluid Flow Problems Using an Immersed- 42, pp. 471–489.
Boundary Finite-Volume Method on Non-Staggered Grids,” Numer. Heat [32] Raman, S. K., 2012, “Fluid flow and Heat Transfer Over an Elliptic
Transfer, Part B, 48, pp. 1–24. Cylinder Using Immersed Boundary Method,” M.S. thesis,
[22] Su, S. W., Lai, M. C., and Lin, C. A., 2007, “An Immersed Boundary Tech- Department of Applied Mechanics, Indian Institute of Technology,
nique for Simulating Complex Flows with Rigid Boundary,” Comput. Fluids, Madras, India.
36, pp. 313–324. [33] Dennis, S. C. R., and Dunwoody, J., 1966, “The Steady Flow of a Viscous Fluid
[23] Brown, D. L., Cortez, L., and Minion, M., 2001, “Accurate Projection Methods Past a Flat Plate,” J. Fluid Mech., 24, pp. 577–595.
for the Incompressible Navier–Stokes Equations,” J. Comput. Phys., 168, pp. [34] Mittal, R., and Balachandar, S., 1996, “Direct Numerical Simulation of Flow
464–499. Past Elliptic Cylinders,” J. Comput. Phys., 124, pp. 351–367.
[24] Shin, S. J., Huang, W. X., and Sung, H. J., 2008, “Assessment of Regularized [35] Kim, M. S., and Sengupta, A., 2005, “Unsteady Viscous Flow Over Elliptic
Delta Functions and Feedback Forcing Schemes for an Immersed Boundary Cylinder at Various Thickness With Different Reynolds Number,” J. Mech. Sci.
Method,” Int. J. Numer. Methods Fluids, 58, pp. 263–286. Technol., 19, pp. 877–486.

111201-10 / Vol. 135, NOVEMBER 2013 Transactions of the ASME

DownloadedViewFrom:
publicationhttp://fluidsengineering.asmedigitalcollection.asme.org/
stats on 02/10/2014 Terms of Use: http://asme.org/terms