International Journal of Heat and Mass Transfer 55 (2012) 5941–5958

Contents lists available at SciVerse ScienceDirect

International Journal of Heat and Mass Transfer

journal homepage: www.elsevier.com/locate/ijhmt

Non-Newtonian power-law flow and heat transfer across a pair of side-by-side

circular cylinders
N.S.K. Chaitanya, A.K. Dhiman ⇑
Department of Chemical Engineering, Indian Institute of Technology Roorkee, Roorkee 247 667, India

a r t i c l e i n f o a b s t r a c t

Article history: Flow and heat transfer of non-Newtonian power-law fluids across a pair of identical circular cylinders in
Received 19 January 2012 side-by-side arrangement are investigated numerically by solving the continuity, momentum and energy
Received in revised form 16 May 2012 equations along with the appropriate boundary conditions. The numerical calculations are performed in
Accepted 6 June 2012
an unconfined computational domain for the following range of physical parameters: Reynolds number,
Available online 3 July 2012
Re = 1–40 and power-law index, n = 0.4–1.8 (covering shear-thinning, n < 1; Newtonian, n = 1 and shear-
thickening, n > 1 behaviors) for gap ratio, T/D = 1.5–4.0 at a constant Prandtl number of 50. The global
Keywords:
characteristics such as drag coefficients and average Nusselt number, etc. are calculated and the repre-
Side-by-side circular cylinders
Non-Newtonian fluids
sentative streamline and isotherm contours are presented for the above range of conditions. It has been
Streamlines found that the individual and overall drag coefficients decrease and the average Nusselt number
Isotherms increases with Reynolds number for all T/D and n considered here. The heat transfer is found higher in
Drag and Nusselt number shear-thinning fluids than Newtonian fluids and followed by shear-thickening fluids for 1.5 6 T/D 6 4.0
and 1 6 Re 6 40.
Ó 2012 Elsevier Ltd. All rights reserved.

1. Introduction The present work attempts to understand the flow and heat trans-
fer characteristics of the non-Newtonian power-law fluids around
The study of forced convection flow and heat transfer across a two side-by-side circular cylinders at low Reynolds numbers for
pair of circular cylinders in side-by-side arrangements have many various distances between the centers of the cylinders (i.e., gap
engineering and real life applications such as offshore structures, ratio).
heat exchangers, fuel and control rods in nuclear reactors, piers Several experiments were performed in the past to investigate
and bridge pilings, oil and gas pipelines, cooling-tower arrays, sus- the Newtonian flow characteristics around two stationary circular
pension bridges and so forth. Owing to the importance of these cylinders of equal diameter arranged side-by-side [3]. The experi-
applications, the flow and heat transfer characteristics past a pair mental/numerical investigations carried out by Farrant et al. [4],
of circular cylinders in side-by-side arrangements for Newtonian Meneghini et al. [5], Chen et al. [6], Kang [7], Kondo and
fluids has been extensively researched in the literature over Matsukuma [8] provide extensive results for Newtonian flow
considerable ranges of Reynolds and Prandtl numbers. On the other across a pair of side-by-side circular cylinders at intermediate
hand, many industrial processes often include materials like Reynolds numbers and establish the dependence of the flow char-
polymer melts, multiphase mixtures, high molecular weight acteristics on the Reynolds number (Re) and the distance between
substances, emulsions and suspensions, etc. that exhibit non- the cylinders (T/D). Ding et al. [9] numerically studied the flow
Newtonian behaviors like shear-thickening, shear-thinning, visco- using a mesh-free least square based finite difference method for
elasticity and yield stress [1]. Also, the understanding of flow and Re = 100 and 200 and reported the hydrodynamic coefficients for
thermal characteristics of submerged cylindrical objects find their T/D of 1.5, 2.5 and 4.0. Kun et al. [10] numerically investigated
use in novel designs of compact heat exchange devices, mixing the flow using an unstructured spectral element method for
impellers, in rake filters used for non-Newtonian slurries, in various non-dimensional distances (i.e., T/D) for Re < 100. They
designing the support structures exposed to non-Newtonian fluids observed a total of nine different kinds of wake patterns and re-
such as muds, in the formation of weld lines in polymer processing ported lift and drag coefficients for various wake patterns.
operations [2], in the use of thin cylinders and wires as measure- Mizushima and Ino [11] investigated the stability and the transi-
ment probes and sensors in non-Newtonian flows and so forth. tion of flow past a pair of circular cylinders in a side-by-side
arrangement for Reynolds numbers below 100 by numerical simu-
lations and linear stability analyses. Based on the gap between the
⇑ Corresponding author. Tel.: +91 1332 285890, mobile: +91 9410329605.
two cylinders (T/D), they classified the instability modes of steady
E-mail addresses: [email protected], [email protected] (A.K. Dhiman).

0017-9310/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ijheatmasstransfer.2012.06.005
5942 N.S.K. Chaitanya, A.K. Dhiman / International Journal of Heat and Mass Transfer 55 (2012) 5941–5958

Notations

CD total drag coefficient, 2F D =qU 21 D Re Reynolds number, qDn U 2n

1 =m
CDf friction drag coefficient, 2F Df =qU 21 D T distance between the centers of the cylinders, m
CDp pressure drag coefficient, 2F Dp =qU 21 D t time, s
cp specific heat of the fluid, J/kg K Dt time step, t=ðD=U 1 Þ
dP pressure loss coefficient, 1qDUp2 U1 free stream velocity of the fluid at the inlet, m/s
2 1
FD drag force per unit length of the cylinder, N/m Vx components of the velocity in the x-direction, m/s
FDf viscous drag force per unit length of the cylinder, N/m Vy components of the velocity in the y-direction, m/s
FDp pressure drag force per unit length of the cylinder, N/m x streamwise coordinate, m
h average convective heat transfer coefficient, W/m2 K Xd downstream distance, m
H height of the computational domain in the lateral direc- Xu upstream distance, m
tion, m y transverse coordinate, m
I2 second invariant of the rate of strain tensor, s2
jh Nu
Colburn heat transfer factor, RePr 1=3
Greek symbols
e rate of deformation tensor, s1
k thermal conductivity of the fluid, W/m K
d smallest grid size, m
L length of the computational domain in the axial direc-
g power-law viscosity, Pa s
tion, m
h temperature, K
m power-law consistency index, Pa sn
hw temperature at the surface of the cylinder, K
n power-law index, dimensionless
h1 temperature of the fluid at the inlet, K
Nu average Nusselt number, hD=k
q density of the fluid, kg/m3
p pressure, Pa
n1
mc s extra stress tensor, Pa
Pr Prandtl number, k p UD1

symmetric flow into four groups and reported the critical Reynolds Reynolds numbers. However, in spite of having many engineering
number for each mode of instability. Liang et al. [12] used spectral applications, the non-Newtonian flow and heat transfer over a pair
difference method to develop a two-dimensional solver and pre- of side-by-side circular cylinders have not been investigated in the
dicted four wake patterns after investigating nine different spac- literature. Therefore, the aim of this study is to investigate the non-
ings between 1.1 and 4.0 at a fixed Reynolds number of 100. Ryu Newtonian flow and heat transfer across a pair of side-by-side
et al. [13] carried out investigations by applying a finite-volume cylinders for the range of conditions: Re = 1–40, T/D = 1.5–4,
method using a multi-block grid technique for Re = 100 at various n = 0.4–1.8 (thereby including shear-thinning, n < 1; Newtonian,
non-dimensional gap spacings between 0.1 and 15. They reported n = 1 and shear-thickening, n > 1 behaviors) for a Prandtl number
five vortex interaction categories and found that the changes of of 50. The values of the Prandtl number of the order of 50 are
hydrodynamic coefficients due to variations in separation are well frequently encountered in chemical, petroleum and oil related
characterized by the five different vortex flow patterns. Recently, engineering applications [1,18,19].
Vakil and Green [14] carried out simulations for the values of Rey-
nolds number between 1 and 20 with gap spacing varied between 2. Problem statement, governing equations and boundary
0.05 and 30. They observed four distinct flow behaviors and re- conditions
ported the variation of drag and lift coefficients with gap spacing
at different values of Reynolds number. The 2-D flow of non-Newtonian power-law fluids across a pair
In contrast to flow investigations, laminar forced convection of circular cylinders in side-by-side arrangement is schematically
heat transfer from side-by-side circular cylinders is not investi- displayed in Fig. 1. At the inlet, the flow is uniform and isothermal
gated in the same extent for low and intermediate Reynolds num- with velocity U1 and temperature h1. The cylinders are considered
bers. Moshkin and Sompong [15] numerically investigated the heat to be at a constant temperature of hw (>h1). The length and the
transfer from two rotating circular cylinders for Reynolds numbers height of the computational domain in axial and lateral directions
up to 40 and Pr = 0.7–50 and reported that the increase in the value are represented by L (= Xu + Xd) and H, respectively. The circular
of Prandtl number resulted in an appreciable increase in the aver- obstacles are located symmetrically about the mid-plane at an up-
age Nusselt number and the rate of heat transfer decreases with stream distance of Xu from the inlet to the center of the cylinder
the speed of cylinder’s rotation. Han et al. [16] numerically studied and at a downstream distance of Xd from the outlet to the cylinder
the heat transfer from two nearby isothermal circular cylinders of center. The distance between the centers of the cylinders is T.
equal diameter immersed in a uniform cross-flow for Re = 120 at The governing equations for the problem under consideration
Pr = 0.7 and reported that significant changes in the characteristics can be written as given by Bird et al. [20] as
of heat transfer are noticed depending on how they are positioned, Equation of continuity
resulting in quantitative changes in heat transfer coefficients of
both cylinders. Yoon et al. [17] numerically investigated the 2-D
rV ¼0 ð1Þ
laminar forced convection heat transfer past two rotating circular Equation of momentum
cylinders in a side-by-side arrangement at various rotational  
oV
speeds for four different gap spacings of 3, 1.5, 0.7 and 0.2 at a con- q þ V  rV  n  r  r ¼ 0 ð2Þ
stant Reynolds number of 100 and a Prandtl number of 0.7 (air). ot
Thus, based upon the above discussion, it can be summarized Equation of Energy
here that sufficient information is now available in the open liter-  
oh
ature on the Newtonian flow and heat transfer characteristics qcp þ V  rh  kr2 h ¼ 0 ð3Þ
around a pair of side-by-side cylinders at low and intermediate ot
 N.S.K. Chaitanya, A.K. Dhiman / International Journal of Heat and Mass Transfer 55 (2012) 5941–5958 5943

Fig. 1. Schematics of the flow past a pair of circular cylinders in side-by-side arrangement.

where q is the density of the fluid, V is the velocity vector with com- oV x oV y oh
¼ 0; ¼ 0; ¼0 ð11Þ
ponents Vx and Vy in x- and y-directions, respectively, n is the body ox ox ox
force, r is the stress tensor and h is the temperature. The numerical simulation starts with initial conditions: Vx = 0,
The stress tensor r is defined as the sum of the isotropic pres- Vy = 0, p = 0 and h = h1, the governing Eqs. (1)–(7) along with
sure p and the deviatoric stress tensor s and is given by Eq. (4) boundary (Eqs. (8)–(11)) and initial conditions are solved for the
r ¼ pI þ s ð4Þ primitive variables, i.e., velocity (Vx and Vy), pressure (p) and tem-
perature (h) fields. The drag coefficients and the average Nusselt
The rheological equation of state for incompressible fluids is given numbers are calculated here by using ANSYS Fluent code and their
as related mathematical expressions can be found elsewhere [21,22].
s ¼ 2geðVÞ ð5Þ
3. Numerical methodology
where e(V), the components of the rate of strain tensor and are de-
fined by Eq. (6)
In this work, the grid generation and subsequent simulations
1 are carried out by using computational fluid dynamics (CFD) solver
eðVÞ ¼ ½ðrVÞ þ ðrVÞT  ð6Þ
ANSYS FLUENT. The computational grid generated here has both
2
uniform and non-uniform distributions with the mesh being fine
For power-law fluids, the viscosity g is given by Eq. (7)
near the cylinders and coarse away from the cylinders. The imme-
 ðn1Þ=2 diate space around each of the cylinders has been divided into
I2
g¼m ð7Þ eight faces, where the smallest grid size of 0.01D is utilized. The
2
grid spacing away from these faces increases exponentially from
where n is the power-law index of the fluid (n < 1 implies to a 0.01D to 0.5D up to the domain size of 20D in both x- and y-direc-
shear–thinning fluid; n = 1 implies Newtonian fluid; n > 1 implies tions. In the remaining domain, a uniform grid spacing of 0.5D is
to a shear-thickening fluid), m is the power-law consistency index implemented.
and I2 is the second invariant of the rate of strain tensor e, which The 2-D, steady/unsteady, segregated laminar solver has been
is given as I2 ¼ 2ðe2xx þ e2yy þ e2xy þ e2yx Þ. The components of the rate employed to solve the incompressible flow equations on the non-
of strain tensor are related to the velocity components in Cartesian staggered grid arrangement. The second order upwind scheme is
coordinates as follows: used to discretize convective terms of momentum and energy
  equations. The SIMPLE scheme is used for solving pressure-velocity
oV x oV y 1 oV x oV y
exx ¼ ; eyy ¼ and exy ¼ eyx ¼ þ decoupling. The second order implicit time-integration scheme is
ox oy 2 oy ox used and the dimensionless time step (Dt) has been fixed at 0.01
The physical boundary conditions for the present flow system can as for any smaller values of time step the changes in the values
be written as follows (see Fig. 1): of global characteristics are found to be negligible. The constant
 At the inlet boundary: The flow is uniform and isothermal and is density and the non-Newtonian power-law viscosity models are
flowing from left to right, used for steady/unsteady simulations. The simulations are carried
V x ¼ U 1 ; V y ¼ 0 and h ¼ h1 ð8Þ out until the residuals satisfy the convergence criteria of the order
of 1015 for steady state and of 1020 for unsteady state.
 On top and bottom boundaries: Slip boundary conditions are In order to investigate the effect of grid size on the physical
used, parameters, three non-uniform grids of grid sizes of 137,200 cells,
oV x oh 205,700 cells and 301,700 cells are used with the smallest grid
¼ 0; V y ¼ 0 and ¼0 ð9Þ
oy oy sizes (d) of 0.03D, 0.01D and 0.005D clustered around the circular
cylinders respectively. The domain considered here corresponds
 On circular cylinders: No-slip boundary condition is used with
to an upstream distance, Xu = 20D, downstream distance,
cylinders held at a constant temperature of hw
Xd = 60D and computational domain height, H = 40D. The simula-
V x ¼ 0; V y ¼ 0 and h ¼ hw ð10Þ
tions are carried out at Re = 1 and 40, Pr = 50, T/D = 4.0 for n = 0.4,
 At the exit boundary: The default outflow condition available in 1 and 1.8 (Table 1). The percentage relative changes in the values
ANSYS FLUENT is used. This condition is similar to the homoge- of the total drag coefficient for the cases of n = 0.4, 1 and 1.8 at
neous Neumann condition, i.e., Re = 1 are found to be about 0.096%, 0.076% and 0.01% for the grid
5944 N.S.K. Chaitanya, A.K. Dhiman / International Journal of Heat and Mass Transfer 55 (2012) 5941–5958

Table 1
Grid dependence study for Re = 1 and 40, Pr = 50, T/D = 4.0, n = 0.4, 1 and 1.8.

Total number of cells in the Minimum grid size Total drag coefficient Average Nusselt number Total drag coefficient Average Nusselt number
domain (d) (CD) (Nu) (CD) (Nu)
Re = 1 Re = 40
n = 0.4
301700 0.005D 23.0741 3.7488 1.3295 19.4527
205700 0.01D 23.2233 3.7332 1.3302 19.5564
137200 0.03D 23.2449 3.7488 1.3374 20.2995
n=1
301700 0.005D 11.4633 2.7805 1.7094 14.4761
205700 0.01D 11.3674 2.7727 1.7095 14.5049
137200 0.03D 11.3588 2.7745 1.7103 14.8986
n = 1.8
301700 0.005D 7.2731 2.6555 2.0177 12.4905
205700 0.01D 7.2740 2.6555 2.0176 12.5015
137200 0.03D 7.2731 2.6567 2.0175 12.6940

with 137,200 cells and about 0.67%, 0.84% and 0.04% for the grid 1.50% and 1.93% for Xu = 30D as compared to Xu = 20D, respectively.
with 301,700 cells as compared to the grid with 205,700 cells, And at Re = 40, the corresponding percentage relative changes for
respectively. And at Re = 40, the corresponding percentage relative n = 0.4, 1 and 1.8 are found to be about 2.49%, 4.70% and 7.2% for
changes for n = 0.4, 1 and 1.8 are found to be about 1.24%, 0.05% Xu = 10D and 0.36%, 0.83% and 1.25% for Xu = 30D as compared to
and 0.005% for the grid with 137,200 cells and 0.5%, 0.007% and Xu = 20D, respectively. Similarly, the relative percentage changes
0.005% for the grid with 301,700 cells as compared to the grid with in the values of the average Nusselt number for the cases of
205,700 cells, respectively. Similarly, the percentage relative n = 0.4, 1 and 1.8 at Re = 1 are found to be about 1.65%, 4.10%
changes in the values of the average Nusselt number for the cases and 2.67% for Xu = 10D and about 0.26%, 0.48% and 0.45% for
of n = 0.4, 1 and 1.8 at Re = 1 are found to be about 0.42%, 0.07% and Xu = 30D as compared to Xu = 20D, respectively. And at Re = 40,
0.034% for the grid with 137,200 cells and about 0.98%, 0.28% and the corresponding percentage relative changes for n = 0.4, 1 and
0.002% for the grid with 301,700 cells as compared to the grid with 1.8 are about 1.40%, 1.51% and 1.36% for Xu = 10D and 0.21%,
205,700 cells, respectively. And at Re = 40, the corresponding per- 0.27% and 0.25% for Xu = 30D as compared to Xu = 20D, respectively.
centage relative changes for n = 0.4, 1 and 1.8 are found to be about It can be observed that the results vary considerably when the up-
3.79%, 2.70% and 1.54% for the grid with 137,200 cells and 0.50%, stream distance is changed from 10D to 20D, but the change from
0.20% and 0.10% for the grid with 301,700 cells as compared to 20D to 30D doesn’t induce much difference in the results and
the grid with 205,700 cells, respectively. Thus, the grid size of hence, an upstream distance of 20D is used here.
205,700 cells is used for T/D = 4.0 and the grid sizes of 81,416 cells The influence of the downstream distance (Xd) is examined for
and 117,800 cells are found to be adequate for T/D values of 1.5 and Xd = 40D (191,700 cells), 60D (205,700 cells) and 80D (219,700
2.5, respectively in the present work. cells) at Re = 1 and 40, T/D = 4.0, Pr = 50 for n = 0.4, 1 and 1.8 (Table
Further, the influence of the upstream distance (Xu) on the val- 3). The relative percentage changes for both total drag coefficient
ues of physical parameters is investigated for Xu = 10D (189,950 and average Nusselt number are found to be negligible at Re = 40
cells), Xu = 20D (205,700 cells) and Xu = 30D (224,950 cells) at for all n considered here. At Re = 1, for n = 0.4, 1 and 1.8, the per-
Re = 1 and 40, T/D = 4.0, Pr = 50 for n = 0.4, 1 and 1.8 (Table 2). centage relative changes in the total drag coefficient are found to
The relative percentage changes in the values of the total drag coef- be about 0.001%, 0.011% and 0.011% for Xd = 40D and 0.35%,
ficient for the cases of n = 0.4, 1 and 1.8 at Re = 1 are found to be 0.25% and 0.002% for Xd = 80D when compared to Xd = 60D, respec-
about 0.62%, 13.71% and 17.56% for Xu = 10D and about 0.87%, tively. Similarly, at Re = 1, for n = 0.4, 1 and 1.8, the percentage

Table 2
Upstream dependence study (Xu) for Re = 1 and 40, T/D = 4.0, Pr = 50 at different values of n.

Total number of cells in the Upstream distance Total drag coefficient Average Nusselt number Total drag coefficient Average Nusselt number
domain (Xu) (CD) (Nu) (CD) (Nu)
Re = 1 Re = 40
n = 0.4
189950 10D 23.7642 3.7948 1.3633 19.8301
205700 20D 23.2233 3.7332 1.3302 19.5564
224950 30D 23.2632 3.7431 1.3253 19.5144
n=1
189950 10D 12.9262 2.8862 1.7898 14.7232
205700 20D 11.3674 2.7727 1.7095 14.5049
224950 30D 11.1896 2.7592 1.6954 14.4650
n = 1.8
189950 10D 8.5514 2.7264 2.1633 12.6709
205700 20D 7.2740 2.6555 2.0176 12.5015
224950 30D 7.1334 2.6434 1.9924 12.4702
 N.S.K. Chaitanya, A.K. Dhiman / International Journal of Heat and Mass Transfer 55 (2012) 5941–5958 5945

Table 3
Downstream dependence study (Xd) for Re = 1 and 40, T/D = 4.0, Pr = 50 at different values of n.

Total number of cells in the Downstream distance Total drag coefficient Average Nusselt number Total drag coefficient Average Nusselt number
domain (Xd) (CD) (Nu) (CD) (Nu)
Re = 1 Re = 40
n = 0.4
191700 40D 23.2235 3.7324 1.3302 19.5562
205700 60D 23.2233 3.7332 1.3302 19.5564
219700 80D 23.3067 3.7487 1.3302 19.5563
n=1
191700 40D 11.3662 2.7726 1.7094 14.5048
205700 60D 11.3674 2.7727 1.7095 14.5049
219700 80D 11.3966 2.7751 1.7095 14.5049
n = 1.8
191700 40D 7.2732 2.6554 2.0173 12.5011
205700 60D 7.2740 2.6555 2.0176 12.5015
219700 80D 7.2738 2.6554 2.0176 12.5015

Table 4
Computational domain height dependence study (H) for Re = 1 and 40, T/D = 4.0 Pr = 50 at different values of n.

Total number of cells in the Computational domain Total drag coefficient Average Nusselt number Total drag coefficient Average Nusselt number
domain height (H) (CD) (Nu) (CD) (Nu)
Re = 1 Re = 40
n = 0.4
191700 30D 23.9135 3.8257 1.3458 19.6801
205700 40D 23.2233 3.7332 1.3302 19.5564
279200 80D 23.1654 3.7282 1.3221 19.4952
n=1
191700 30D 12.8747 2.8885 1.7502 14.6069
205700 40D 11.3674 2.7727 1.7095 14.5049
219700 80D 11.1698 2.7302 1.6886 14.4549
n = 1.8
191700 30D 8.4308 2.7274 2.1013 12.5825
205700 40D 7.2740 2.6555 2.0176 12.5015
219700 80D 7.1204 2.6217 1.9738 12.4617

relative changes in the average Nusselt number are found to be between the vertical boundaries for the wall effects to be considered
about 0.023%, 0.003% and 0.002% for Xd = 40D and for Xd = 80D negligible. The influence of the height of the computational domain
the changes were 0.41%, 0.086% for n = 0.4 and 1 and for n = 1.8 (H) has been investigated (Table 4) for the values of H = 30D
the change was negligible when compared to Xd = 60D, respec- (191,700 cells), 40D (205,700 cells) and 80D (279,200 cells) for the
tively. Intuitively, the downstream should have more length scale parameters mentioned above (i.e., Re = 1 and 40, T/D = 4.0, Pr = 50,
than the upstream because of the disturbance in the flow in the up- n = 0.4, 1 and 1.8). The relative percentage changes in the values of
stream is less, but at the downstream the separation of the flow the total drag for the cases of n = 0.4, 1 and 1.8 at Re = 1 are found
and the vortex shedding from the cylinders occur. Therefore, even to be about 2.97%, 13.25% and 15.90% for H = 30D and about 0.24%,
though the differences in the results for various downstream dis- 1.73% and 2.11% for H = 80D as compared to H = 40D, respectively.
tances are relatively negligible, the downstream distance of 60D And at Re = 40, the corresponding percentage relative changes for
is used in this study for better visualization of the streamline and n = 0.4, 1 and 1.8 are found to be about 1.17%, 2.38% and 4.14% for
isotherm contours. H = 30D and 0.61%, 0.35% and 0.32% for H = 80D as compared to
The present case being that of unconfined cylinders, the vertical H = 40D, respectively. Similarly, the relative percentage changes in
boundaries determine the computational area. It has been a well the values of the average Nusselt number for the cases of n = 0.4, 1
established fact that a minimum distance of about 30D is required and 1.8 at Re = 1 are found to be about 2.48%, 4.17% and 2.71% for
H = 30D and about 0.14%, 1.53% and 1.27% for H = 80D as compared
to H = 40D, respectively. And at Re = 40, the corresponding percent-

Table 5
Validation of results for drag coefficient of side-by-side cylinders at Re = 100 for T/
D = 1.5 and 4.0.
Table 6
Value of T/D Upper or Lower cylinder Present work Ding et al. [9] Validation of average Nusselt number for side-by-side cylinders at Re = 100, Pr = 0.7
for T/D = 2.5 and 4.0.
T = 1.5D Upper cylinder 1.4901 1.53
Lower cylinder 1.4656 1.51 Value of T/D Present work Yoon et al. [17]
T = 4.0D Upper cylinder 1.4727 1.514 ± 0.013 T = 2.5D 5.2456 5.17
Lower cylinder 1.4726 1.514 ± 0.013 T = 4.0D 5.2404 5.19
5946 N.S.K. Chaitanya, A.K. Dhiman / International Journal of Heat and Mass Transfer 55 (2012) 5941–5958

Table 7
Validation of drag coefficient and average Nusselt number for an unconfined single cylinder for Re = 1 and 40 for n = 0.6, 1 and 1.8.

Re n Present work Sivakumar et al. [21] Present work Bharti et al. [22]
Drag coefficient Average Nusselt number
1 0.6 17.3273 17.2388 3.1392 3.1304
1.0 10.4060 10.3555 2.7021 2.6690
1.8 5.8715 5.8537 2.5352 2.5350
40 0.6 1.3380 1.3368 16.4478 16.3920
1.0 1.5012 1.5013 14.0579 13.9163
1.8 1.6686 1.6674 12.3062 12.1804

age relative changes for n = 0.4, 1 and 1.8 are about 0.63%, 0.70% and H = 40D vary greatly. On the other hand, when H = 40D is compared
0.65% for H = 30D and 0.31%, 1.23% and 2.01% for H = 80D as with an even large domain height, H = 80D only slight variation in
compared to H = 40D, respectively. It can be observed here that the results has been observed. Hence, a height of 40D has been em-
H = 30D is not adequate since the results when compared with ployed here for the computational domain.

(a)T/D=1.5, n=0.4 24 (d)T/D=2.5, n=0.4 24 (g)T/D=4.0, n=0.4 24

30 30 30
22 22 22

25 25 25
20 20 20
20 22 20 22 20 22

20 20 20

15 15 15

10 10 10
20 25 30 35 20 25 30 35 20 25 30 35

(b)T/D=1.5, n=1 24 (e)T/D=2.5, n=1 24 (h)T/D=4.0, n=1 24

30 30 30
22 22 22

25 25 25
20 20 20
20 22 20 22 20 22

20 20 20

15 15 15

10 10 10
20 25 30 35 20 25 30 35 20 25 30 35

(c)T/D=1.5, n=1.8 24 (f)T/D=2.5, n=1.8 24 (i)T/D=4.0, n=1.8 24

30 30 30
22 22 22

25 25 25
20 20 20
20 22 20 22 20 22

20 20 20

15 15 15

10 10 10
20 25 30 35 20 25 30 35 20 25 30 35

Fig. 2. Streamline contours around a pair of cylinders in side-by-side arrangement along with magnified views of the top cylinder at Re = 1 for varying values of T/D and n.
 N.S.K. Chaitanya, A.K. Dhiman / International Journal of Heat and Mass Transfer 55 (2012) 5941–5958 5947

4. Results and discussion T = 1.5D and about 2.6% for T = 4.0D (Table 5). Similarly, the drag
values for the upper cylinder differ by about 2.7% for T/D = 1.5
The 2-D unconfined flow of non-Newtonian power-law fluids and about 2.6% for T/D = 4.0 (Table 5). Whence, the heat transfer re-
(0:4 6 n 6 1:8), covering shear-thinning (n < 1), Newtonian (n = 1) sults are compared for T/D = 2.5 and T/D = 4.0 at Re = 100 and
and shear-thickening (n > 1) behaviors, is simulated here in the full Pr = 0.7 for n = 1, the values of the average Nusselt number are
computational domain for Reynolds number, Re = 1–40 and gap ra- found to differ about 1.5% and 1.0%, respectively (Table 6) with
tio, T/D = 1.5, 2.5 and 4.0 at Prandtl number, Pr = 50. The validation the values reported by Yoon et al. [17].
of present numerical results with literature values is reported in Further, simulations have been carried out for steady non-New-
the following subsection. tonian flow across an unconfined single cylinder for Re = 1 and 40,
Pr = 50 and n = 0.6, 1 and 1.8, and the results are compared to those
4.1. Validation of results reported by Sivakumar et al. [21] for drag coefficients and of Bharti
et al. [22] for average Nusselt numbers (Table 7). The values of the
The validation of the current framework is done by comparing drag coefficients for n = 0.6, 1 and 1.8 are found to differ only by
with the results of Ding et al. [9] and Yoon et al. [17] as presented 0.51%, 0.48% and 0.30%, respectively at Re = 1 and by 0.09%, 0.01%
in Tables 5 and 6. The unsteady fluid flow studies conducted at and 0.07%, respectively at Re = 40. Similarly, the average Nusselt
Re = 100 for T/D values of 1.5 and 4.0 reveal that the present drag numbers for n = 0.6, 1 and 1.8 are found to differ only by 0.28%,
values compared for the lower cylinder differ by about 3.0% for 1.22% and 0.01%, respectively at Re = 1 and by 0.34%, 1.01% and

(a)T/D=1.5, n=0.4 24 (d)T/D=2.5, n=0.4 24 (g)T/D=4.0, n=0.4 24

30 30 30
22 22 22

25 25 25
20 20 20
20 22 20 22 20 22

20 20 20

15 15 15

10 10 10
20 25 30 35 20 25 30 35 20 25 30 35

(b)T/D=1.5, n=1 24 (e)T/D=2.5, n=1 24 (h)T/D=4.0, n=1 24

30 30 30
22 22 22

25 25 25
20 20 20
20 22 20 22 20 22

20 20 20

15 15 15

10 10 10
20 25 30 35 20 25 30 35 20 25 30 35

(c)T/D=1.5, n=1.8 24 (f)T/D=1.5, n=1.8 24 (i)T/D=4.0, n=1.8 24

30 30 30
22 22 22

25 25 25
20 20 20
20 22 20 22 20 22

20 20 20

15 15 15

10 10 10
20 25 30 35 20 25 30 35 20 25 30 35

Fig. 3. Streamline contours of flows at Re = 20 for varying values of T/D and n.

5948 N.S.K. Chaitanya, A.K. Dhiman / International Journal of Heat and Mass Transfer 55 (2012) 5941–5958

1.02%, respectively at Re = 40. In addition, the validation of present ing the streamline contours (as shown in Figs. 2–4 for Re = 1–40,
solution methodology for steady unconfined Newtonian flow and n = 0.4–1.8 and T/D = 1.5–4.0) along with the magnified views for
heat transfer [23], unsteady unconfined non-Newtonian flow [24] the top cylinder. Overall, the flow patterns can be classified into
and steady confined non-Newtonian flow [24] over a single circular three steady patterns: (i) Single bluff-body steady pattern, (ii)
cylinder are available in our recent studies [23,24]. Transitional steady pattern and (iii) Separated double body steady
Thus, the results obtained in the present study are in excellent pattern, and one unsteady pattern: (i) Single bluff-body periodic
agreement with the results available in the literature and hence pattern for the range of conditions covered here. The categoriza-
the results presented henceforth can be deemed trustworthy. This tion of flow patterns here follows the nomenclature used by Kun
validates the present numerical solution procedure. Unfortunately, et al. [10]. For a better understanding, the flow patterns are dis-
no experimental/numerical results are available in the open litera- cussed individually for the values of T/D at different Re & n and
ture on the non-Newtonian flow and heat transfer around a pair of then finally the combined effects at different flow parameters are
circular cylinders in side-by-side arrangement. summarized.

4.2. Flow patterns 4.2.1. Influence of Re and n on flow patterns at T/D = 1.5
At Re = 1, for shear-thinning fluids (n < 1), the flow is uniform
The flow patterns of non-Newtonian power-law fluids across a and no flow separation is observed, as can be seen in Fig. 2(a).
pair of side-by-side circular cylinders are characterized by analyz- For Re > 1, the flow patterns just downstream of the cylinders

(a)T/D=1.5, n=0.4 24 (d)T/D=2.5, n=0.4 24 (g)T/D=4.0, n=0.4 24

30 30 30
22 22 22

25 25 25
20 20 20
20 22 20 22 20 22

20 20 20

15 15 15

10 10 10
20 25 30 35 20 25 30 35 20 25 30 35

(b)T/D=1.5, n=1 24 (e)T/D=2.5, n=1 24 (h)T/D=4.0, n=1 24

30 30 30
22 22 22

25 25 25
20 20 20
20 22 20 22 20 22

20 20 20

15 15 15

10 10 10
20 25 30 35 20 25 30 35 20 25 30 35

(c)T/D=1.5, n=1.8 24 (f)T/D=2.5, n=1.8 24 (i)T/D=4.0, n=1.8 24

30 30 30
22 22 22
Y

25 25 25
20 20 20
20 22 20 22 20 22

20 20 20

15 15 15

10 10 10
20 25 30 35 20 25 30 35 20 25 30 35

Fig. 4. Streamline contours of flows at Re = 40 for varying values of T/D and n.

 N.S.K. Chaitanya, A.K. Dhiman / International Journal of Heat and Mass Transfer 55 (2012) 5941–5958 5949

i.(a) T/D=1.5, Re=10, n=1.8 i.(b) T/D=1.5, Re=10, n=1.8 i.(c) T/D=1.5, Re=10, n=1.8
2.4
30 30

25 25

CD
2.35
20 20

15 15

2.3
10 10 300 350 400 450
20 25 30 35 20 25 30 35
Time, t
ii.(a)T/D=1.5, Re=11, n=1.8 ii.(b) T/D=1.5, Re=11, n=1.8 ii.(c) T/D=1.5, Re=11, n=1.8
30 30 2.6

25 25

CD
20 20 2.5

15 15

10 10 2.4
20 25 30 35 20 25 30 35 300 350 400 450
Time, t

Fig. 5. Streamlines, isotherms and drag coefficient depicting the onset of unsteady behavior for T/D = 1.5 and n = 1.8 at Re = 10 and 11.

begin resembling the flow in a channel with sudden expansion. observed for n P 1.4 and for Re = 10 the vortices are observed for
This can be observed to be the tendency of the fluid to form sepa- n > 1. And, the size of the steady wake increases with Re and/or
rate wakes for each cylinder and henceforth, will be termed as n. For Re > 10, the size of the steady wake keeps increasing with
transitional steady patterns. The transitional steady patterns are Reynolds number, until a critical Reynolds number is reached,
observed for all the values of power-law index, i.e., n < 1 at above which unsteady behavior in shear-thickening fluids (n > 1)
Re = 5. With increasing the value of the Reynolds number is observed. The transition from steady to unsteady flow patterns
(Re > 5), the tendency towards the formation of separate wakes in- can be seen in Fig. 5 for n = 1.8 along with the variation of CD with
creases as observed from the transitional steady patterns of Re = 10 time and Table 8 presents the transitional Re range for various val-
and 20 (Fig. 3a). And, this tendency increases with decreasing ues of n. The onset of unsteady behavior occurs between Re = 10–
shear-thinning behavior. As the Reynolds number is further in- 11 and between Re = 19–20 for n = 1.8 and 1.1, respectively. For
creased (Re > 20), vortices start appearing behind each of the cylin- all other shear-thickening fluids with values of n between 1.8
ders. And, the early formation of vortices is observed for higher and 1.1, the limits of transition from steady to unsteady flow are
shear-thinning behavior. The appearance of twin vortices indicates between Re = 11 and 20. Also, higher the value of power-law index,
the formation of separate wakes for each of the cylinders and n earlier is the transition. The unsteady flow pattern observed is
hence the flow pattern is termed as separated double-body steady periodic and resembles the unsteady periodic pattern of a single
pattern. For 30 6 Re 6 40, the flow patterns are separated double- cylinder. And, hence the flow pattern is termed as single bluff-body
body steady patterns for all n < 1 (Fig. 4a) and the sizes of vortices periodic pattern. For 20 6 Re 6 40 and n > 1, the flow patterns ob-
increase with Reynolds number and/or shear-thinning behavior. served are single bluff-body periodic patterns.
On the other hand, the flow behavior of shear-thickening fluids The flow patterns of Newtonian fluids (n = 1) for T/D = 1.5 are
(n > 1) at T/D = 1.5 significantly differs from that of shear-thinning found to be similar to those of shear-thickening fluids (Figs. 2–
fluids (n < 1). For shear-thickening fluids (n > 1) at Re = 1, the flow 4b). No flow separation is observed for Re 6 5 (Fig. 2b). The ten-
is uniform for low values of n. However, as the shear-thickening dency towards the wake formation is observed to increase with
behavior increases, the flow gets separated and leads to the forma- Re and the single bluff-body steady patterns are observed for
tion of wakes. The wakes are only observed at n = 1.8 for Re = 1
(Fig. 2c). This wake formation is characterized by the formation
Table 8
of two elongated vortices which are symmetrical about the mid
The onset of unsteady behavior for n P 1 at T/D = 1.5.
plane and a little apart from both the cylinders. The flow pattern
resembles the steady wake formation similar to the case of a single Power-law index (n) Transitional Re range
cylinder and is termed as the single bluff-body steady pattern. 1.8 10–11
Increasing the value of the Reynolds number (Re > 1) brings forth 1.4 13–14
1.1 19–20
the formation of wakes even for fluids with lesser shear-thickening
1 26–27
behavior. For instance, at Re = 5, the appearance of vortices are
5950 N.S.K. Chaitanya, A.K. Dhiman / International Journal of Heat and Mass Transfer 55 (2012) 5941–5958

i.(a) T/D=1.5, Re=26, n=1 i.(b) T/D=1.5, Re=26, n=1 i.(c) T/D=1.5, Re=26, n=1
30 30 1.92

25 25

CD
20 20 1.91

15 15

10 10 1.9
600 650 700
20 25 30 35 20 25 30 35
Time, t
ii.(a) T/D=1.5, Re=27, n=1 ii.(b) T/D=1.5, Re=27, n=1 ii.(c) T/D=1.5, Re=27, n=1
1.91
30 30

25 25

CD
1.9
20 20

15 15

1.89
10 10 600 650 700
20 25 30 35 20 25 30 35
Time, t

Fig. 6. Streamlines, isotherms and drag coefficient depicting the onset of unsteady behavior for T/D = 1.5 and n = 1 at Re = 26 and 27.

Re = 10 and 20 (Fig. 3b). The wake size is found to increase with Re 4.2.3. Influence of Re and n on flow patterns at T/D = 4.0
for 10 6 Re 6 26. The onset of unsteady behavior is found here at For T/D = 4.0, the flow characteristics are observed to be similar
Re = 27 (Fig. 6 and Table 8) and hence, the single bluff-body peri- to those of T/D = 2.5 (Fig. 2–4g–i) for the range of physical param-
odic pattern is observed for 27 6 Re 6 40. The variation of CD with eters studied. Here, the transitional steady flow patterns are ob-
time can also be seen in Fig. 6. The present Newtonian results are served for 1 6 Re 6 10, which show higher tendency of transition
also consistent with the results reported elsewhere [14]. For in- than those of T/D = 2.5 for 0.4 6 n 6 1.8. If compared with the
stance, the uniform flow pattern is observed at Re = 5 and for gap ratio, T/D of 2.5, the appearance of twin vortices behind each
Re = 20, the single bluff-body steady pattern is observed in the of the cylinders is observed earlier at T/D = 4.0 (for Re > 10). And,
two studies. the separated double-body wakes observed for 20 6 Re 6 40 are
also found to be larger in case of T/D = 4.0 than those of T/D = 2.5
for any particular values of n (0.4 6 n 6 1.8). Further, similar to
4.2.2. Influence of Re and n on flow patterns at T/D = 2.5 T/D = 2.5, the onset of unsteady behavior is not observed for
For T/D = 2.5, the flow characteristics are observed to be similar T/D = 4.0 at any value of Re and n covered here.
for shear-thinning (n < 1), shear-thickening (n > 1) and Newtonian Overall, the flow patterns for shear-thinning fluids (n < 1) are
(n = 1) fluids for the range of parameters investigated here. The found steady (for the range of conditions as 1 6 Re 6 40 and
transitional steady flow patterns are observed at Re = 1, as can be 1:5 6 T=D 6 4:0) and if the vortex formation occurs it results in
seen in Fig. 2d–f. The transition tendency of the flow increases with separated double-body steady patterns. The early formation of vor-
power-law index, n (i.e., from shear-thinning fluids to Newtonian tices is observed for higher shear-thinning fluids for T/D = 1.5;
fluids and subsequently to shear-thickening fluids). Increasing whereas, for T/D = 2.5 and 4.0 the early formation of vortices is ob-
the value of the Reynolds number (Re > 1) accelerates the transi- served for lesser shear-thinning fluids. As the different trends of
tion tendency as the streamlines observed for Re = 10 show higher flow patterns are observed for T/D = 1.5 and T/D > 1.5 (i.e.,
transition than those for Re = 5. This transition tendency finally 2:5 6 T=D 6 4:0) for 0.4 6 n 6 1.8. For instance, for T/D = 1.5, with
leads to the formation of vortices behind each of the cylinders decreasing n from 1.8–0.4, the single bluff-body behavior depletes
for Re > 10 and the early appearance of vortices is observed in flu- and gets converted into separated double-body behavior (as ob-
ids with higher values of power-law index, n. And, the flow pat- served for shear-thinning fluids). On the other hand, for T/D = 2.5
terns corresponding to the formation of twin vortices are the and 4.0, the separated double-body behavior is observed for all n
separated double body steady patterns, as can be seen in Fig. 3d– and depletes as the value of n decreases from n = 1.8–0.4.
f (for Re = 20) and Fig. 4d–f (for Re = 40). The sizes of the vortices Further, for shear-thickening (n > 1) and Newtonian (n = 1) flu-
behind the cylinders are found to increase with Reynolds number ids, the flow patterns are steady only for T/D = 2.5 and 4.0 for
and/or power-law index for 10 < Re 6 40 and n = 0.4–1.8. The onset 1 6 Re 6 40. Whereas, for T/D = 1.5, the flow is steady until a criti-
of unsteady behavior is not observed at T/D = 2.5 for the range of cal Reynolds number is reached, above which the flow patterns are
conditions embraced in this study. found unsteady. Also, for T/D = 1.5, the steady wakes behind the
 N.S.K. Chaitanya, A.K. Dhiman / International Journal of Heat and Mass Transfer 55 (2012) 5941–5958 5951

cylinders in shear-thickening and Newtonian fluids correspond to the centerline (Figs. 7–9c, f, i). It has also been found that lower
single bluff-body steady patterns, in contrast to the separated dou- the value of power-law index (n) for shear-thinning fluids, the
ble-body steady patterns observed for T/D ratios of 2.5 and 4.0. more inclined are the isotherms towards the centerline. On
the other hand, in the case of shear-thickening fluids (n > 1), the
4.3. Thermal patterns inclination of isotherms away from the centerline increases with
power-law index, n.
The heat transfer characteristics are studied here by using iso- Obviously, the inclination of isotherms in case of Newtonian
therm contours, as shown in Figs. 7–9 along with the magnified fluids (n = 1) is in between shear-thinning (n < 1) and shear-thick-
views for a top cylinder. The heat transfer from the cylinders is ening (n > 1) behaviors for all T/D and Re (Figs. 7–9b, e, h). Due to
guided by the flow and hence, the isotherm patterns can be clas- inclination towards the centerline, the isotherms of shear-thinning
sified in the same way as the streamline patterns mentioned fluids (n < 1) are more interactive and compact than shear-thicken-
above by single bluff-body, transitional and separated double ing fluids (n > 1). Hence, the separation of isotherms readily occurs
body steady patterns and single bluff-body periodic pattern (Sec- in shear-thickening fluids, and the tendency to form separated
tion 4.2). It can be seen from these figures that for all T/D and Re double-body wakes is higher than shear-thinning fluids. Increasing
studied here, the isotherms of shear-thinning fluids (n < 1) are in- the value of gap ratio (T/D) from 1.5 to 4.0 also favors the separa-
clined towards the centerline (Figs. 7–9a, d, g), whereas the iso- tion of isotherms, and hence the formation of separated double-
therms of shear-thickening fluids (n > 1) are inclined away from body wakes.

(a)T/D=1.5, n=0.4 24 (d)T/D=2.5, n=0.4 24 (g)T/D=4.0, n=0.4 24

30 30 30
22 22 22

25 25 25
20 20 20
20 22 20 22 20 22

20 20 20

15 15 15

10 10 10
20 25 30 35 20 25 30 35 20 25 30 35

(b)T/D=1.5, n=1 24 (e)T/D=2.5, n=1 24 (h)T/D=4.0, n=1 24

30 30 30
22 22 22
Y

25 25 25
20 20 20
20 22 20 22 20 22

20 20 20

15 15 15

10 10 10
20 25 30 35 20 25 30 35 20 25 30 35

(c)T/D=1.5, n=1.8 24 (f)T/D=2.5, n=1.8 24 (i)T/D=4.0, n=1.8 24

30 30 30
22 22 22

25 25 25
20 20 20
20 22 20 22 20 22

20 20 20

15 15 15

10 10 10
20 25 30 35 20 25 30 35 20 25 30 35

Fig. 7. Isotherms around a pair of circular cylinders in side-by-side configuration at Re = 1 for varying values of T/D and n.
5952 N.S.K. Chaitanya, A.K. Dhiman / International Journal of Heat and Mass Transfer 55 (2012) 5941–5958

(a)T/D=1.5, n=0.4 24 (d)T/D=2.5, n=0.4 24 (g)T/D=4.0, n=0.4 24

30 30 30
22 22 22

25 25 25
20 20 20
20 22 20 22 20 22

20 20 20

15 15 15

10 10 10
20 25 30 35 20 25 30 35 20 25 30 35

(b)T/D=1.5, n=1 24 (e)T/D=2.5, n=1 24 (h)T/D=4.0, n=1 24

30 30 30
22 22 22

25 25 25
20 20 20
20 22 20 22 20 22

20 20 20

15 15 15

10 10 10
20 25 30 35 20 25 30 35 20 25 30 35

(c)T/D=1.5, n=1.8 24 (f)T/D=2.5, n=1.8 24 (i)T/D=4.0, n=1.8 24

30 30 30
22 22 22

25 25 25
20 20 20
20 22 20 22 20 22

20 20 20

15 15 15

10 10 10
20 25 30 35 20 25 30 35 20 25 30 35

Fig. 8. Isotherms around a pair of circular cylinders in side-by-side configuration at Re = 20 for varying values of T/D and n.

4.4. Dependence of the drag on Re, n and T/D value of T/D affects the flow patterns and hence the drag
coefficients.
The frictional drag (CDf) and the pressure drag (CDp) both con- Fig. 10a–c present the variation of pressure drag coefficient (CDp)
tribute to the total hydrodynamic drag (CD) exerted by the fluid with Re and n for T/D ratios of 1.5, 2.5 and 4.0. The value of CDp
on the cylinders (i.e., CD = CDp + CDf). The variation of individual decreases with increase in Re for all n and T/D. The variation in
and total drag coefficients depends here on Re, n and T/D and pre- the values of CDp with n for T/D = 1.5 are found different from that
sented in Figs. 10 and 11 for a lower cylinder. The drag coefficients of T/D = 2.5 and 4.0. At T/D = 1.5, the values of CDp decrease with
for the upper cylinder are found identical in magnitude to those of increasing n for 1 6 Re < 20 and thereafter (for 20 6 Re 6 40) CDp
the lower cylinder and the overall drag coefficient of the system decreases for 0.4 6 n 6 1, but increases for 1 6 n 6 1.8. On the other
would be the double of the total drag coefficient of top or bottom hand, for T/D = 2.5 and 4.0, the values of CDp decrease with n for
cylinder. The drag coefficients are known to be functions of Rey- 1 6 Re 6 20; however, for 20 < Re 6 40 CDp increases for
nolds numbers. For the flow of constant density and constant 0.4 6 n 6 1 and decreases for 1 6 n 6 1.8. As the value of T/D
velocity, the inertial term of Re remains constant, while the viscous increases from 1.5 to 4.0, the value of CDp increases with T/D for
term (proportional to U n1 ) increases or decreases depending upon 1 6 Re 6 10; however, it decreases with T/D for 10<Re 6 40 for all n.
the value of power-law index, n. Therefore, various trends of indi- The viscous drag coefficient (CDf) is found higher for shear-
vidual and overall drag coefficients are observed (as discussed be- thinning fluids than Newtonian and shear-thickening fluids at
low) due to their complex dependence on Re and n. In addition, the low Reynolds numbers (Re  5); whereas, an opposite trend is
 N.S.K. Chaitanya, A.K. Dhiman / International Journal of Heat and Mass Transfer 55 (2012) 5941–5958 5953

(a)T/D=1.5, n=0.4 24 (d)T/D=2.5,n=0.4 24 (g)T/D=4.0, n=0.4 24

30 30 30
22 22 22

25 25 25
20 20 20
20 22 20 22 20 22

20 20 20

15 15 15

10 10 10
20 25 30 35 20 25 30 35 20 25 30 35

(b)T/D=1.5, n=1 24 (e)T/D=2.5, n=1 24 (h)T/D=4.0,n=1 24

30 30 30
22 22 22

25 25 25
20 20 20
20 22 20 22 20 22

20 20 20

15 15 15

10 10 10
20 25 30 35 20 25 30 35 20 25 30 35

(c)T/D=1.5, n=1.8 24 (f)T/D=2.5, n=1.8 24 (i)T/D=4.0, n=1.8 24

30 30 30
22 22 22

25 25 25
20 20 20
20 22 20 22 20 22

20 20 20

15 15 15

10 10 10
20 25 30 35 20 25 30 35 20 25 30 35

Fig. 9. Isotherms around a pair of circular cylinders in side-by-side configuration at Re = 40 for varying values of T/D and n.

observed for Re > 5 for all values of T/D (Fig. 10d–f). The value of CDf Fig. 10g–i depict the variation of CDp/CDf at different values of Re,
decreases with increasing Re for all T/D and n. The rate of declina- T/D, n and show that the contribution of pressure drag coefficient,
tion in the values of CDf with Re decreases with increasing n as the CDp to total drag coefficient, CD is considerably higher in shear-
shear-thinning fluids exhibit stronger dependence on Re than New- thinning fluids than Newtonian and shear-thickening fluids and
tonian and shear-thickening fluids. Therefore, for Re > 5, the values it increases with increasing Re. Fig. 11a–c depict the variation of
of CDf for shear-thickening fluids are found to be higher than New- the total drag coefficient, CD with Re, n and T/D along with the mag-
tonian and shear-thinning fluids. This results in a crossover of CDf nified views (as shown in Fig. 11a1–c1). The variations of the total
values at Re  5. Furthermore, the point of crossover in the values drag coefficient (CD) are found similar to that of viscous drag coef-
of CDf varies slightly with T/D ratio. For instance, the crossover in ficient (CDf), but the contribution of pressure drag coefficient (CDp)
the values of CDf is just above Re = 5 for T/D = 1.5; whereas, for to total drag coefficient shifts the point of crossover to a high value
T/D = 2.5 and 4.0, the crossover is observed just below Re = 5. Also, of Reynolds number. For instance, the crossover in the values of CD
the values of CDf increase from T/D = 1.5 to 2.5 for all Re and n. But, for shear-thinning (n < 1) and shear-thickening (n > 1) fluids occurs
the CDf values of T/D = 4.0 are found greater than those of T/D = 2.5 for Re > 10, whereas the crossover in the values of CDf occurs at
for 1 6 Re 6 10; however, the opposite trend is observed for Re  5 for all T/D ratios considered here.
10<Re 6 40. This decline in the values of CDf indicates that the Next, simple expressions have been developed to represent the
dependency on Re is stronger at T/D = 4.0 than T/D = 2.5, since total drag coefficient as a function of Reynolds number, Re and
the rate of decrease in the values of CDf is higher for T/D = 4.0. power-law index, n for the range of settings studied here. After a
5954 N.S.K. Chaitanya, A.K. Dhiman / International Journal of Heat and Mass Transfer 55 (2012) 5941–5958

Fig. 10. The variation of pressure drag coefficient (CDp), viscous drag coefficient (CDf) and CDp/CDf with Re for n = 0.4–1.8 and T/D = 1.5–4.0.

thorough experimentation, it is observed that two different expres- 4.5. Dependence of the average Nusselt number on Re, n and T/D
sions: (i) for 1 6 Re<20 and (ii) for 20 6 Re 6 40 can be used so as to
minimize the maximum and average deviations and in order to The variation of the surface average Nusselt number (Nu) with
determine the drag for the intermediate values of physical param- Re, n and T/D for a constant Prandtl number of 50 is depicted in
eters. The expressions are as follows: Fig. 12a–c. Similar to a single circular cylinder [25], the average
Nusselt number and hence, the heat transfer increases with increas-
(i) For 1 6 Re < 20, ing Re for fixed T/D and n due to increasing contribution of convec-
tion. As expected, the heat transfer in shear-thinning fluids (n < 1) is
  found to be considerably higher than the Newtonian fluids (n = 1)
b Re and followed by shear-thickening fluids (n > 1) for the range of con-
CD ¼ a þ d ð12Þ
nc ditions as 1 6 Re 6 40, 0:4 6 n 6 1:8 and 1:5 6 T=D 6 4:0. For
shear-thinning fluids, as the value of power-law index, n decreases
(ii) For 20 6 Re 6 40,
shearing between the cylinders increases, this lowers the effective
  viscosity and increases the rate of heat transfer.
b
CD ¼ a þ Red ð13Þ The average Nusselt numbers are found to be higher for
nc
T/D = 2.5 than T/D = 1.5 for any Re and n studied here. But, the
difference in the values of Nu for T/D = 1.5 and 2.5 decrease with
The corresponding values of the fitted constants (a, b, c and d) along increasing Re and more rapidly in shear-thinning fluids than New-
with their maximum and average deviations are given in Table 9. tonian and shear-thickening fluids, indicating stronger dependence
 N.S.K. Chaitanya, A.K. Dhiman / International Journal of Heat and Mass Transfer 55 (2012) 5941–5958 5955

Fig. 11. The variation of the total drag coefficient, CD with Re for n = 0.4–1.8 and T/D = 1.5–4.0 (a–c) and the magnified views of the point of crossover (a1–c1).

Table 9
Correlation fitted constants (a, b, c and d) for determining CD at various T/D ratios.

T/D Re a b c d Average deviation (%) Maximum deviation (%)

1.5 1 6 Re<20 2.474 9.699 1.187 0.618 4.59 11.6
20 6 Re 6 40 3.102 0.195 0.341 0.507 3.56 7.84
2.5 1 6 Re<20 3.085 11.341 1.162 0.599 3.72 10.4
20 6 Re 6 40 6.711 2.575 0.094 0.175 1.32 3.00
4.0 1 6 Re<20 3.168 13.832 0.995 0.596 4.61 7.76
20 6 Re 6 40 7.426 3.244 0.077 0.153 1.27 3.37

of shear-thinning fluids on Re compared to T/D ratio. On the other law index covered. The difference in the values of Nu for T/D = 2.5
hand, the values of Nu at T/D = 4.0 are found to be higher than and 4.0 is minimal in comparison to the difference in values of T/D
those of T/D = 2.5 only for Re < 10 and an opposite trend in the val- ratios of 1.5 and 2.5 and thus indicating the declining effect of T/D
ues of Nu is observed for 10 6 Re 6 40 for all the values of power- ratio on average Nusselt number.
5956 N.S.K. Chaitanya, A.K. Dhiman / International Journal of Heat and Mass Transfer 55 (2012) 5941–5958

Table 10
Correlation fitted constants (a, b and c) for determining jh at various T/D ratios.

T/D a b c Average Maximum

deviation (%) deviation (%)
1.5 0.522 0.440 0.240 4.13 9.77
2.5 0.739 0.550 0.244 2.81 11.00
4.0 0.786 0.580 0.245 3.31 12.77

The present numerical results for the jh factor can be repre-

sented by the following correlation for all the range of parameters
covered in this study

a
jh ¼ ð15Þ
Reb nc
The best values of the fitted constants (a, b and c) for various T/D
ratios are given in Table 10 along with their maximum and average
deviations.
In general, the average Nusselt number increases with Prandtl
number for any Re and n for the flow across a single circular cylin-
der [25]. The heat transfer results obtained for the present system
are similar to those reported for a single cylinder and hence the
same could be attributed to the variation of the heat transfer char-
acteristics with Prandtl number. The specific flow and temperature
characteristics at varying values of Pr could be considered as a fu-
ture scope of research.
Increase in heat transfer is often accompanied by increase in
pressure loss across the domain. And hence, the dimensionless
pressure loss coefficient, dP has been calculated and depicted in
Figs. 13a–c, which is defined as

Dp
dP ¼ 1 ð16Þ
2
qU 21
It has been observed that the variation trends of dP with Re, n
and T/D are similar to those of CD. For any Re and T/D, the Nu of
shear-thinning fluids is higher to those of Newtonian and shear-
thickening fluids. On the other hand, the values of dP are higher
for shear-thinning fluids for Re < 5, but are lower than Newtonian
and shear-thickening fluids for Re > 5 for any T/D. Hence for
Re > 5, the lower the value of n, the higher is the heat transfer
and lower is the pressure loss. Thus, shear-thinning fluids with
lower value of power-law index are preferable heat transfer agents
for Re > 5. Now, for any Re, increasing T/D from 1.5 to 2.5 both the
values of Nu and dP are found to increase. At Re = 1, the increase in
Nu from 1.5 to 2.5 is about 40% for any n. On the other hand, the
increase in dP is about 15% for shear-thinning fluids, and about
25% for Newtonian and shear-thickening fluids. With increasing
Re, the percentage increase in Nu decreases for all n. But, the per-
Fig. 12. The variation of the surface average Nusselt number, Nu with Re for n = 0.4– centage increase in dP increases with Re up to Re 6 10 after which
1.8 and T/D = 1.5–4.0.
it decreases until Re 6 40. Also, for the increase in T/D from 1.5 to
2.5, the percentage increase in both Nu and dP increases with
The maximum augmentation of heat transfer is observed here increasing value of power-law index for any Re. Overall, the deci-
at T/D = 1.5 for shear-thinning fluids. For instance, the percentage sion of choice between T/D = 1.5 and 2.5 depends on design
increase in the value of the average Nusselt number is found about requirements, since the increase in heat transfer results an increase
42% (at Re = 40) as the value of power-law index changes from 1 to in pressure loss.
0.4 for T/D = 1.5. On the other hand, the maximum enhancement in Now for increase in T/D = 2.5 to 4.0, both Nu and dP are higher
the value of the average Nusselt number is found about 36% (at for T/D = 4.0 until Re 6 10. Here, the percentage increase in heat
T/D = 2.5) and about 35% (at T/D = 4.0) for Re = 40. transfer reduces to near about 1% for Re = 10 from about 5% at
The Colburn heat transfer factor, jh is introduced here in order to Re = 1. But, the percentage increase in dP remains almost the same
establish a functional relationship between the Reynolds number, for Re = 1and 10. Now, for 10<Re 6 40, the values of Nu and dP are
power-law index and the average Nusselt number and is defined observed to be greater for T/D = 2.5 than 4.0. Here, the percentage
below increase in heat transfer from T/D = 2.5 to 4.0 is very low and only
about 1%, but the percentage increase in dP is as high as 30% in
Nu shear-thickening fluids and about 4% in shear-thinning fluids. For
jh ¼ ð14Þ
Re  Pr1=3 any Re, the percentage increase in dP and Nu increases with
 N.S.K. Chaitanya, A.K. Dhiman / International Journal of Heat and Mass Transfer 55 (2012) 5941–5958 5957

30 (a) T/D=1.5 30 (b) T/D=2.5 5. Conclusion

25 n=0.4 25 The 2-D flow and heat transfer of non-Newtonian (power-law)

n=0.6
n=1 fluids across a pair of circular cylinders in side-by-side arrange-
n=1.4
20 n=1.8 20 ment are studied for the varying values of Reynolds numbers
(1 6 Re 6 40), power-law indices (0.4 6 n 6 1.8), gap ratios
dP

dP
15 15 (1:5 6 T=D  4:0) for a Prandtl number of 50. The extensive
numerical results are presented such as viscous, pressure and total
10 10 drag coefficients and average Nusselt number. For T/D = 1.5, the
flow and heat transfer remain steady for shear-thinning fluids
5 5 (n < 1) for 1 6 Re 6 40. On the other hand, for the shear-thickening
fluids (n > 1) and Newtonian fluids (n = 1), the flow and heat trans-
0 0 fer exhibit unsteady behavior above the critical Reynolds number.
0 10 20 30 40 0 10 20 30 40
Re However, the flow and heat transfer are found to be steady for
Re
higher value of T/D ratios of 2.5 and 4.0 for all Re and n embraced.
(c) T/D=4.0 The value of the total drag coefficient is initially higher for shear-
30
thinning fluids than Newtonian and shear-thickening fluids for
Re  10, but a crossover in the values of drag occurs for Re > 10
25
after which an opposite trend in the values of total drag coefficient
20 is observed for all T/D. The average Nusselt numbers are found
higher for shear-thinning fluids than Newtonian and shear-thick-
ening fluids for all Re and T/D. The maximum enhancement in
dP

15
the heat transfer is found about 42% for Re = 40 and n = 0.4 at
10 T/D = 1.5.

5 References

[1] R.P. Chhabra, J.F. Richardson, Non-Newtonian Flow and Applied Rheology,
0 second ed., Butterworth-Heinemann, Oxford, 2008.
0 10 20 30 40
Re [2] R.K. Gupta, Polymer and Composites Rheology, second ed., Marcel Dekker, New
York, 2000.
[3] D. Sumner, S.S.T. Wong, S.J. Price, M.P. Paidoussis, Fluid behaviour of side-by-
Fig. 13. Variation of the pressure loss coefficient, dP with Re for n = 0.4–1.8 and T/
side circular cylinders in steady cross-flow, J. Fluids Struct. 13 (1999) 309–338.
D = 1.5–4.0. [4] T. Farrant, M. Tan, W.G. Price, A cell boundary element method applied to
laminar vortex shedding from circular cylinders, Computers Fluids 30 (2001)
211–236.
[5] J.R. Meneghini, F. Saltara, C. Siquiera, J.A. Ferrari Jr., Numerical simulation of
flow interference between two circular cylinders in tandem and side-by-side
increasing the power-law index (n) when T/D is increased from 2.5 arrangements, J. Fluids Struct. 15 (2001) 327–350.
to 4.0. In summary, T/D = 4.0 is preferable over T/D = 2.5 for [6] L. Chen, J.Y. Tu, G.H. Yeoh, Numerical simulation of turbulent wake flows
10 < Re 6 40, though the heat transfer is lower, because of the behind two side-by-side cylinders, J. Fluids Struct. 18 (2003) 387–403.
[7] S. Kang, Characteristics of flow over two circular cylinders in a side-by-side
considerably lower pressure loss characteristics observed in the arrangement at low Reynolds numbers, Phys. Fluids 15 (2003) 2486–2498.
case of T/D = 4.0. For Re < 10, the choice depends on design [8] N. Kondo, D. Matsukuma, Three-dimensional computation for flow around two
requirements. circular cylinders in side-by-side arrangement by a third-order upwind finite
element method, Int. J. Comput. Fluid Dyn. 19 (2005) 37–43.
Next, the heat transfer characteristics determined in the present [9] H. Ding, C. Shu, K.S. Yeo, D. Xu, Numerical simulation of flows around two
analysis could be related to real-time heat exchange devices, circular cylinder by mesh-free least square-based finite difference methods,
where the optimum set of physical parameters (Re, n, T/D) for a Int. J. Numer. Meth. Fluids 53 (2007) 305–332.
[10] L. Kun, M.A. Dong-Jun, S. De-Jun, Y. Xie-Yuan, Wake patterns of flow past a pair
prescribed heat exchange system could be determined. The analy-
of circular cylinders in side-by-side arrangements at low Reynolds numbers, J.
sis of flow patterns at various T/D provides extensive data regard- Hydrodyn. 19 (2007) 690–697.
ing the sizes of wakes that play an important role in increasing the [11] J. Mizushima, Y. Ino, Stability of flows past a pair of circular cylinders in a side-
pressure drag and hence the overall drag. A basic estimate of by-side arrangement, J. Fluid Mech. 595 (2008) 491–507.
[12] C. Liang, S. Premasuthan, A. Jameson, High-order accurate simulation of low-
streamlining required to reduce the drag for the cylinders could Mach laminar flow past two side-by-side cylinders using spectral difference
be made based on the analysis provided in this study. The onset method, Comput. Struct. 87 (2009) 812–827.
of unsteady flow results provided here, could be useful in the con- [13] S. Ryu, S.B. Lee, B.H. Lee, J.C. Park, Estimation of hydrodynamic coefficients for
flow around cylinders in side-by-side arrangement with variation in
struction of structures, cables, etc., that are exposed to power-law separation gap, Ocean Eng. 36 (2009) 672–680.
fluids. [14] A. Vakil, S.I. Green, Two-dimensional side-by-side circular cylinders at
Furthermore, the flow and heat transfer characteristics ob- moderate Reynolds numbers, Comput. Fluids 51 (2011) 136–144.
[15] N.P. Moshkin, J. Sompong, Numerical simulation of flow and forced convection
served here for T/D ratios of 2.5 and 4.0 corresponding to wake for- heat transfer in cross flow of incompressible fluid over two rotating circular
mations could be directly attributed to the flow across multiple cylinders, Suranaree J. Sci. Technol. 17 (2009) 87–104.
cylinders (3 or more), since it has been observed that the flow [16] T. Han, K.S. Yang, K. Lee, Heat transfer characterization of two isothermal
circular cylinders in proximity, ASME Trans. J. Heat Transfer 132 (2010)
and thermal patterns are separated double-body steady patterns, 0345041–0345044.
where each cylinder behaves as a single cylinder in cross-flow. [17] H.S. Yoon, J.H. Seo, J.H. Kim, Laminar forced convection heat transfer around
On the other hand, for the case of T/D = 1.5, since it has been two rotating side-by-side circular cylinder, Int. J. Heat Mass Transfer 53 (2010)
4525–4535.
observed that the flow and thermal patterns are single bluff-body
[18] R.P. Chhabra, Bubbles, Drops and Particles in Non-Newtonian Fluids, second
patterns and the flow around both the cylinders resembles flow edition., CRC Press, Boca Raton, 2006.
around a single cylinder. Thus, the inclusion of one or more cylin- [19] A.K. Dhiman, R.P. Chhabra, V. Eswaran, Effects of Reynolds and Prandtl
ders would induce flow interference into the current domain and numbers on heat transfer across a square cylinder in the steady flow regime,
Num, Heat Transfer, part A 49 (2006) 717–731.
determining the specific flow and heat transfer characteristics for [20] R.B. Bird, W.E. Stewart, E.N. Lightfoot, Transport Phenomena, second ed.,
such settings is a future scope of research. Wiley, New York, 2001.
5958 N.S.K. Chaitanya, A.K. Dhiman / International Journal of Heat and Mass Transfer 55 (2012) 5941–5958

[21] P. Sivakumar, R.P. Bharti, R.P. Chhabra, Steady flow of power-law fluids across [24] S. Bijjam, A.K. Dhiman, CFD analysis of two-dimensional non-Newtonian
an unconfined elliptical cylinder, Chem. Eng. Sci. 62 (2007) 1682–1702. power-law flow across a circular cylinder confined in a channel, Chem. Eng.
[22] R.P. Bharti, P. Sivakumar, R.P. Chhabra, Forced convection heat transfer from an Comm. 199 (2012) 767–785.
elliptical cylinder to power-law fluids, Int. J. Heat Mass Transfer 51 (2008) [25] R.P. Bharti, R.P. Chhabra, V. Eswaran, Steady forced convection heat transfer
1838–1853. from a heated circular cylinder to power-law fluids, Int. J. Heat Mass Transfer
[23] V. Sharma, A.K. Dhiman, Heat transfer from a rotating circular cylinder in the 50 (2007) 977–990.
steady regime, Therm. Sci. 16 (2012) 79–91.