Unsteady wake dynamics and heat transfer in forced and mixed convection past

a circular cylinder in cross ow for high Prandtl numbers

Sandip Sarkar
a
, Amaresh Dalal
b
, G. Biswas
c,
a
Research and Development Division, Tata Steel, Jamshedpur 831 001, India
b
Department of Mechanical Engineering, Indian Institute of Technology Guwahati, Guwahati 781 039, India
c
Department of Mechanical Engineering, Indian Institute of Technology Kanpur, Kanpur 208 016, India
a r t i c l e i n f o
Article history:
Received 18 August 2010
Received in revised form 14 February 2011
Accepted 4 March 2011
Available online 12 April 2011
Keywords:
Mixed convection
Circular cylinder
Vortex shedding
Prandtl number
Aiding buoyancy
Finite element method
a b s t r a c t
This paper investigates the combined effect of Prandtl number and Richardson number on the wake
dynamics and heat transfer past a circular cylinder in crossow using a SUPG based nite element
method. The computations are carried out for 80 < Re < 180, 0.7 < Pr < 100 and 0 6 Ri 6 2. The results
have been presented for both forced and mixed convection ows. In the case of forced convection, crowd-
ing of temperature contours with reduced spatial spread is observed for increasing Prandtl numbers. The
local and average Nusselt numbers are found to increase with increasing Reynolds number and Prandtl
number. The average Nusselt number and Colburn factor are found to vary as Re
0.548
Pr
0.373
and
Re
0.452
, respectively. The extrapolated results of the average Nusselt number for low and high Reynolds
numbers are found to match quite well with the available results in literature. Effect of Prandtl number
shows various interesting phenomena for the mixed convective ows. Increasing the Prandtl numbers
resulted in decreasing deection and strength in the wake structures. The effect of baroclinic vorticity
production during vortex shedding has been demonstrated at the vicinity of the cylinder and near wake.
The Strouhal number is found to decrease with increasing Prandtl number, in the case of buoyancy
induced ow. The effect of increasing Prandtl number is manifested as the stabilizing effect in the ow.
This is, perhaps, the rst time that such behavior for the Prandtl number is being reported. Additionally it
is observed that the average Nusselt number decreases with increasing Richardson number.
2011 Elsevier Ltd. All rights reserved.
1. Introduction
Flow past a circular cylinder and the phenomenon of vortex
shedding have been the subject of voluminous computational
and experimental studies. A substantial body of work has been
devoted for understanding the spatial structure of cylinder wake,
vortex shedding and other wake ows. Furthermore several studies
have been undertaken to understand the heat transfer phenome-
non at different ow conditions. In such studies, the momentum
and heat transfer under cross ow situations have been identied
as the subjects of considerable theoretical and practical impor-
tance. The theoretical and numerical models are practically ideal-
izations of several industrially important applications such as
heat exchanger tubes, chimney stacks, cooling towers, measuring
probes and sensors. To maintain better product quality and higher
productivity, sufcient knowledge of heat transfer is very impor-
tant in the application of food processing industries, where food
substances are thermally treated as high Prandtl number uids
(up to Pr = 100).
Literature shows that during the bluff-body ows, formation of
more or less coherent structures is observed at the wake. These
structures are advected downstream for different ow conditions
to form global wake characteristics. Vortex shedding appears in
the form of von Karman Vortex Street when the Reynolds number
exceeds the critical value of 45 (approx). Investigations show that
the vortex shedding process can be disturbed and several new
structures could be found if the ow eld is perturbed by external
forces. Such forces cause the ow to enter in an early transition
stage by forming complex ow structures. One of such forces is
the thermal buoyancy force. The inuence of thermal buoyancy
leads to mixed convective ows. The effect of Prandtl number plays
a signicant role in heat transfer and wake characteristics of mixed
convective ows.
Some numerical studies are available on the effect of Prandtl
number on heat transfer and wake characteristics for both forced
and mixed convective ows. Several studies are available in the
literature using air as the working uid. Sanitjai and Goldstein
[1] determined heat transfer characteristics in forced convection
from circular cylinder in cross ow for 2 10
3
< Re < 9 10
4
and
0017-9310/$ - see front matter 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijheatmasstransfer.2011.03.032

Corresponding author. Present address: Central Mechanical Engineering

Research Institute (CSIR), Durgapur 713 209, India. Tel.: +91 343 2546749; fax:
+91 512 2597656.
E-mail addresses: director@cmeri.res.in, gtm@iitk.ac.in (G. Biswas).
International Journal of Heat and Mass Transfer 54 (2011) 35363551
Contents lists available at ScienceDirect
International Journal of Heat and Mass Transfer
j our nal homepage: www. el sevi er . com/ l ocat e/ i j hmt
7 < Pr < 176. They showed that the distribution of local heat
transfer changes with the above range of Reynolds numbers and
the angular location over the cylinder, but after Re > 5 10
3
the lo-
cal heat transfer increases with the Prandtl number. Juncu [2]
numerically analyzed the forced convective heat transfer around
two tandem circular cylinders at low Reynolds numbers (1 < Re <
30) for various Prandtl numbers (0.1 < Pr < 100). The heat transfer
characteristics for the evolution of the system for RePr > 1 showed
completely different behavior than that of RePr < 1 in their study.
As in the former case, convection dominates over conduction,
while for the latter case conduction dominates over convection.
Perkins and Leppert [3] studied local heat transfer coefcients
from a uniformly heated cylinder using water as a working uid
under crossow situations for 2 10
3
6 Re 6 12 10
4
and 1 6
Pr 6 7. Based on the experiments, they gave several correlations
for local and average Nusselt numbers over the cylinder for both
constant wall temperature and constant wall heat ux conditions.
They also analyzed the effect of the variation of the uid viscosity
and free stream turbulence across the boundary layer on Nusselt
number distributions.
For square cylinders, numerical investigations pertaining to the
effect of Reynolds and Prandtl numbers on heat transfer are avail-
able for both steady and unsteady regime [4,5]. The heat transfer
characteristics for an isolated square cylinder subjected to a low
Reynolds number range of 145 and a very high Prandtl number
range of 0.74000 was reported by Dhiman et al. [6]. In their
numerical study they showed the dependence of both Reynolds
and Prandtl number on heat transfer coefcients. They also gave
correlations on Colburn heat transfer factor as a function of
Reynolds number for both constant wall temperature and constant
wall heat ux case, respectively.
Biswas et al. [7] studied unsteady mixed convection heat trans-
fer in a horizontal channel with a built-in square obstacle using air
as the working uid. Their results show that the mixed convection
can initiate periodicity and asymmetry in the wake at lower
Reynolds numbers, in contrast to forced convection alone. Using
the same medium, Singh et al. [8] studied the inuence of favor-
able and adverse buoyancy in a vertical channel with a built-in
circular cylinder. For Richardson numbers less than 0.15, the ow
was characterized by broadening of the wake. Richardson numbers
greater than 0.15, revealed separation delay and attached twin vor-
tices behind the cylinder. Experimental investigation by Dumou-
chel et al. [9] showed the dependence of effective Reynolds
number and effective temperature to characterize the wake
structure.
The dualism between buoyancy and viscous forces may lead to
various interesting phenomena. For the case of heated cylinder un-
der cross ow conditions it has been numerically found by Biswas
and Sarkar [10] that hydrodynamic instabilities grow and ow be-
comes unsteady periodic if the uid is severely inuenced by ther-
mal buoyancy. It has been reported that angle of separation and
average Nusselt number oscillates with the vortex shedding fre-
quency. Lange et al. [11] numerically calculated momentum and
heat transfer from a circular cylinder in laminar cross ow for
the range 10
4
6 Re 6 200 and for temperature loadings of
1.0031.5. Kieft et al. [12] reported both numerically and experi-
mentally the effect of heat input on the wake characteristics and
vortex structure from a horizontal circular cylinder in cross ow.
At a xed Reynolds number of 75 and by setting the range of Rich-
ardson number 0 to 1, they analyzed the different vortex character-
istics and production of baroclinic vorticity production terms. All
the above mentioned investigations mainly deal with constant Pra-
ndtl number uids.
Another investigation by Kieft et al. [13] described the vortex
shedding process in cross ow past a circular cylinder for a
Reynolds number of 75 and a Grashof number range between 0
and 5000. They reported that the strength of the vortices shed from
the upper half of the cylinder is stronger than those of lower half.
Shi et al. [14] numerically simulated the effect of heating on the
ow and heat transfer in a two dimensional laminar ow past a
circular cylinder, considering the variation of uid properties with
temperature. In the range of 0:001 6 Re 6 170, they showed the
dependence of various overheat ratio on vortex shedding
frequency.
It is obvious from the above discussion that there is a need for
investigating the combined effect of Prandtl and Richardson num-
bers on the wake characteristics and heat transfer from a circular
cylinder in the unsteady range. In the present investigation, the
combined effect of Prandtl and Richardson numbers on wake char-
acteristics and heat transfer for a circular cylinder in crossow is
investigated numerically. Blockage ratio is chosen as D/H=0.05.
For the simulation, a SUPG based Finite Element technique as
[15] is used. Assuming the ow situation to be completely
two-dimensional, the ranges of Reynolds and Prandtl numbers
Nomenclature
D cylinder diameter
f vortex shedding frequency
g acceleration due to gravity
h heat transfer coefcient
Gr Grashof number
gbTwT1D
3
m
2
_ _
j
c
Colburn heat transfer factor
Nu
h
time averaged local Nusselt number on the cylinder sur-
face
Nu
avg
average Nusselt number over the cylinder surface
p non-dimensional pressure
Pr Prandtl number
m
a
_ _
Re Reynolds number
qU1D
l
_ _
Ri Richardson number
Gr
Re
2
_ _
t non-dimensional time
T temperature
u, v non-dimensional velocity components in x and y
directions
U
1
inlet velocity
x, y non-dimensional coordinates
Greek symbols
a thermal diffusivity of a uid
b thermal expansion coefcient
l viscosity of a uid
q density of a uid
H non-dimensional temperature, i.e.
TT1
TwT1
_ _
s time period for a complete cycle
m kinematic viscosity of the uid
h angular location on the cylinder surface from the
forward stagnation point
x vorticity
Subscripts
1 inlet condition
w wall
avg average
S. Sarkar et al. / International Journal of Heat and Mass Transfer 54 (2011) 35363551 3537
are chosen as 80 6 Re 6 180 and 0:7 6 Pr 6 100. At a representa-
tive Reynolds number of 100, the combined effect of Prandtl num-
ber and Richardson number is shown at the same Prandtl number
range, by varying Richardson number between 1:0 6 Ri 6 2:0. For
forced convective ows, the results are shown for twelve Prandtl
numbers corresponding to each Reynolds number. Some of the re-
sults are validated from the existing literature. For forced convec-
tion we propose a new correlation of average Nusselt number by a
power law relationship as Nu
avg
= CRe
m
Pr
n
. Another correlation is
also established on Colburn factor as a function of Reynolds num-
ber. For mixed convective ows, the results are shown for the Pra-
ndtl number range considered. Effects of Prandtl number on vortex
characteristics are analyzed by different vortex identifying param-
eters. Effects of heat input and wake characteristics are described
using Baroclinic vorticity production terms.
g
Freeslip wall
Freeslip wall
O
u
t
f
l
o
w

s
e
c
t
i
o
n
D
I
n
l
e
t

s
e
c
t
i
o
n
v,y
u,x
H
X
o
X
i
L = X
i
+ X
o
Fig. 1. Computational domain.
(a)
(b)
Fig. 2. Typical grid used for computations: (a) full view and (b) close view.
Re
N
u
a
v
g
10 15 20 25 30
2
4
6
8
Juncu (2007), Pr = 1.0
Juncu (2007), Pr = 10.0
Present Calculation, Pr = 1.0
Present Calculation, Pr = 10.0
Re
N
u
a
v
g
10 15 20 25 30
8
10
12
14
16
18
Juncu (2007), Pr = 100.0
Present Calculation, Pr = 100.0
(a)
(b)
Fig. 3. Comparison of average Nusselt number: (a) Pr = 1 and 10 and (b) Pr = 100.
Table 1
Meshes and corresponding numbers of elements and nodes.
Mesh Nodes Elements C
N
M1 23,184 22,832 180
M2 26,520 26,144 200
M3 30,134 29,732 216
Table 2
Grid independence test results.
Mesh St Nu
avg
M1 0.151953 20.45218
M2 0.1529052 20.33673
M3 0.1525628 20.30758
3538 S. Sarkar et al. / International Journal of Heat and Mass Transfer 54 (2011) 35363551
The outline of the rest of this paper is as follows: We start by
dening the model formulation by reviewing the governing equa-
tions for two-dimensional incompressible uid ow and heat
transfer in Section 2. The problem setup is described along with
the boundary and initial conditions. Next, the SUPG based nite
element solution methodology has been referred from the previous
publications of the authors. Starting from the grid generation
methodology, grid independence study and code validation have
been described in detail. In Section 3, computational results for
ow and heat transfer for the forced convection case are presented
and discussed. Simulations for the ow and heat transfer for the
mixed convection case are analyzed and represented in Section 4.
In Section 5, a few concluding remarks are made.
2. Formulation of the problem
2.1. Governing equations and boundary conditions
The two-dimensional incompressible NavierStokes and energy
equations in the Cartesian coordinate system form the governing
equations of the ow. The buoyancy-driven ow from the heated
surface interacts with the laminar main ow to yield mixed con-
vection conditions. The dimensionless equations for continuity,
momentum and energy with the Boussinesq approximation may
be expressed in the following form:
@u
@x

@v
@y
0 1
X
Y
10 12 14 16 18
8
9
10
11
12
X
Y
10 12 14 16 18
8
9
10
11
12
X
Y
10 12 14 16 18
8
9
10
11
12
X
Y
10 12 14 16 18
8
9
10
11
12
X
Y
10 12 14 16 18
8
9
10
11
12
X
Y
10 12 14 16 18
8
9
10
11
12
X
Y
10 12 14 16 18
8
9
10
11
12
X
Y
10 12 14 16 18
8
9
10
11
12
(a) (b)
(c) (d)
(e) (f)
(g) (h)
Fig. 4. Instantaneous contours of H for Re = 120 and Pr = 30 at (a) t, (c) t
s
4
, (e) t
2s
4
, and (g) t
3s
4
. Instantaneous contours of streamlines for Re = 120 and Pr = 30 at (b) t, (d)
t
s
4
, (f) t
2s
4
, and (h) t
3s
4
.
S. Sarkar et al. / International Journal of Heat and Mass Transfer 54 (2011) 35363551 3539
@u
@t
u
@u
@x
v
@u
@y

@p
@x

1
Re
@
2
u
@x
2

@
2
u
@y
2
_ _
2
@v
@t
u
@v
@x
v
@v
@y

@p
@y

1
Re
@
2
v
@x
2

@
2
v
@y
2
_ _
RiH 3
@H
@t
u
@H
@x
v
@H
@y

1
RePr
@
2
H
@x
2

@
2
H
@y
2
_ _
4
where Richardson number Ri gbDHD=U
2
1
. Here, D is the length
scale and U
1
is the velocity scale of the problem under consider-
ation. Fig. 1 shows the computational domain used in this investiga-
tion. Conforming to a blockage ratio of 0.05, the corresponding
distance between the top and bottom boundaries is 20D.
At the inlet, which is located at X
i
= 10D upstream of the center
of the cylinder, a uniform ow was prescribed (u = 1, v = 0, H = 0).
At the outlet, which is located at X
o
= 20D downstream of the cen-
ter of the cylinder, a homogeneous Neumann boundary condition
for the velocity components (u and v) and temperature (H) was
used. No-slip conditions were prescribed on the cylinder. On the
cylinder surface, uniform temperature (H = 1) was prescribed. At
the upper and lower conning walls, symmetry conditions simu-
lating a frictionless wall @u=@y v 0 and zero heat ux
@H=@y 0 were used. The normal derivative of pressure was
set to zero on all boundaries except at the outlet where pressure
equals the local ambience and this way the Dirichlet condition
was set.
2.2. Finite element formulations
To solve the governing equations numerically, a stable nite
element method was used. The Eulerian velocity correction ap-
proach, based on the Projection scheme identical with the Marker
and Cell (MAC) method of Harlow and Welch [16] was used to
solve the governing equations for the mass, momentum and en-
ergy. An explicit time-stepping numerical technique with the
Streamline Upwind PetrovGalerkin (SUPG) formulation [17] was
adopted. This present work is an extension of the earlier work of
Maji and Biswas [15], Biswas and Sarkar [10] and the detail formu-
lation has been discussed there. The method has been successfully
applied to solve complex problems [10,1819].
2.3. Grid generation and code validation
The computational domain is discretized into small curvilinear
four-noded elements. The grid is generated based on the multi
block technique using the transnite interpolation (algebraic
method) further smoothed by a partial differential equation of
elliptic type. The velocity components, pressure and temperature
are collocated at each node of the element. A typical grid is shown
X
Y
10 12 14 16
9
10
11
12
X
Y
10 12 14 16
9
10
11
12
X
Y
10 12 14 16
9
10
11
12
X
Y
10 12 14 16
9
10
11
12
X
Y
10 12 14 16
9
10
11
12
X
Y
10 12 14 16
9
10
11
12
X
Y
10 12 14 16
9
10
11
12
X
Y
10 12 14 16
9
10
11
12
X
Y
10 12 14 16
9
10
11
12
Pr = 1.0 Pr = 40.0 Pr = 80.0
(a) Re=80
Pr = 1.0 Pr = 30.0 Pr = 60.0
(b) Re=140
Pr = 1.0 Pr = 50.0 Pr = 100.0
(c) Re=180
Fig. 5. Time-averaged isotherms at different Prandtl numbers: (a) Re = 80, (b) Re = 140 and (c) Re = 180 (close-up view).
3540 S. Sarkar et al. / International Journal of Heat and Mass Transfer 54 (2011) 35363551
in Fig. 2a. The grid is rened around the cylinder (see Fig. 2b) and
made a little coarse in the far eld.
This SUPG-based nite element code has been tested satisfacto-
rily on a number of benchmark problems, such as the ow in a lid-
driven cavity, the ow over a backward-facing step, the ow in a
buoyancy-driven cavity and ow over a circular cylinder in cross
ow under inuence of buoyancy. The results are presented else-
where [10,18] and they are found satisfactory. In the present work,
some of the results are validated with available literature [2], nd-
ing very good matches. Fig. 3a and b shows the comparison results
of the variation of average values of Nusselt number, Nu
avg
with
Reynolds number for different Prandtl numbers. The present
results are found to be in good agreement with the literature.
2.4. Grid independence test
The grid independence test was carried out with respect to the
Strouhal number and time-average values of average Nusselt num-
bers at a xed Reynolds number of 100, Prandtl number of 30 and
Richardson number of 0. Corresponding to the xed blockage ratio
of 0.05, three different meshes were generated viz. M1, M2 and M3.
Table 1 shows the meshes and corresponding numbers of elements
and nodes used in this study. Here C
N
is the numbers of nodes over
the cylinder surface. The results of grid independence have been
presented in Table 2.
From Table 2 it is evident that Strouhal number shows varia-
tions of 0.22%, whereas time average Nusselt number has 0.14%
with meshes M2 and M3. So for the present computations, the
grid-independent situation is achieved for mesh M2, which is com-
putationally economical for all different cases studied in the pres-
ent investigation.
3. Flow and heat transfer for the forced convection case
The nite element simulations were carried out for Re = 80180
in the steps of 20 and Pr = 0.7, 1 and 10100 in the steps of 10. As

N
u

0 30 60 90 120 150 180

0
10
20
30
40
Pr = 1.0
Pr = 20.0
Pr = 40.0
Pr = 50.0

N
u

0 30 60 90 120 150 180

0
20
40
60
Pr = 1
Pr = 10
Pr = 30
Pr = 40
(a)
(b)
Fig. 6. Variation of time-averaged local Nusselt number on the cylinder surface for
different Prandtl numbers: (a) Re = 80 and (b) Re = 140.
Table 3
Nu
avg
for different Pr and at a xed Re = 160.
Pr Nu
avg
10 17.7147
30 26.3705
60 34.1832
90 39.7764
Pr
N
u
a
v
g
0 20 40 60 80 100
10
20
30
40
Re = 80
Re = 100
Re = 120
Re = 140
Re = 160
Re = 180
Fig. 7. Variation of average Nusselt number with Prandtl numbers for different
Reynolds numbers.
Re
N
u
a
v
g
10 20 30 40
0
10
20
Present, Eq. (7), Pr = 0.1
Juncu (2007), Pr = 0.1
Present, Eq. (7), Pr = 0.7
Biswas and Sarkar (2009), Pr = 0.7
Present, Eq. (7), Pr = 10
Juncu (2007), Pr = 10.0
Present, Eq. (7), Pr = 20.0
Srinivas et al.(2009), Pr = 20.0
Present, Eq. (7), Pr = 50.0
Srinivas et al.(2009), Pr = 50.0
Fig. 8. Comparisons of Nu
avg
extrapolated from the proposed Eq. (7) and numerical
data from literature for 10 6 Re 6 45 and different Prandtl numbers.
S. Sarkar et al. / International Journal of Heat and Mass Transfer 54 (2011) 35363551 3541
described earlier, under forced convective situations, the effect of
Prandtl number strongly inuences the heat transfer characteris-
tics. Here we observe its effect on heat transfer by some non-
dimensional parameters and those are the quantitative parameters
signifying heat transfer, viz., the local Nusselt number and the
average Nusselt number. The results are shown for dynamic steady
state conditions. In order to ensure that the system reached a
dynamic steady state, we carried out the simulations at least for
about 60 periodic cycles till the asymptotic value of vortex
shedding frequency was reached. Typical value for vortex shedding
frequency at Re = 100 is found to be varying asymptotically as
0:151 0:0025 after 33 periodic cycles.
Fig. 4ah shows the instantaneous isotherms and streamlines in
a complete vortex shedding period captured at time instants
t; t
s
4
; t
2s
4
and t
3s
4
at a Reynolds number of 120 and Prandtl
number of 30. It is evident from Fig. 4 that there is less spatial
spreading of isotherms and they are found to be crowded nearer
to the cylinder surface. This typical structure of isotherms
(Fig. 4a, c, e and g) is due to the effect of high Prandtl number.
Whereas the streamlines (Fig. 4b, d, f and h) have usual feature
showing alternate shedding from upper and lower half of the cyl-
inder shoulders.
Fig. 5ac shows the time-averaged isotherms around the heated
cylinder for different Prandtl numbers at Re = 80, 140 and 180. The
time-averaged streamline delineating the wake bubble is incorpo-
rated in the isotherms plots as a thick line. The usual features of
the heat transfer characteristics are explored by these gures. The
isotherms appear to be symmetrical about the line of symmetry
for Pr = 1. It is also evident from these plots that the isotherms are
congregated at the front stagnation region, signifying the highest
rate of heat transfer from the cylinder surface thus having the larg-
est value of Nusselt number. With the increase in Prandtl number,
temperature gradients of isotherms increases around the cylinder
surface, and these are incarcerated to a smaller region. This clearly
suggests that the quantum of the resistance to heat transfer is con-
stricted to a thin layer of uid lump. The same observation was also
reported by Sahu et al. [5], for square cylinders. As the Reynolds
number increases, temperature contours show higher temperature
gradients in the near-wake region and the isotherms cluster [10].
This means that an increasing Reynolds number, signifying a higher
uid velocity, sets a higher temperature gradient, leading to en-
hanced heat transfer from the cylinder surface.
Table 4
Empirical correlation of the average Nusselt number.
Researcher Empirical correlation Range of Re Range of Pr
Sanitjai and Goldstein [1]
Nu
avg
0:446Re
0:5
Pr
0:35
0:5286:5e
Re=5000

5
0:031Re
0:8

1=5
Pr
0:42
8
2 10
3
10
5
0.7176
Sharma and Sukhatme [22]
Nu
avg
0:62Re
0:505
9
12004700
Kramers [23]
Nu
avg
0:42Pr
0:20
0:57Re
0:50
Pr
0:31
10
510
3

Fand [24]
Nu
avg
0:35 0:34Re
0:5
0:15Re
0:58
Pr
0:31
11
0.110
5

Churchill and Bernstein [25] Nu

avg
0:3
0:62Re
1=2
Pr
1=3
1 0:4=Pr
2=3
_ _
1=4
1
Re
282; 000
_ _ _ _
4=5
12 10
2
10
7
RePr > 2
Re
N
u
a
v
g
1000 2000 3000 4000
10
20
30
40
50
Kramers (1946)
Fand (1965)
Sharma and Sukhatme (1977)
Churchill and Bernstein (1977)
Sanitjai and Goldstein (2004)
Khan et al. (2005)
Present, Eq. (7)
Fig. 9. Comparisons of Nu
avg
predicted by various empirical correlations for
1000 6 Re 6 4700 and Pr = 0.7.
Re
j
c
80 100 120 140 160 180
0.04
0.05
0.06
0.07
0.08
Pr = 0.7
Pr = 1.0
Pr = 10.0
Pr = 20.0
Pr = 30.0
Pr = 40.0
Pr = 50.0
Pr = 60.0
Pr = 70.0
Pr = 80.0
Pr = 90.0
Pr = 100.0
Fig. 10. Variation of j
c
with Reynolds numbers for different Prandtl numbers.
3542 S. Sarkar et al. / International Journal of Heat and Mass Transfer 54 (2011) 35363551
The local heat transfer from the cylinder surface to the uid is
quantied by local Nusselt number. The local Nusselt number,
based on cylinder diameter is dened as:
Nu
h

@H
@n

along the cylinder surface

5
where n denotes the direction normal to the cylinder surface and h
is the polar angle. The location h = 0 corresponds to the front stag-
nation point facing the incoming ow. The local Nusselt number
computed in this unsteady range is time-averaged over ten periods
of vortex shedding cycle.
The variation of the time-averaged local Nusselt number over
the cylinder surface is shown in Fig. 6a and b for Re = 80 and 140
for different ranges of Prandtl numbers. It is evident from these
plots that the local Nusselt number increases with increasing
Reynolds numbers, whereas there is a signicant increase in local
Nusselt number distribution near the front stagnation region with
increasing Prandtl numbers. The maximum value of the local
Nusselt number occurs at front stagnation point and decreases
gradually over the cylinder surface. The local Nusselt number
distribution at the rear stagnation region is slightly different and
there is a gradual increase up to the rear stagnation point after
reaching its minimum value nearer to the point of separation. This
can be explained by the isotherm patterns, which show that with
increasing Prandtl numbers, the isotherms get closer to the rear
stagnation region to the axis of symmetry. This indicates higher
temperature gradient and accordingly higher local Nusselt number
distribution.
The average Nusselt number has been calculated by time aver-
aging the spatially averaged local Nusselt number over the cylinder
surface. Table 3 represents the average Nusselt number for differ-
ent Prandtl numbers and at a xed Reynolds number of 160. It can
be seen from Table 3 that the average Nusselt number increases
with increasing Prandtl numbers. This is due to the fact that
increasing Prandtl number sets higher temperature gradients over
the cylinder surface and thereby higher rate of heat transfer.
The variations of average Nusselt number with Prandtl numbers
for different Reynolds numbers is shown in Fig. 7. The average Nus-
selt number increases with increasing Prandtl numbers. For the
present range of Reynolds and Prandtl numbers, it is found that
maximum value of average Nusselt number occurs at Re = 180
and Pr = 100 whereas the minimum value corresponds to Re = 80
and Pr = 0.7.
Khan et al. [20] proposed the following empirical equation for
the determination of average Nusselt number
Nu
avg
0:593Re
0:500
Pr
0:333
6
For the present investigation we have proposed the empirical
relation for average Nusselt number by least square curve t is
as follows:
Nu
avg
0:459Re
0:548
Pr
0:373
for 80 6Re 6180 and 0:7 6Pr 6100
7
The numbers of data points and associated root mean square er-
ror for this t are 66 and 0.03. This equation is in close agreement
with Eq. (6) proposed by Khan et al. [20].

0 120 240 360

-20
-10
0
10
20
Pr = 0.7
Pr = 10.0
Pr = 100.0

0 120 240 360

-20
-10
0
10
20
Pr = 0.7
Pr = 10.0
Pr = 100.0

0 120 240 360

-20
-10
0
10
20
Pr = 0.7
Pr = 10.0
Pr = 100.0
(a) (b)
(c)
Fig. 11. Variation of time-averaged wall vorticity along the cylinder surface for different Prandtl numbers; (a) Ri = 1, (b) Ri = 1.5 and (c) Ri = 2.
S. Sarkar et al. / International Journal of Heat and Mass Transfer 54 (2011) 35363551 3543
In Fig. 8, the variation of Nu
avg
with Re predicted by Eq. (7) is
compared with the numerical data from the literature for the stea-
dy ow10 6 Re 6 45 and for different Prandtl numbers. The pre-
dicted results display good agreement with those of Juncu [2],
Biswas and Sarkar [10] and Srinivas et al. [21].
Based on experimental and analytical studies of ow over circu-
lar cylinder, many researchers proposed correlations for average
Nusselt number. A few correlations for different range of Reynolds
and Prandtl number are presented in Table 4.
The correlations in Table 4 appear to have different power indi-
ces for Re and Pr. These anomalies in indices arise due to experi-
mental limitations, blockage ratios and the range of Reynolds and
Prandtl numbers.
Fig. 9 compares the average Nusselt number (Nu
avg
) extrapo-
lated from Eq. (7) with the formulae given in Eqs. (6), (8)(12).
For a xed Prandtl number of 0.7, the range of Reynolds number
for this comparison is chosen as 10004700. It is found that the
predicted results of Sanitjai and Goldstein [1] under-predict the
average Nusselt number for the Re range considered. Values of
Nu
avg
predicted by the proposed Eq. (7) displays satisfactory agree-
ment with those predicted by Khan et al. [20] and Churchill and
Bernstein [25]. At Re = 4700, Nu
avg
predicted by Eq. (7) is 6.78%
smaller than predicted by Sharma and Sukhatme [22] and 32%
higher compared to the prediction of Sanitjai and Goldstein [1].
The average Nusselt number varies with both Reynolds number
and Prandtl number. It is possible to appease the variations of
Reynolds number and Prandtl number to a single curve by dening
a factor j
c
, namely the Colburn heat transfer factor and which is
dened in the literature [4,5] as:
j
c

Nu
avg
Re Pr
1=3
13
The values of j
c
are computed for the present range of Reynolds and
Prandtl number and the results are presented in Fig. 10. The j
c
factor
is found to be decreasing with increasing Reynolds number and the
variation is exponential. By using the power t, an empirical
relation is obtained for the j
c
factor and described by:
j
c
0:521Re
0:452
for 80 6 Re 6 180 14
The associated root mean square error for this t is 0.06. The
above Eq. (14) can be used as a single correlating equation for heat
transfer at the present range of Reynolds and Prandtl numbers.

N
u

0 30 60 90 120 150 180

0 30 60 90 120 150 180
0
20
40
60
Pr = 0.7
Pr = 10.0
Pr = 50.0
Pr = 100.0
1.25 1.26
62.4
62.6

N
u

0
20
40
60
Ri = 1.0
Ri = 1.5
Ri = 2.0
(a)
(b)
Fig. 12. Variation of time averaged local Nusselt number over the cylinder surface
for (a) Ri = 1.0 and different Prandtl numbers and (b) Pr = 100 and different
Richardson numbers.
100
28.5
28.6
Pr
N
u
a
v
g
0 20 40 60 80 100
10
20
30
Ri = 0.0
Ri = 1.0
Ri = 1.5
Ri = 2.0
Fig. 13. Variation of average Nusselt number with Prandtl numbers for different
Richardson numbers.
Ri
L
f
0 1 2
6
9
12 Pr = 0.7
Pr = 10.0
Pr = 100.0
Pr = 0.7
Pr = 10.0
Pr = 100.0
Upper Vortex
Lower Vortex
Fig. 14. Variation of vortex formation length with Richardson numbers.
3544 S. Sarkar et al. / International Journal of Heat and Mass Transfer 54 (2011) 35363551
4. Flow and heat transfer for the mixed convection case
Effect of thermal buoyancy is induced in the ow eld by
varying Richardson number. The value of Richardson number is
zero for forced convective ow. To investigate the effect of the
Prandtl number on mixed convective ow, we have carried out
simulations at a xed Reynolds number of 100. The range of
Prandtl number and Richardson number are chosen as 0:7 6
Pr 6 100 and 1 6 Ri 6 2 respectively. The combined effect of Rich-
ardson number and Prandtl number affects the wake dynamics sig-
nicantly along with heat transfer. The periodicity in the ow eld
is perturbed and vortex motions are disturbed if the ow eld is
signicantly inuenced by the effect of Prandtl number in buoy-
ancy induced eld. This may nally cause delay in formation of
vortices from boundary layer during the vortex shedding process.
The vortex shedding mechanism is described in detail in Biswas
and Sarkar [10].
The scalar vorticity transport equation for mixed convective
ow can be written as,
@x
@t
u
@x
@x
v
@x
@y

1
Re
@
2
x
@x
2

@
2
x
@y
2
_ _
Ri
@H
@x
15
If we compare the Eq. (15) with a general vorticity transport
equation, it is found that the second term of the right hand side
of the Eq. (15) represents a source term. Here we dene it as
g
b
Ri
@H
@x
16
Kieft et al. [12] dened the term g
b
as the baroclinic production
term, representing a production of vorticity within the ow eld
due to a non-zero temperature gradient perpendicular to the grav-
ity vector. Comparing this to the case of forced convective ow,
addition of heat will cause a change in the total circulation [12].
This circulation is dened as the integrated vorticity over the entire
ow domain [26]. For higher heat induced processes by varying
Richardson and Prandtl number it is expected that stability of vor-
tex street will be inuenced by the vorticity production.
In the present study we will investigate the combined effect of
Prandtl and Richardson number on the vortex wake by considering
two situations: First the near wake dynamics by analyzing wall
vorticity, Nusselt number and vortex formation length. Next, the
far wake dynamics by vortex trajectories, vortex strength. To fur-
ther investigate the effect of Prandtl number, we will analyze the
baroclinic vorticity production in the vicinity to the cylinder by
computing local and total period-averaged ux of vorticity. In the
far wake, we will describe the vortex formation process by com-
puting the contours of g
b
in a complete vortex shedding cycle.
X
Y
15 20 25 30
9
10
11
12
Pr = 0.7, Ri = 1.0
Pr = 10.0, Ri = 1.0
Pr = 100.0, Ri = 1.0
Pr = 0.7, Ri = 1.5
Pr = 10.0, Ri = 1.5
Pr = 100.0, Ri = 1.5
X
Y
15 20 25 30
10
11
12
13
Pr=0.7, Ri=1.0
Pr=10.0, Ri=1.0
Pr=100.0, Ri=1.0
Pr=0.7, Ri=1.5
Pr=10.0, Ri=1.5
Pr=100.0, Ri=1.5
(a)
(b)
Fig. 15. Vortex trajectories for different Prandtl numbers and Richardson numbers:
(a) upper vortex and (b) lower vortex.
X

p
15 20 25 30
1
2
3
Pr = 0.7, Ri = 1.0
Pr = 10.0, Ri = 1.0
Pr = 100.0, Ri = 1.0
Pr = 0.7, Ri = 1.5
Pr = 10.0, Ri = 1.5
Pr = 100.0, Ri = 1.5
X

p
15 20 25 30
-3
-2.5
-2
-1.5
-1
-0.5
Pr = 0.7, Ri = 1.0
Pr = 10.0, Ri = 1.0
Pr = 100.0, Ri = 1.0
Pr = 0.7, Ri = 1.5
Pr = 10.0, Ri = 1.5
Pr = 100.0, Ri = 1.5
(a)
(b)
Fig. 16. Variations of vortex strength along the spatial direction for different
Prandtl numbers and Richardson numbers: (a) Upper vortex and (b) lower vortex.
S. Sarkar et al. / International Journal of Heat and Mass Transfer 54 (2011) 35363551 3545
The effect of Prandtl number on the ow periodicity will be inves-
tigated at the end. It is known that for the mixed convection cases,
at times, early transition to three-dimensionality may creep in
[27]. However, in the present investigation, we have assumed the
ow to be two-dimensional for all the cases.
4.1. Near wake dynamics and heat transfer
4.1.1. Effect of heat on wall vorticity
We carried out the investigations to analyze the effect of Ri and
Pr on vorticity generated at the cylinder surface. For this analysis
we have computed time averaged wall vorticity at the cylinder sur-
face for different Ri and Pr. The time averaged wall vorticity x
h
, is
dened as:
x
h

1
s
_
s
0
xdt 17
Fig. 11 illustrates that the time averaged wall vorticity decreases
sharply over the cylinder from the front stagnation point. Its mini-
mum value occurs at a location of h % 145 but, for Pr = 0.7, the loca-
tion of minima is at h % 120, then increases sharply to its peak value
at h % 210 after that it decreases to a second minima at h % 290 and
then increases slightly. Increasing Richardson number causes the
magnitude of negative vorticity at the upper half of the cylinder
shoulder (0 < h < 140) to decrease. Within the points of separation
(140 < h < 260) the magnitude of positive wall vorticity added to
the ow regime. It has also been found that the magnitude of
negative vorticity at the lower vortex region nearer to the bottom
separation point (260 < h < 360) decreases and positivity increases.
This establishes the fact that heat addition causes the negative wall
vorticity to decrease and positive to increase.
The increase of heat addition causes the magnitude of positive
vorticity to increase and negative vorticity to decrease. Another
important fact is that there is an asymmetry in the vorticity
distribution over the cylinder surface. This asymmetry in vorticity
distribution is due to the different magnitude of heating in the
shear layers generated from the upper and lower surfaces of the
cylinder.
4.1.2. Variation of time-averaged local and average Nusselt number
The variation of time-averaged local Nusselt number (Nu
h
) over
the cylinder surface for Ri = 1.0 and different Prandtl numbers is
Y

L
8 9 10 11 12
-2
0
2
4
6
Pr = 0.7
Pr = 10.0
Pr = 100.0
Y

L
8 9 10 11 12
-2
0
2
4
Pr = 0.7
Pr = 10.0
Pr = 100.0
Y

L
8 9 10 11 12
-2
0
2
4
Pr = 0.7
Pr = 10.0
Pr = 100.0
(a) (b)
(c)
Fig. 17. Variation of c
L
x
along the vertical cross section at x = 11 for different Prandtl numbers; (a) Ri = 1, (b) Ri = 1.5, and (c) Ri = 2.
Ri

T
0 1 2
0
0.4
0.8
1.2
Pr = 0.7
Pr = 10.0
Pr = 100.0
Fig. 18. Variation of c
T
x
with Richardson numbers for different Prandtl numbers.
3546 S. Sarkar et al. / International Journal of Heat and Mass Transfer 54 (2011) 35363551
shown in Fig. 12a. The time-averaged local Nusselt number in-
creases with increasing Prandtl numbers as expected. Fig. 12b de-
picts the variation of time-averaged local Nusselt number for
different Richardson numbers at a Prandtl number of 100. The
observation can be made from Fig. 12b that there is no such signif-
icant increase of time-averaged local Nusselt number distribution
with increasing Richardson numbers.
Fig. 13 shows the variation of average Nusselt number (Nu
avg
)
with Prandtl numbers for different Richardson numbers. It is ob-
served that Nu
avg
increases with increasing Prandtl number. An-
other observation that can be made from Fig. 13 is that there is
a decrease in average Nusselt number when Richardson number
is increased from zero to one and this decrease is higher for high-
er Prandtl numbers. Apparently, increasing Prandtl number re-
duces heat transfer in mixed convection ow. This is due to the
stabilizing effect of Prandtl number on ow. The same observa-
tion was reported by Srinivas et al. [21] in steady parallel ow
situation.
4.1.3. Vortex formation length
The vortex formation length L
f
is dened as the crosswise dis-
tance between the centers of the cylinder to the vortex center just
after its formation. The location of vortex center just after its for-
mation is calculated in the following way. The area, where the local
vorticity peak is brought about, is enclosed by a contour of vorticity
having a value of 0.3x
max
. Over that area, the exact location of vor-
tex center is calculated by doing area weighted average (method of
locating centroid) over a complete vortex shedding cycle. The var-
iation of L
f
with Richardson number is illustrated in Fig. 14 for dif-
ferent Prandtl numbers. Fig. 14 show that the values of L
f
decrease
with increasing Richardson number for upper vortex and increase
for lower vortex. However the formation length increases with
increasing Prandtl number. This causes vortex motion to be re-
tarded and separation delay. At Prandtl number of 10, the magni-
tude of L
f
is maximum for lower vortex and minimum for upper
vortex.
4.2. Far wake dynamics and heat transfer
4.2.1. Vortex trajectories
Vortex trajectories are the pictorial representations of the
motion of the vortex originating from the boundary layer of the
cylinder shoulder. Methods of calculation of vortex trajectories in-
clude vortex peak, complex eigenvalue, vortex centroid and hybrid
tting. In the present investigation vortex trajectories are calcu-
lated using vortex centroid method for about twelve typical shed-
ding periods and shown in Fig. 15a and b. Results are shown for
both upper (Fig. 15a) and lower (Fig. 15b) vortex rows and for dif-
ferent Prandtl and Richardson numbers. Fig. 15 depicts that in
comparison to the lower, the upper vortex has a higher deection.
X
Y
10 12 14 16 18
8
9
10
11
12
Local extreme
Local extreme
+
X
Y
10 12 14 16 18
8
9
10
11
12
Local extreme
Local extreme
+
+
+ +
X
Y
10 12 14 16 18
8
9
10
11
12
Constriction
+
X
Y
10 12 14 16 18
8
9
10
11
12
Constriction
+
+
+
+
+
X
Y
10 12 14 16 18
8
9
10
11
12
Vorticity Strand
Tip
Vorticity Strand
+
Tip
X
Y
10 12 14 16 18
8
9
10
11
12
Vorticity Strand
Tip
Vorticity Strand
+
Tip
+
+
+
+
+
(a) (b)
(d) (c)
(e) (f)
Fig. 19. Instantaneous contours of x for Pr = 10 and Ri = 2.0 at (a) t, (c) t
s
6
, (e) t
2s
6
, (g) t
3s
6
, (i) t
4s
6
, (k) t
5s
6
(close-up view). Instantaneous contours of g
b
for Pr = 10 and
Ri = 2.0 at (b) t, (d) t
s
6
, (f) t
2s
6
, (h) t
3s
6
, (j) t
4s
6
, (l) t
5s
6
(close-up view).
S. Sarkar et al. / International Journal of Heat and Mass Transfer 54 (2011) 35363551 3547
This shift is probably due the effect of heat in the near wake. The
strongest deection occurs at the end of the domain for both the
vortex rows at Pr = 0.7 and Ri = 1.5. For a xed Richardson number,
lesser severe deection is observed for higher Prandtl number.
Accordingly the lowest deection corresponds to Pr = 100 for the
range of Richardson numbers presented.
4.2.2. Variation of vortex strength
Vortex strength is computed using peak vorticity (x
p
) values
(see [12]). Fig 16a and b represents the variations of vortex
strength along the spatial direction for different Prandtl and Rich-
ardson numbers. The vorticity values for the upper vortex row are
negative and become less negative when travelling downstream
(Fig. 16a). The strength becomes smaller. The situation for lower
vortex rows show decreasing strength (Fig. 16b). This trend of
variations of strength is similar to that reported by Kieft et al.
[12].
4.2.3. Production of baroclinic vorticity during vortex formation
process
The baroclinic production of vorticity g
b
as dened in Eq. (16)
has the typical property describing the vorticity production from
the cylinder shoulder to the far wake. The sign of production term
is a determining factor to investigate its effect on vorticity produc-
tion. Kieft et al. [12] described elaborately the sign of g
b
. In their
paper they showed that at the up stream side there is a positive
baroclinic vorticity production due to positive temperature
gradient @H=@x > 0, whereas at the rear stagnation region this
production term becomes negative due to the negative gradient
of temperature @H=@x < 0. They also reported that the values
of g
b
were found to be varying strongly during vortex shedding
cycle.
In the present investigation the effect of baroclinic production
of vorticity at the vicinity of the cylinder shoulder (x < 11) is ana-
lyzed by computing local and total time-averaged vorticity ux
for varying Richardson number and Prandtl number.
X
Y
10 12 14 16 18
8
9
10
11
12
Constriction
+
X
Y
10 12 14 16 18
8
9
10
11
12
+
+ +
+
+
X
Y
10 12 14 16 18
8
9
10
11
12
Constriction
+
X
Y
10 12 14 16 18
8
9
10
11
12
+
+
+
+
X
Y
10 12 14 16 18
8
9
10
11
12
+
X
Y
10 12 14 16 18
8
9
10
11
12
+
+
+
+
+
+
+
(g) (h)
(k) (j)
(m) (l)
Fig. 19 (continued)
Pr
S
t
0 20 40 60 80 100
0.15
0.16
0.17
Ri = 1.0
Ri = 1.5
Ri = 2.0
Fig. 20. Variation of Strouhal number with Prandtl number.
3548 S. Sarkar et al. / International Journal of Heat and Mass Transfer 54 (2011) 35363551
The local time-averaged vorticity ux is dened at the perpen-
dicular cross section, x = 11 as:
c
L
x
y

x11

1
s
_
s
0
uxy; tj
x11
dt 18
The physical signicance of c
L
x
culminates the effect of the baroclin-
ic vorticity production nearer to the vicinity of the cylinder surface
along with the convection of the vorticity produced at the cylinder
wall. Another signicance of c
L
x
is that the quantity of vorticities
generated upstream of the cross section x = 11 can be implicitly
captured at the shed vortex structure during vortex shedding cycle
[12]. Fig. 17ac shows the variation of c
L
x
along the vertical cross
section for different Richardson number and Prandtl number. The
c
L
x
value for positive vorticities increases with increasing Richard-
son number while the c
L
x
value for negative vortices decreases.
However c
L
x
value decreases with increasing Prandtl number and
this decrease is higher for higher Richardson numbers.
Here we dene the total time-averaged vorticity ux carried
through the cross section at x = 11 as:
c
T
x

x11

_
12D
8D
c
L
x
y

x11
dy 19
The variations of c
T
x
with Richardson number for different Prandtl
number are shown in Fig. 18. The total time-averaged vorticity ux
increases linearly with increasing Richardson number. This explains
the fact that heat addition increases the quantity of positive vortic-
ity relative to negative vorticity in the region x > 11. But increasing
Prandtl number reduces the value of c
T
x
, and minimum value corre-
sponds to maximum Prandtl number. This typical nature of Prandtl
number variation establishes the fact that increasing Prandtl num-
ber reduces the frequencies of vorticities to be passed around that
cross section. Therefore vortex motions are retarded and vortices
are tending to attach more at the cylinder surface. This nally leads
to a stabilizing effect on the ow.
We have analyzed vortex structure formation during a complete
vortex shedding cycle at the spatial region 11 < x < 18 with the
contours of x and g
b
. Fig. 19 shows the instantaneous contours
of vorticities (Fig. 19a, c, e, g, i and k) and g
b
(Fig. 19b, d, f, h, j
and l) in a complete vortex shedding period captured at time in-
stants t; t
s
6
; t
2s
6
; t
3s
6
; t
4s
6
and t
5s
6
at Pr = 10 and Ri = 2.
The contours show that within that region the characteristics are
totally different than that of x < 11.
The shedding of upper vortex starts with the formation of vor-
tex strand (at x % 12, y % 10.4, Fig. 19a) with which the tip of the
strand is located at an area of negative baroclinic vorticity produc-
tion (at x % 12, y % 10.4, Fig. 19b). So the contribution of g
b
is to
form the initiation of vortex structure. A tiny area containing posi-
tive g
b
is found at a location x % 10.8, y % 10.2. During the consec-
utive vortex shedding cycle, this area moves farther to a location
where the constriction of the vortex strand takes place (Fig. 19i).
Periodically when the area containing positive g
b
almost vanishes
(Fig. 19d), the area containing negative g
b
evolves in dimensions
and strength (Fig. 19d, f, h and j). This leads us to the conclusion
that baroclinic vorticity production term sums up with the
negative upper vortex structure just after its formation, during
the initiation of vortex structure formation cycle. Consequently,
for the remaining shedding process the area containing positive
X
Y
10 15 20 25
5
10
15
X
Y
10 15 20 25
5
10
15
X
Y
10 15 20 25
5
10
15
X
Y
10 15 20 25
5
10
15
X
Y
10 15 20 25
5
10
15
X
Y
10 15 20 25
5
10
15
(a) (b)
(d) (c)
(e) (f)
Fig. 21. Instantaneous contours of vorticity at Ri = 2.0 for (a) Pr = 0.7, (c) Pr = 10, (e) Pr = 80 and instantaneous contours of streamlines at Ri = 2.0 for (b) Pr = 0.7, (d) Pr = 10, (f)
Pr = 80.
S. Sarkar et al. / International Journal of Heat and Mass Transfer 54 (2011) 35363551 3549
g
b
, attached with the negative g
b
, grows in strength and nally
becomes nearly equal to the area containing negative g
b
. Therefore
the baroclinic production term does not affect the absolute
strength of vortex structure.
Formation of the lower vortex starts one stage later than that of
the upper vortex (Fig. 19e, j and i). A pair of positive and negative
g
b
is located in an area of newly formed vortex structure and at a
location of x % 11.5, y % 9.4 and x % 13, y % 9.5 (Fig. 19h). The value
of negative g
b
is stronger at the initiation of the vortex formationcy-
cle. This can be shown in Fig. 19h and j at the locations of x % 11.5,
y % 9.4 and x % 13.7, y % 9.4. So the high amount of negative vortic-
ity is formed in the lower vortex structure. Finally it leads to weak-
ening the strength of the lower vortex as compared to that of the
forced convective ow situation. Consequently for the remaining
shedding process, the positive g
b
increases within the newly formed
vortex structure and nally its inuence turns to be greater or sim-
ilar than that of negative g
b
, causing net vortex strength to be same.
4.3. Variation of Strouhal number with Prandtl number
The frequency of vortex shedding is characterized by Strouhal
number. Strouhal number is a nondimensional number and
dened by St = fD/U
1
, where f is the dimensional vortex shedding
frequency. The quantitative value of St is calculated by taking the
Fast Fourier Transform (FFT) of the temporal evolution of a lift sig-
nal when the ow reaches dynamic steady state. The highest peak
of the harmonics in the FFT portrait represents the corresponding
Strouhal number in the ow. For the present range of Prandtl
and Richardson numbers, the values of Strouhal number is
computed to investigate the effect of Prandtl number on vortex
shedding frequency. The graphical representation of variation of
Strouhal number with Prandtl number for different Richardson
number is presented in Fig. 20. It can be seen that for all Richard-
son numbers, Strouhal number decreases with increasing Prandtl
number. It is observed that an increasing Prandtl number value
delays the vortex formation process and reduces ow uctuations.
Therefore the higher Prandtl number has a stabilizing effect on
ow in mixed convection ow. This can also be observed from
the instantaneous contours of vorticities and streamlines.
Fig. 21af depicts instantaneous contours of vorticity and stream-
line at Pr = 0.7, 10 and 80, at a xed Ri = 2.0. From vorticity and
streamline contours, it is found that the upward drift of vortex
structures and streamlines decreases with increasing Prandtl num-
bers. The uctuations of streamline and the size of the vortex
structures reduce with increasing Prandtl numbers. Almost sym-
metrical vortex structures are observed at Pr = 80. This phenome-
non established the fact stated by Schlichting and Gersten [28]
that the heating or cooling of a body has some inuence on the
boundary layer stability. As, for an example, heating of liquid
boundary layer has a stabilizing effect since viscosity decreases
with temperature. Furthermore, a value of Ri around unity
represents a strong effect of buoyancy. Flow is altered primarily
by buoyancy, and variation in viscosity is of secondary importance,
if the working uid is gas (example, air). The fundamental differ-
ence between the effect of buoyancy and that due to viscosity
variation is the following: Buoyancy can generate ow, suppress
ow, control velocity uctuations selectively with respect to the
direction of gravity and modify the pressure eld directly. The
effect of viscosity, in this regard, is indirect since at best, it changes
the resistance to ow.
5. Conclusions
A stabilized SUPG based nite element technique is employed
to investigate the effect of Prandtl number and Richardson number
on the wake dynamics and heat transfer past a circular cylinder in
crossow. The results are presented for both forced and mixed
convective ows for 80 6 Re 6 180, 0:7 6 Pr 6 100 and 0 6 Ri 6 2.
Simulation results for forced convection exhibit the following
characteristics:
During vortex shedding cycle at Re = 120 and Pr = 30, isotherms
show less spatial spreading and are found to be crowded nearer
to the cylinder surface.
At Pr = 1, the time averaged isotherms show symmetric distri-
bution around the line of symmetry. The isotherms are found
to be steeper with increasing Prandtl number around the cylin-
der surface, and are incarcerate to a smaller region. With
increase in Reynolds number, isotherms become steeper in
the near-wake region.
The time-averaged local Nusselt number increases with increas-
ing Reynolds number. The response of local Nusselt number dis-
tribution at the rear stagnation region is found to be increasing
up to the rear stagnation point after reaching its minimum
value nearer to the point of separation. The average Nusselt
number increases with increasing Reynolds number and Prandtl
number.
A new correlation is presented for average Nusselt number
using least square technique. The average Nusselt number fol-
lows the relation Nu
avg
= 0.459Re
0.548
Pr
0.373
for 80 6 Re 6 180
and 0:7 6 Pr 6 100. The validity of this equation is shown for
both low and high Reynolds number range for different Prandtl
numbers. Extrapolated results from the equation are compared
with the analytical, numerical and experimental results avail-
able in literature.
Another correlation is proposed on Colburn heat transfer factor
which is expressed as j
c
= 0.521Re
0.452
for 80 6 Re 6 180.
The combined effect of Prandtl and Richardson number on wake
dynamics and heat transfer has been studied at a representative
Re = 100. The following are the salient features:
The upper vortex trajectory shows higher deection than the
lower vortex trajectory. For both upper and lower vortex rows,
the strongest deection occurs at the end of the domain at
Pr = 0.7 and Ri = 1.5. The lowest deection is found at Pr = 100.
The average Nusselt number (Nu
avg
) increases with increasing
Prandtl number. With increasing Richardson number from zero
to one, there is a decrease in average Nusselt number.
The formation length L
f
decreases for upper vortex and
increases for lower vortex. The formation length increases with
an increase in Prandtl number.
Along the spatial direction, the vortex strength of upper vortex
increases and the vortex strength of lower vortex decreases.
Vortex shedding cycles are described by calculating baroclinic
vorticity production term g
b
. Close to the cylinder wall
(x < 11) local c
L
x
and total c
T
x
time-averaged vortex uxes
are calculated. The combined effect of Prandtl number and
Richardson number are analyzed. It is found that increasing
Richardson number increases the c
L
x
value for the positive
vorticities and reduces for negative vorticities. The value of c
T
x
increases linearly with increase in Richardson number. Increas-
ing Prandtl number reduces the value of c
T
x
, and minimum
value corresponds to maximum Prandtl number. In the far wake
(11 < x < 18) effect of baroclinic vorticity production termg
b
has
been described in a complete vortex shedding cycle.
Under the inuence of superimposed thermal buoyancy, Strou-
hal number decreases with increasing Prandtl number. The
upward drift of vortex structures and streamlines decreases
with increasing Prandtl numbers. The effect of increasing Pra-
ndtl number stabilizes the ow.
3550 S. Sarkar et al. / International Journal of Heat and Mass Transfer 54 (2011) 35363551
Acknowledgements
The authors would like to thank the management of TATA
STEEL, India for providing computational facilities to carry out this
research work.
References
[1] S. Sanitjai, R.J. Goldstein, Heat transfer from a circular cylinder to mixtures of
water and ethylene glycol, Int. J. Heat Mass Transfer 47 (2004) 47854794.
[2] G. Juncu, A numerical study of momentum and forced convection heat transfer
around two tandem circular cylinders at low Reynolds numbers. Part II: Forced
convection heat transfer, Int. J. Heat Mass Transfer 50 (2007) 37993808.
[3] H.C. Perkins, G. Leppert, Local heat-transfer coefcients on a uniformly heated
cylinder, Int. J. Heat Mass Transfer 7 (1964) 143158.
[4] A.K. Dhiman, R.P. Chhabra, V. Eswaran, Flow and heat transfer across a
conned square cylinder in the steady ow regime: effect of Peclet number,
Int. J. Heat Mass Transfer 48 (2005) 45984614.
[5] A.K. Sahu, R.P. Chhabra, V. Eswaran, Effects of Reynolds and Prandtl numbers
on heat transfer from a square cylinder in the unsteady ow regime, Int. J. Heat
Mass Transfer 52 (2009) 839850.
[6] A.K. Dhiman, R.P. Chhabra, A. Sharma, V. Eswaran, Effects of Reynolds and
Prandtl numbers on heat transfer across a square cylinder in the steady ow
regime, Numer. Heat Transfer A 49 (2006) 717731.
[7] G. Biswas, H. Laschefski, N.K. Mitra, M. Fiebig, Numerical investigation of
mixed convection heat transfer in a horizontal channel with built-in square
cylinder, Numer. Heat Transfer A 18 (1990) 173188.
[8] S. Singh, G. Biswas, A. Mukhopadhyay, Effect of thermal buoyancy on the ow
through a vertical channel with a built-in circular cylinder, Numer. Heat
Transfer A 34 (1998) 769789.
[9] F. Dumouchel, J.C. Lecordier, P. Paranthoen, The effective Reynolds number of a
heated cylinder, Int. J. Heat Mass Transfer 41 (1998) 17871794.
[10] G. Biswas, S. Sarkar, Effect of thermal buoyancy on vortex shedding past a
circular cylinder in cross ow at low Reynolds numbers, Int. J. Heat Mass
Transfer 52 (2009) 18971912.
[11] C.F. Lange, F. Durst, M. Breuer, Momentum and heat transfer from cylinders in
laminar crossow at 10
4
6 Re 6 200, Int. J. Heat Mass Transfer 41 (1998)
34093430.
[12] R.N. Kieft, C.C.M. Rindt, A.A. van Steenhoven, G.J.F. van Heijst, On the wake
structure behind a heated horizontal cylinder in cross-ow, J. Fluid Mech. 486
(2003) 189211.
[13] R.N. Kieft, C.C.M. Rindt, A.A. van Steenhoven, Near-wake effects of heat input
on the vortex-shedding mechanism, Int. J. Heat Fluid Flow 28 (2007) 938947.
[14] J.-M. Shi, D. Gerlach, M. Breuer, G. Biswas, F. Durst, Heating effect on steady
and unsteady horizontal laminar ow of air past a circular cylinder, Phys.
Fluids 16 (12) (2004) 43314345.
[15] P.K. Maji, G. Biswas, Analysis of ow in the spiral casing using a streamline
upwinding PetrovGalerkin method, Int. J. Numer. Methods Eng. 45 (1999)
147174.
[16] F.H. Harlow, J.E. Welch, Numerical calculation of time-dependent viscous
incompressible ow of uid with free surface, Phys. Fluids 8 (1965) 2182
2188.
[17] A.N. Brooks, T.J.R. Hughes, Streamline upwind PetrovGalerkin formulations
for convection dominated ows with particular emphasis on the
incompressible NavierStokes equations, Comput. Methods Appl. Mech. Eng.
32 (1982) 199259.
[18] K. Arul Prakash, G. Biswas, B.V. Rathish Kumar, Thermal hydraulics of the
spallation target module of an accelerator driven sub-critical system: a
numerical study, Int. J. Heat Mass Transfer 49 (2006) 46334652.
[19] N. Senthil Kumar, G. Biswas, A nite element study of the onset of vortex
shedding in a ow past two-dimensional circular cylinder, Prog. Comput. Fluid
Dynam. 8 (2008) 288298.
[20] W.A. Khan, J.R. Culham, M.M. Yovanovich, Fluid ow around and heat transfer
from an innite circular cylinder, ASME J. Heat Transfer 127 (2005) 785790.
[21] A.T. Srinivas, R.P. Bharti, R.P. Chhabra, Mixed convection heat transfer from a
cylinder in power-law uids: effect of aiding buoyancy, Ind. Eng. Chem. Res. 48
(2009) 97359754.
[22] T.S. Sharma, S.P. Sukhatme, Local heat transfer from a horizontal cylinder to air
in cross ow: inuence of free convection and free stream turbulence, Int. J.
Heat Mass Transfer 20 (1977) 5156.
[23] H.A. Kramers, Heat transfer from spheres to owing media, Physica 12 (1946)
6180.
[24] R.M. Fand, Heat transfer by forced convection from a cylinder to water in
crossow, Int. J. Heat Mass Transfer 8 (1965) 9951010.
[25] S.W. Churchill, M. Bernstein, A correlating equation for forced convection from
gases and liquids to a circular cylinder in cross ow, ASME J. Heat Transfer 99
(1977) 300306.
[26] R.N. Kieft, C.C.M. Rindt, A.A. van Steenhoven, Heat induced transition of a
stable vortex street, Int. J. Heat Mass Transfer 45 (2002) 27392753.
[27] W.J.P.M. Mass, C.C.M. Rindt, A.A. van Steenhoven, The inuence of heat on the
3D-transition of the von Karman vortex street, Int. J. Heat Mass Transfer 46
(2003) 30693081.
[28] H. Schlichting, K. Gersten, Boundary Layer Theory, eighth revised and enlarged
ed., McGraw-Hill, New York, 2000.
S. Sarkar et al. / International Journal of Heat and Mass Transfer 54 (2011) 35363551 3551