Eur. Phys. J.

Plus (2017) 132: 532

DOI 10.1140/epjp/i2017-11791-2
THE EUROPEAN
PHYSICAL JOURNAL PLUS
Regular Article

Flow and heat transfer over a row of multiple semi-circular

cylinders: selection of optimum number of cylinders and effects
of gap ratios

Neeraj Parthasarathy1 , Amit Dhiman1,a , and Sandip Sarkar2,3

1
Department of Chemical Engineering, Indian Institute of Technology Roorkee, Roorkee - 247 667, India
2
Research Development and Scientific Services, Tata Steel Limited, Jamshedpur - 831001, India
3
Department of Mechanical Engineering, Indian Institute of Science, Bangalore - 560012, India

Received: 20 June 2016

Published online: 20 December 2017 – 
c Società Italiana di Fisica / Springer-Verlag 2017

Abstract. The present work aims at studying the laminar flow and heat transfer characteristics of a
Newtonian fluid over a row of semi-circular cylinders placed in uniform cross-flow configuration. The effects
of spacing to diameter (gap) ratio, varied from 1 to 10, on the flow and heat transfer patterns have been
studied at a Reynolds number of 100 for air as the working fluid. There has been no significant interaction
in the flow at a spacing ratio greater than 4 and each semi-circular cylinder behaved more or less like the
situation of a single cylinder. At lower gap ratios, the flow becomes more complicated due to the shear
layer interactions that exist in the flow. The streamline and isotherm contours have been presented and
discussed in detail for various gap ratios. The variation in flow behavior is explored further by studying the
drag coefficient and lift coefficient signals of various semi-circular cylinders. The presence of a secondary
frequency at low gap ratios that has adverse effect on the drag coefficient has been confirmed. The variation
of measured global quantities, namely the drag coefficient, the Strouhal number and the Nusselt number
have been studied and discussed in detail.

1 Introduction

The flow and heat transfer over bluff-bodies of various cross-sectional shapes like circular, semi-circular, square and
triangular have been a topic of sizable research, both experimentally and numerically, over the past few decades, due
to its varied applications in different fields of engineering. Out of the shapes mentioned, the circular cylinder shape
has gained maximum attention due to the fact that it is the simplest of the shapes with symmetry in both axes and
its vast application in various flow related processes. This has been summarized and reviewed in a detailed fashion in
the literature [1–3]. It is also evident that much of this work has been confined to a single cylinder as it presents as the
ideal case as well, the complexity in the flow is minimal. Though study of flow and heat transfer over a single cylinder
offers better understanding of how the shape of the body, and other factors like the Reynolds number, fluid behavior,
buoyancy, etc. affect the flow and thermal fields, its applications are very limited. A more realistic and pragmatic case
is multiple cylinders which are encountered in various flow and heat transfer applications like heat exchangers, power
generators, cooling towers, chimneys, etc. Keeping the applications aside, the necessity to study the flow is even more
justified by the complexity involved in the flow physics due to the interactions found in the wakes and vortex street
formed and the effects these have on the heat transfer.
The arrangement of multiple cylinders can be of three types, namely side-by-side, tandem and staggered configu-
rations. For each of these configurations, the factor that decides the variation of the flow phenomenon is the spacing
between the cylinders, i.e. spacing to diameter ratio in the cases of side-by-side and tandem configurations and trans-
verse to longitudinal distance in case of a staggered configuration. Much of the work on multiple cylinders (two or
more) has been concentrated on circular and square cross-sections. Flow over two cylinders is the simplest case and
has been studied in detail for circular and square cylinders. A few studies have also extended this to three or more
cylinders placed in a row parallel (tandem) and perpendicular (side by side) to the flow.
a
e-mail: [email protected] (corresponding author)
Page 2 of 23 Eur. Phys. J. Plus (2017) 132: 532

On the other hand not much attention has been showed to its counterpart, a semi-circular cylinder which is a
combination of a curved surface and a flat edge. A peek into the literature on semi-circular cylinders suggests that
only very limited information is available which mainly pertains to flow over a single object though the applications of
semi-circular cylinder are numerous like in novel heat exchanger design (better heat transfer characteristics) [4], food
processing, formation of weld-lines in polymer processing, cooling of electronic circuits and chips of various shapes, air
pre-heaters, probes and flow sensors etc. The numerous applications combined with the complexity in flow that arises
due to the flat edge and the fact that, unlike a circular cylinder, for a semi-circular cylinder the angle of incidence
of flow also comes into play, all these factors make the study of flow and heat transfer over semi-circular cylinders a
topic of research. The literature also suggests that, the work on flow over multiple semi-circular cylinders has been
very minimal and not tested for a range of conditions as done in the cases of circular and/or square cylinders. Hence,
the present study aims at studying the flow and heat transfer across multiple semi-circular cylinders.

2 Previous work

An important factor that influences the flow over two or more cylinders is the spacing between them. Detailed studies
are available on the influence of gap ratio (spacing to diameter ratio) on the flow over multiple circular cylinders.
Ishigai et al. [5], Bearman and Wadcock [6] and Williamson [7] did one of the few early experimental works which
have thrown light on the intricacies involved in the flow field around two circular cylinders at various gap ratios. They
observed that below a critical gap ratio (ε/D), the vortex shedding frequencies of the two cylinders differed from
each other but had certain harmonics. It was also noted that below the critical gap ratio, there existed instability
in the flow due to a repulsive force that existed between the cylinders. Above the critical gap ratio, the flow had a
synchronous in-phase or anti-phase behavior. The Von Karman Vortex Street from the flow past two or more cylinders
develops different patterns with regard to the spacing between the cylinders. These various patterns were studied
for a Reynolds number (Re) range of 500 to 3000 and for gap ratios ranging from 1 to 6 by Sumner et al. [8] and
they registered the development of three different flow patterns for various gap ratios; Single bluff-body flow (low
gap ratios), Biased synchronized shedding (intermediate gap ratios) and Symmetric synchronized shedding (high gap
ratios). Similar studies were made for a low Re range, the laminar regime, by Kang [9] and Liu et al. [10]. They
identified a total of nine patterns, four belonging to the steady laminar regime and five to the unsteady regime. The
results were similar to that of studies on high range of Re mentioned above hence concluding that vortex shedding
patterns are independent of the flow regime though there might be differences in the flow physics. Meneghini et al. [11]
numerically studied the flow over a pair of circular cylinders at various gap ratios and different configurations. They
presented the work over a Re range of 100 to 200 and gap ratios of 1.5, 2 and 3. For a side-by-side arrangement, a
power spectral analysis of the lift coefficient revealed a single peak corresponding to the vortex shedding frequency for
a gap ratio of 3, and in case of the lower gap ratios, a broad-banded spectrum with a peak which was not notably sharp
was obtained. Other significant works include [12–15] which depict the flow over a pair of cylinders for the Re ranging
from low to intermediate and at different values of spacing to diameter ratios. They have also detailed the physics
behind the vortex interactions and the shedding patterns and the bifurcation phenomenon, where both in-phase and
anti-phase sheddings occur at few gap ratios, has also been studied. The works discussed till now have restricted their
study mainly to the flow patterns generated at different working parameters, Chaitanya and Dhiman [16] and Daniel
and Dhiman [17] carried out forced and mixed convection studies, respectively, at low Re (1–40) for power-law fluids
over a pair of cylinders.
The flow over a pair of cylinders provides a basic idea on the flow interactions and vortex patterns, but it does
not generalize the case for flow over multiple cylinders (more than two). Though for larger gap ratios, the flow might
resemble that of over a pair of cylinders but at lower gap ratios the complexity increases manifolds. Tamada and
Fujikawa [18] were one of the first to address this issue wherein they used Oseen’s equation of motion to solve a
two-dimensional (2D) fluid flow over a row of infinite number, equal and parallel circular cylinders at low Re. They
observed that the drag force on any particular cylinder in the row was always greater than that for a single cylinder.
They attributed this to the interference in flow that existed between the cylinders. The same problem was solved using
the Stokes equation by Miyagi [19] who presented a numerical discussion on the variation of the drag coefficient with
the spacing to diameter ratio and also developed a simple equation to correlate both the parameters. Fornberg [20]
carried out his study on incompressible flow over a row of cylinders for Re ≤ 700, but limited his study to high gap
ratios ranging from 5 to ∞. He observed the wakes to increase in length at low values of Re, but at a certain gap ratio
of 16. The wakes that remained slender up to this point seemed to widen rapidly with increase in the value of the gap
ratio. He also concluded that at ∞ gap ratio, each cylinder behaved similarly to a single cylinder. An experimental
study conducted by Ishigai and Nishikawa [21] to study the flow structure developed when a gas flows over a row of
circular tube banks and came to a conclusion that at higher gap ratios, an anti-phase behavior was seen in the vortex
shedding patterns which became in-phase at lower gap ratios and with further reduction a jet flow phenomenon was
observed which coalesced due to the Coanda effect. A numerical study over a row of cylinders conducted by Huang
Eur. Phys. J. Plus (2017) 132: 532 Page 3 of 23

et al. [22] for unsteady laminar flow of water at a Re of 150 indicated an out-of-phase shedding of vortex shedding
patterns for gap ratios > 4 and the interaction between wakes was observed for gap ratios < 2.5.
Yamamoto and Hattori [23] were the first to analyze the flow and heat transfer of water flowing over a row of
heated circular tubes. The study was carried out for center-to-center distances of 2, 2.5, 3.3 and ∞ between the tubes,
and over a Re range of 75 to 500. The average Nusselt number was found to increase with decrease in center-to-center
distance owing to the increase in local velocities of the fluid around the cylinder.
Studies on flow and heat transfer over semi-circular shaped bluff bodies have gained interest only in the recent
years due to its applications in novel heat exchanger designs, food processing industries, processing of polymer suspen-
sions, submarines with flat base, air pre-heaters with semi-circular fins, micro channel heat exchangers (semi-circular
channels) etc. The studies available are mainly restricted to fluid flow over a single body but they give a clear and
an in-detail picture on the flow and thermal behavior of different types of fluids over and around a semi-circular
cylinder. Boisaubert et al. [24], Nada et al. [25] and Koide et al. [26] were ones of the very few to study the flow over
semi-circular bluff body experimentally. The former [24] tried to establish the effect of the body shape on the initial
development of a vortex using a semi-circular bluff body placed in two different configurations i.e. curved and flat
faces facing the flow. It was found that the Re at which the vortex shedding initiated was 190 for the curved face
facing the flow and 140 for the flat side facing the flow. The latter [25] examined the effect of the angle of attack on the
flow and heat transfer over a semi-circular cylinder. The average Nusselt number value was found maximum for the
curved face facing the flow at all angles of attack and the value of the average Nusselt number increased with increase
in the angle of attack. Koide et al. [26] studied the effect of the cross-sectional configuration of a cylindrical obstacle
using a circular, a semi-circular and a triangular cylinder to examine the role of the separation point movement on the
vortex shedding excitation. The vortex shedding excitation seems on all the obstacles, though the oscillation behavior
was considerably different among them. Numerical analysis of flow and thermal characteristics over a semi-circular
cylinder was initiated by Chandra and Chhabra [27–29] whose works included establishing the critical Re for the onset
of wake formation (0.55 < Re < 0.6) and the onset of vortex shedding (39.5 < Re < 40) for the curved side facing the
flow. Bhinder et al. [30] carried out the work by [27] numerically and found the existence of three separation zones
based on the angle of incidence for Re of 100 and using air (Pr = 0.71) as the working fluid. The works mentioned
till now discussed the flow in an unconfined domain, the same in a confined domain was studied by Kumar et al. [31],
Sukesan and Dhiman [32], Kumar and Dhiman [33] where the effects of blockage ratio (level of confinement) on the
flow and heat transfer were analyzed.
Thus, the detailed literature study clearly indicates that the attention that has been given to studying the flow
around multiple cylinders has not been equally spread to all cross-sectional shapes. No prior work is reported on flow
over two or more semi-circular cylinders though its applications in the heat transfer related processes are numerous.
Hence, the current study aims at filling the gap left in the literature by conducting a study on the flow and forced
convection heat transfer over a row of semi-circular tubes placed in a cross-flow configuration at a Re of 100 and
different spacing to diameter ratio ranging from 1 to 10. This work shall help better understand the influence that the
spacing between the cylinders has on the flow and heat transfer phenomena around semi-circular cylinders.

3 Mathematical formulation

The schematic representation of the problem being considered is displayed in fig. 1. Seven semi-circular cylinders of
the same dimensions (diameter “D”) are placed parallel to each other (in a single row) with the axis perpendicular
to the direction of flow and the curved surface facing the oncoming flow. The domain under consideration is of height
H, and the inlet and outlet at the distances of Xu (upstream) and Xd (downstream), respectively. The fluid stream
is considered to flow at a uniform velocity (U∞ ) with a uniform temperature (T∞ ). The semi-circular cylinders are
maintained at a constant temperature of Tw which is greater than that of the fluid temperature. The temperature
difference between the fluid and the object walls, i.e. (Tw − T∞ ) is kept low such that there is no variation in the
thermo-physical properties of the fluid with temperature and also the viscous dissipation effect can be neglected. The
edge-to-edge spacing (ε) remains the same between adjacent objects.

3.1 Governing equations

The dimensionless governing equations that represent a 2D unsteady laminar flow of an incompressible Newtonian
fluid are as follows.
Continuity equation:
∂u ∂v
+ = 0. (1)
∂x ∂y
Page 4 of 23 Eur. Phys. J. Plus (2017) 132: 532

Fig. 1. Schematic diagram.

Momentum equations:
 
∂u ∂(uu) ∂(uv) ∂p 1 ∂2u ∂2u
+ + =− + + 2 , (2a)
∂t ∂x ∂y ∂x Re ∂x2 ∂y
 2 
∂v ∂(uv) ∂(vv) ∂p 1 ∂ v ∂2v
+ + =− + + 2 . (2b)
∂t ∂x ∂y ∂y Re ∂x2 ∂y
Energy equation:  
∂θ ∂(uθ) ∂(vθ) 1 ∂2θ ∂2θ
+ + = + . (3)
∂t ∂x ∂y Re Pr ∂x2 ∂y 2
The dimensionless parameters are defined as follows:
u∗ v∗ p∗ T − T∞ x∗ y∗ t∗
u= , v= , p= ∗
, θ= , x= , y= , t=
U∞ U∞ ρU∞ Tw − T∞ D D (D/U∞ )
ρU∞ D μCp
Re = , Pr = . (4)
μ k

3.2 Boundary conditions

The boundary conditions that are imposed on the imaginary boundaries of the unconfined domain defined in the
present study are represented (in their dimensionless forms) as follows.
At the inlet: A uniform velocity of the fluid in the x-direction at a temperature of T∞ , i.e.
u = 1, v = 0, θ = 0. (5)
At the semi-circular cylinders walls: A no-slip condition is applied for the fluid flow and a constant temperature
condition is applied to the walls, i.e.
u = 0, v = 0, θ = 1. (6)
At the outlet: An outflow boundary condition is applied which is mathematically represented as
∂u ∂v ∂θ
= = = 0. (7)
∂x ∂x ∂x
At the upper and lower domain boundaries: A symmetric condition is applied, which is mathematically represented
as
∂u ∂θ
v= = = 0. (8)
∂y ∂y
Eur. Phys. J. Plus (2017) 132: 532 Page 5 of 23

3.3 Drag and lift coefficients

The overall drag force acting on the body due to the fluid flow is the sum of pressure and viscous drags. The drag
coefficient which is a direct representation of the drag force acting on the body is given by
FD
CD = CDP + CDF = 1 2
. (9)
2 ρU∞ D

While the drag force acts in the direction of the fluid flow, there is a force acting in the lateral or perpendicular
direction to the flow which is named as the lift force. This lift force is represented in terms of the lift coefficient which
is given by
FL
CL = 1 2 . (10)
2 ρU ∞D

3.4 Strouhal number

The Strouhal number is the dimensionless form of the vortex shedding frequency (f ) and is mathematically represented
as
fD
St = . (11)
U∞

3.5 Nusselt number

The dimensionless form of the heat transfer coefficient and thereby a direct indicator of the amount of heat transfer
is the Nusselt number,
hD
N uL = . (12)
k
While the above equation defines the local Nusselt number which provides insight into how the heat transfer changes
on the surface of the semi-circular cylinder, the overall Nusselt number is the surface averaged value of the local
Nusselt number that is required for various engineering applications. This is given by the following equation:

havg D 1
N uavg = = N uL · ds. (13)
k s s

4 Numerical methodology
The present study is carried out using a commercial flow modeling software ANSYS FLUENT. The grid on which the
study is conducted is made of unstructured quadrilateral elements of non-uniform spacing and is generated using
GAMBIT. The solver chosen is a 2D, unsteady, laminar and segregated one to solve the flow of and the forced convection
heat transfer from an incompressible Newtonian fluid. The discretization of the convective terms in the momentum
and energy equations is achieved by using QUICK, a third-order upwind scheme. The SIMPLE scheme is used to avoid
the coupling in pressure and velocity terms. A second order Adam-Bashforth implicit method is used to discretize
the time-dependent terms in the governing equations. The algebraic equations developed from the discretization of
the governing equations are then solved by the Gauss-Seidel method with the help of an algebraic multi-grid (AMG)
method based solver. Absolute convergence criteria of 10−15 have been used for the continuity, momentum equations
and 10−18 for that of the energy. Further, the convergence of a simulation is deemed to be achieved if a stable periodicity
is observed in the time signals of the lift and drag coefficients, and the Nusselt number values.

5 Choice of domain and grid

The domain and the grid fineness are matters of concern in the numerical study of a flow problem. They have a
profound effect on the consistency and the accuracy of the results. Thus, a prudent choice needs to be made in regard
to these parameters so as to reduce the numerical deviations in the results. The dimensions of the domain under
consideration as represented in fig. 1 need to be dealt with initially. Tables 1–3 summarize the values of different
engineering parameters at different values of the domain dimensions. As the Re comfortably represents a laminar
unsteady flow regime, the effect of the downstream distance is felt more than that of the upstream distance on the
Page 6 of 23 Eur. Phys. J. Plus (2017) 132: 532

Table 1. Effect of upstream distance on the values of flow and heat transfer parameters.

Upstream distance (Xu ) CD N uavg St

10 1.8696 6.1355 0.1613

15 1.8622 6.1328 0.1614

Table 2. Effect of downstream distance on the values of flow and heat transfer parameters.

Downstream distance (Xd ) CD N uavg St

30 1.8698 6.1355 0.1613

40 1.8656 6.135 0.1616

50 1.8646 6.1343 0.1616

Table 3. Effect of domain height on the values of flow and heat transfer parameters.

Height CD N uavg St

n(ε + D) 1.8656 6.1350 0.1616

(n + 1)(ε + D) 1.8491 6.1235 0.1604

results. The domain test has been carried out at a spacing to diameter ratio of 10 owing to the requirement of the
largest domain at this gap ratio. For the upstream distance, as observed from table 1, the percentage deviations in
the values are 0.4% for the mean drag coefficient, 0.04% for the average Nusselt number, and 0.06% for the Strouhal
number. The same can be said with table 2 (downstream test), where the values are comparable in all three cases with
a maximum deviation of about 0.22% in the mean drag coefficient value seen in case of 30D. Considering the fact,
however, that at lower gap ratios the disturbances in flow downstream would increase, a downstream length of 40D
is considered. The height of the domain as observed from the study seemed to have an effect on the objects at both
ends. While the deviations noted in other objects and the overall drag coefficient, the Nusselt number and the Strouhal
number are less. Though there is a considerable increase in the domain size for the domain height of (n + 1)(ε + D) to
obtain better accuracy in the results, this height has been chosen, where n is the number of semi-circular cylinders.
The grid structure used for the existing study is depicted in fig. 2. The grid is quadrilateral and unstructured
with finer elements near the objects in order to capture the boundary layer separation, near wake dynamics, and
other changes in the flow and heat transfer that are exceptionally more prominent in the regions near the objects.
The mesh in the regions behind each object is kept dense especially in order to better capture the flow and thermal
patterns at lower gap ratios where the interactions between layers of fluid are very predominant. Also, the wake
formation, boundary layer separation, and heat transfer patterns near the semi-circular cylinder at lower gap ratios
are very complex. Thus, the grid test is performed at the smallest spacing being studied in the problem, i.e., ε/D = 1.
Minimum grid spacing (δ/D) of 0.01 and 0.008 near the objects were chosen and the results of the test which have been
presented in table 4 suggest that the accuracy offered for the corresponding increase in the number of cells suggest
that a minimum dimensionless grid size of 0.01 is more suitable for further simulations to be carried out.

6 Results and discussion

The present study is carried out for the flow of air (Pr = 0.71) over a row of semi-circular cylinders at Re = 100.
The flow and heat transfer phenomena are numerically analyzed at spacing (ε) to diameter (D) ratios of 1, 2, 4,
6, 8 and 10. Prior to discuss the results obtained, it is important to validate the numerical methodology considered
Eur. Phys. J. Plus (2017) 132: 532 Page 7 of 23

Fig. 2. Representation of the grid distribution at ε/D = 4: (a) full computational domain; (b) closer-view around two adjacent
semi-circular cylinders; (c) closer-view around a single semi-circular cylinder.

Table 4. Effect of grid spacing on flow and heat transfer parameters.

δ
Control volumes on an object surface Total number of cells D
CD N uavg St

248 239197 0.01 3.9496 7.0667 0.2720

322 325467 0.008 3.9631 7.1285 0.2698

by comparing the results obtained from studying the flow and heat transfer around a single semi-circular cylinder
with the results available in the literature [27,30]. The reason for carrying out the validation on a single cylinder is
due to the lack of any available literature on the physical problem as defined in the present work. Tables 5 and 6
compare the values obtained from the present work with that given in the aforementioned works [27,30]. The work by
Chandra and Chhabra [27] involves flow and forced convection heat transfer at low Re and as the values in table 5
suggest, the results obtained are in excellent agreement with the literature. The study by Bhinder et al. [30] has been
carried out at a Re corresponding to that being studied here. The comparison of results shows a deviation of less
than 1%.
Page 8 of 23 Eur. Phys. J. Plus (2017) 132: 532

Table 5. Comparison of the results obtained in the present study with Chandra and Chhabra [27] for a single semi-circular
cylinder at different Re and Pr = 0.7.

CD N uavg
Re
Literature Present Literature Present

5 3.832 3.833 1.504 1.512

10 2.710 2.710 1.994 1.999

20 1.993 1.993 2.683 2.708

30 1.691 1.691 3.132 3.141

39.5 1.520 1.519 3.531 3.552

Table 6. Comparison of the results obtained in the present study with Bhinder et al. [30] for a single semi-circular cylinder at
Re = 100 and Pr = 0.71.

Output parameters Literature Present

CD 1.7992 1.8098

N uavg 6.0458 6.0918

Table 7. Values of mean drag coefficient, average Nusselt number, and Strouhal number for 5, 7 and 9 semi-circular cylinders
placed in a row for ε/D = 4.

Number of semi-circular cylinders (n) CD N uavg St

5 2.1380 6.3268 0.1807

7 2.1643 6.3472 0.1809

9 2.1788 6.3544 0.1816

6.1 Selection of optimum number of semi-circular cylinders

The optimum number of semi-circular cylinders required in order to simulate the flow over the row is stud-
ied by carrying out simulations for 5, 7 and 9 semi-circular cylinders and monitoring the variation in output
parameters like the drag coefficient, the Nusselt number, and the Strouhal number. The simulations are car-
ried out at a gap ratio of 4 and the results are as tabulated in table 7. This particular gap ratio was cho-
sen in accordance with a previous study conducted by Kumar et al. [34] on a row of square cylinders. The
variation in values seems to decrease with increase in the number of objects. With increase of objects in num-
ber from 5 to 7, a variation of 1.22% was observed in the drag coefficient though not much could be observed
in the other two quantities. This decreased to 0.66% when increased from 7 to 9 semi-circular cylinders. With
the help of the results obtained herein, the optimum number of objects required to simulate a row is set at
7 considering the fact that the variation in output parameters with increase in the semi-circular cylinders is
well within the allowable limits. The increase in number of objects also significantly increases the computational
domain and hence the requirement of computational resources and time. Thus, choosing 7 semi-circular cylin-
ders should prove computationally less demanding and it does not compromise the quality of the results ob-
tained.

6.2 Flow and thermal patterns

The study revealed that not much of interaction in flow was visible at higher gap ratios of 10, 8 and 6. The flow
patterns (figs. 3 and 4) for a gap ratio of 10 showed a synchronized in-phase flow behavior behind the semi-circular
cylinders. The flow behavior behind every object was almost equivalent to that behind a single semi-circular cylinder.
Eur. Phys. J. Plus (2017) 132: 532 Page 9 of 23

(a) (b)

80 80
C1 C1

70 70
C2 C2

60 60
C3 C3

50 50

y
C4 C4
y

40 40

C5 C5
30 30

C6 C6
20 20

C7 10 C7
10

0 0
0 10 20 30 40 50 0 10 20 30 40 50
x x

(c) (d)
45.5 45.5

45 45

0.36
44.5 44.5
0.25

44 C4 44 C4
y
y

43.5 43.5 0.63 0.11

43 43

42.5 42.5
9 9.5 10 10.5 11 9 9.5 10 10.5 11 11.5
x x

Fig. 3. Contours representing flow and heat transfer plotted at ε/D = 10: (a) streamlines; (b) isotherms and zoom-in on (c)
streamlines and (d) isotherms around the middle (C4) semi-circular cylinder.

Whereas in the cases of gap ratios 6 and 8, a mixture of both in-phase and anti-phase flow patterns was noted. But
in all three cases (ε/D ≥ 6), the values of the drag coefficient, the Nusselt number, and the Strouhal number were the
same for all the semi-circular cylinders with almost insignificant variation. This is due to the absence of any interaction
in the flow between the semi-circular cylinders. A similarity observed in the cases of flow patterns of 6 and 8 is that the
flow behind the semi-circular cylinders C3, C4, and C5 (as shown in figs. 3–5) always followed an anti-phase pattern
(phase difference of 180◦ ) as shown in fig. 5(a) (flow behind the aforementioned semi-circular cylinders C3 and C4 at a
gap ratio of 6). The observations made here match well with those of Ishigai and Nishikawa [21] and Huang et al. [22].
At a gap ratio of 4, though no interaction in flow is distinctly visible as observed from fig. 6, the values of engineering
parameters (CD , N uavg , and St) varied significantly between the semi-circular cylinders. A symmetry in the values
Page 10 of 23 Eur. Phys. J. Plus (2017) 132: 532

(a) (b)

70 70

C1 C1
60 60

C2 C2
50 50

C3 C3
40 40
C4

y
C4
y

30 30
C5 C5

20 20
C6 C6

10 C7 10 C7

0 0
0 10 20 30 40 49.9998 0 10 20 30 40 50
x x

(c) (d)
37.5 37.5

37 37

0.31
0.19
36.5 36.5 0.75

C4 36 C4
y

36
y

35.5 0.56
35.5
0.06

35 35

34.5 34.5
9 9.5 10 10.5 11 9 9.5 10 10.5 11
x x

Fig. 4. Contours representing flow and heat transfer plotted at ε/D = 8: (a) streamlines; (b) isotherms and zoom-in on (c)
streamlines and (d) isotherms around the middle (C4) semi-circular cylinder.

of the mentioned parameters was noted and an anti-phase behavior was observed between two adjacent semi-circular
cylinders as shown in fig. 4(a) (depicted here are the flow patterns behind objects C3, C4, C5 and C6). An anomaly
is experienced with the objects at the ends (C1 and C7) which seemed to behave in-phase with respect to the objects
adjacent to them. A definitive phase relationship existed in the vortex shedding and flow patterns developed in the
flow over the row of semi-circular cylinders at this gap ratio. There was no or very weak interaction seen between the
layers of fluid flow around and behind the semi-circular cylinders. The observations made here are in good agreement
with previously established results by Williamson [7] for circular cylinders at ε/D = 3.3 at Re = 100 and also by
Chatterjee et al. [35] for a row of square cylinders at a gap ratio of 4 but for Re = 150.
Eur. Phys. J. Plus (2017) 132: 532 Page 11 of 23

(a) (b)

50 C1 50 C1
45 45
C2 C2
40 40

35 C3 35 C3
30 30
C4 C4
y

y
25 25

20 C5 20 C5

15 C6 15 C6
10 10
C7 C7
5 5

0 0
0 10 20 30 40 0 5 10 15 20 25 30 35 40 45
x x

(c) (d)

36.5
36.5

0.06
36
36

0.12
35.5
35.5
0.25

35
y

C4
y

35 C4
0.5
34.5 0.75
34.5

34
34

33.5
9.5 10 10.5 11 11.5 9.5 10 10.5 11 11.5
x x

Fig. 5. Contours representing flow and heat transfer plotted at ε/D = 6: (a) streamlines; (b) isotherms and zoom-in on (c)
streamlines and (d) isotherms around the middle (C4) semi-circular cylinder.

With further reduction in the gap ratio (ε/D < 4), the flow becomes more complex. Lower gap ratios mark the
initiation in the interaction that takes place in the wakes and flow patterns downstream of the semi-circular cylinders
which have a profound effect on the values of the measured parameters like the drag coefficient, the Nusselt number
and other. From fig. 7 it is evident that mixing of flow is seen at a distance of about 5D downstream of the semi-circular
cylinders. This distance is also similar to that reported by Kumar et al. [34] for flow over a row of square cylinders
at Re = 80. The phase behavior in the flow patterns is not distinct as there is a mixture of both anti-phase as well
as in phase. As observed by Miyagi [19] for a row of circular cylinders at Re = 120, the wake patterns alternate i.e.
Page 12 of 23 Eur. Phys. J. Plus (2017) 132: 532

(a) (b)
40 40

C1 C1

30 C2 30 C2

C3 C3

20 C4 20 C4
y

y
C5 C5

10 C6 10 C6

C7 C7

0 0
0 10 20 30 0 10 20 30
x x

(c) (d)
21.5 21.5

21 21
0.06

20.5 20.5 0.50

20 C4 C4 0.44
y

20
y

0.25
19.5 19.5 0.69

19 19

18.5 18.5
9 9.5 10 10.5 11 11.5 9.5 10 10.5 11 11.5
x x

Fig. 6. Contours representing flow and heat transfer plotted at ε/D = 4: (a) streamlines; (b) isotherms and zoom-in on (c)
streamlines and (d) isotherms around the middle (C4) semi-circular cylinder.

the narrow and wider wake appear in an alternating fashion behind the row of cylinders as seen in fig. 7. In addition,
the presence of a secondary frequency which is a consequence of the flow interactions is observed at ε/D = 2 which
has the adverse effect on the behavior of drag signals which is discussed later. The wakes appear distinct immediately
behind the semi-circular cylinders, but this is lost further downstream.
From fig. 8(a), it can be concluded that at a gap ratio of unity the flow becomes complex with lateral as well as
transverse mixing seen in the fluid layers. The distinction that could be made in the wake formation before is completely
lost at this stage. The flow between the semi-circular cylinders follows what is called the jet flow phenomenon (which
is generally seen in flow through narrow gaps) where the fluid entering the gap is discharged out in the form of jets.
Eur. Phys. J. Plus (2017) 132: 532 Page 13 of 23

(a) (b)
24 24

C1 C1
20 20

C2 C2
16 16
C3 C3

12 C4
y

12 C4

y
C5 C5
8 8

C6 C6
4 4
C7 C7

0 0
10 15 20 25
10 15 20 25
x x

(c) (d)
13.5 13.5

13 13

12.5 12.5
0.75

0.5
12 C4 12 C4
y

11.5 11.5
0.31
0.44

11 11
0.1

10.5 10.5
9.5 10 10.5 11 11.5 9.5 10 10.5 11 11.5
x x

Fig. 7. Contours representing flow and heat transfer plotted at ε/D = 2: (a) streamlines; (b) isotherms and zoom-in on (c)
streamlines and (d) isotherms around the middle (C4) semi-circular cylinder.

The jets that are discharged into a closed space are incapable of maintaining the initial discharge direction and hence
are deflected. The direction and angle of deflection is guided by the Coanda effect as explained in [19]. Due to this
deflection, coalescence in flow is observed. The coalesced flow might be the result of interaction between two or more
jets.
The isotherm patterns that can be seen in figs. 3(b)–8(b) are the result of the fluid flow and hence have striking
resemblance to the flow patterns generated at various gap ratios. After studying the patterns closely, it is seen that
there is maximum crowding of the isotherms on the curved faces of the semi-circular cylinders (closer views also
represented in figs. 3(d)–6(d)) which seems to spread out at the rear side. The more crowding of the isotherms at the
Page 14 of 23 Eur. Phys. J. Plus (2017) 132: 532

(a) (b)
16 16

14 14
C1 C1

12 C2 12 C2

10 C3 10 C3

8 C4 8 C4
y

y
6 C5 6 C5

4 C6 4 C6

2 C7 2 C7

0 0
8 10 12 14 16 18 8 10 12 14 16 18
x x

(c) (d)
9 9

8.75 0.18

8.5 8.5 0.35

8.25

8 C4 C4 0.25
y

8
y

7.75

7.5 7.5 0.75 0.5

7.25

7 7
10 9.5 10 10.5
x x

Fig. 8. Contours representing flow and heat transfer plotted at ε/D = 1: (a) streamlines; (b) isotherms and zoom-in on (c)
streamlines and (d) isotherms around the middle (C4) semi-circular cylinder.

curved part points to the fact that heat transfer is higher than that in the rear region. At lower gap ratios, due to the
fluid mixing downstream there is better heat transfer characteristics expected well downstream of the semi-circular
cylinders. This observation is also explained by way of the variation in the Nusselt number in sects. 6.3 to 6.5.
Eur. Phys. J. Plus (2017) 132: 532 Page 15 of 23

(a)
0.8 2
C3 C3
0.6 C4 C4
1.95

0.4

1.9
0.2

CD
CL

0 1.85

-0.2
1.8

-0.4

1.75
-0.6

-0.8 1.7
460 470 480 490 460 470 480 490
t t

(b)
0.8 2.1
C3 C3
0.6 C4 2.05 C4

0.4
2

0.2
1.95
CD
CL

0
1.9
-0.2

1.85
-0.4

1.8
-0.6

-0.8 1.75
1400 1410 1420 1430 1440 1400 1410 1420 1430 1440
t t

Fig. 9. Time history of lift (left) and drag (right) coefficients of two adjacent semi-circular cylinders for (a) ε/D = 10 and (b)
ε/D = 8.

6.3 Time history of flow and heat transfer parameters

The variation in the time signals of drag and lift coefficients is determined for the flow regimes developed at various
gap ratios. The variations here are depicted by considering two common semi-circular cylinders for all gap ratios
studied. At higher gap ratios (≥ 6), the lift signal varies in a sinusoidal fashion with a set phase difference of either 0◦
or 180◦ based on whether the flow behavior is in-phase or anti-phase, respectively. The behavior of the signal in the
case of each semi-circular cylinder is almost akin to that of a single cylinder. The drag coefficient signals are in-phase
for all cases with a frequency that is double that of the lift coefficient signals. These observations can be confirmed
from the signals of drag and lift coefficients presented in figs. 9(a), (b) and 10(a). For a gap ratio of 4 (fig. 10(b)), the
signals are again distinctly sinusoidal and follow a distinct phase relation with the phase difference lying between 0◦
and 360◦ . The time averaged values of the lift signals for each object shows a mild deviation from the normal value
of zero, but the variations in the signals of the semi-circular cylinders seem to complement each other leading to the
Page 16 of 23 Eur. Phys. J. Plus (2017) 132: 532

(a)
0.8 2.2
C3 C3
0.6 C4 2.15 C4

0.4 2.1

0.2 2.05

CD
CL

0 2

-0.2 1.95

-0.4 1.9

-0.6 1.85

-0.8 1.8
1300 1305 1310 1315 1320 1300 1305 1310 1315 1320
t t
(b)
0.8 2.4
C3 C3
0.6 C4 2.35 C4
0.4 2.3

0.2 2.25
CL

CD

0 2.2

-0.2 2.15

-0.4 2.1

-0.6 2.05

-0.8 2
1740 1745 1750 1755 1760 1740 1745 1750 1755 1760
t t
Fig. 10. Time history of lift (left) and drag (right) coefficients of two adjacent semi-circular cylinders for (a) ε/D = 6 and (b)
ε/D = 4.

overall mean value of zero for the lift coefficient. The drag coefficient signals are again in-phase with a definitive phase
relationship as suggested for the lift coefficient. The frequency relation that existed between the signals of the drag
and lift coefficients at higher gap ratios is still maintained.
The complexity in the time signals of the drag and lift coefficients occurs at lower gap ratios (ε/D < 4). At ε/D = 2
though a sinusoidal variation is still present in the lift coefficient signals, damping is observed that repeats itself after
regular intervals. This damping that is observed is alternating for the semi-circular cylinders C3 and C4 as seen in
fig. 11(a). This can be due to the anti-phase behavior seen in the wake patterns. The damping that repeats itself also
suggests the presence of a secondary frequency other than the vortex shedding frequency (primary) as suggested in
Kumar et al. [34]. This secondary frequency is a consequence of the alternating wide and narrow wakes formed behind
the objects, as already suggested in the previous section, and hence it is named as the cylinder interaction frequency.
Till this point, the frequency (primary or vortex shedding) was calculated directly from the lift signal by using the
time period of the oscillations. The Strouhal number was then obtained by using the frequency as St = f D/U∞ .
Eur. Phys. J. Plus (2017) 132: 532 Page 17 of 23

(a)
1.5 3.4
C3 C3
C C
1 4 4
3.2

0.5
3

D
CL

C
2.8
-0.5

2.6
-1

-1.5 2.4
950 1000 1050 1100 1500 1510 1520 1530 1540 1550
t t

(b)
1 5
C3 C3
C C
4 4

0.5 4.5
D
CL

0 4
C

-0.5 3.5

-1 3
850 860 870 880 890 900 850 860 870 880 890 900
t t

Fig. 11. Time history of lift (left) and drag (right) coefficients of two adjacent semi-circular cylinders for (a) ε/D = 2 and (b)
ε/D = 1.

Due to the confluence of various frequencies that exist in the nature of the lift signal, to obtain the shedding frequency
a fast Fourier transform (FFT) is done on the time signal which is translated into the power spectral density plot as
shown in figs. 12(a), (b). The distinct peak observed at the corresponding Strouhal number value of 0.20 indicates the
primary or the dominant frequency. The not so distinct and small peaks observed at lower Strouhal number values
(0.01–0.02) suggest the presence of a second frequency.
Figure 11(b) shows the drag and lift signals for semi-circular cylinders C3 and C4 at ε/D = 1. It can be seen that
the signals are noisy or complex in nature. This can be attributed to the intense interactions present between and
immediately downstream of the semi-circular cylinders and also the jet flow phenomenon that occurs at the gaps. It
can be observed that with a reduction in the gap ratio, the amplitude of the drag signals seem to increase. This and
the complex nature are caused to the presence of the secondary frequency which seems to have a predominant effect
on the drag values. To establish the value of primary frequency, again a FFT is done on the lift signal and the result
is as represented in fig. 12. The peak observed at Strouhal number value of 0.28 corresponds to the primary or vortex
shedding frequency whereas the diffused and small peaks that can be seen at 0.02–0.03 are the result of the secondary
frequency.
Page 18 of 23 Eur. Phys. J. Plus (2017) 132: 532

3
10
101

101
10-1

Power Specctral Density

Power Spectral Desnity

-1
10

-3
-3 10
10

-5
10 10
-5

-7
10
-7
10

10
-9 (a) (b)

10-2 10-1 100 -2

10 10
-1
10
0

St St
Fig. 12. Power spectral density plot of the lift coefficient signal at (a) ε/D = 2 and (b) ε/D = 1.

The surface average Nusselt number signal also varies in accordance with the flow parameters like drag and lift
coefficients (as seen in figs. 13(a)–(f)). A sinusoidal variation is observed in the gap ratio ranging from 2 to 10 though
the signals are not clear and distinct at lower values in this range. Further reducing the gap results in a complex
nature being established in the signals which is indicative of the fact that thermal energy is transported by the flow
phenomenon and hence the behavior shall be in sync with the flow behavior.

6.4 Local Nusselt number

The variation in the local Nusselt number on the semi-circular cylinder surfaces for various gap ratios is represented
in fig. 14. After a keen observation of the plots, it can be noted that the heat transfer is the highest at the stagnation
point irrespective of the spacing between the semi-circular cylinders. The value of the maxima also increases with
the decrease in spacing. The local Nusselt number values at the rear edges seem to increase at lower gap ratios
(especially ≤ 2) with maximum variation at the lowest gap being studied. This is due to the introduction of the jet
flow phenomenon at the gaps and the increase in velocity in the regions near and around the objects at lower spacing
ratios. Also, at higher gap ratios (≥ 4) the variation with respect to each object seems to be similar and hence the
plots coincide. Whereas, at lower spacing or gap ratios the variation of the Nusselt number on the surface of each
object is distinct.

6.5 Overall variation of global quantities

The overall variation of the mean drag coefficient, the average Nusselt number, and the Strouhal number follows a
similar pattern. The values show a very mild increase with decrease in gap ratios at the higher range (≥ 4). With
further decrease in the spacing (< 4), the values increase steeply with the highest variation noted when the gap ratio
is reduced from 2 to 1. This observation shows that the maximum variation in the global parameters is observed when
the flow turns complex in nature. However, the Nusselt number showed an exception to this case as there was a sudden
increase observed when the gap ratio was reduced from 4 to 2. The enhancement in the value of the drag coefficient
is mainly due to the presence of the secondary frequency while that in the Strouhal number is attributed to the flow
interactions. The increase in value of the average Nusselt number could be linked to the increase in velocity due to
the jet flow phenomenon at the gaps as well near and around the semi-circular cylinders. One major observation in
the range of gap ratios studied that relates to findings by [18] is that the values of global parameters on each object
in the roware always greater than those observed in the case of a single semi-circular cylinder placed in the fluid flow.
The variation of the computed global quantities is as shown in figs. 15(a)–(c).
Eur. Phys. J. Plus (2017) 132: 532 Page 19 of 23

(a) (b)
6.14 6.18
C3
C4
6.1375

6.175
6.135
Nu

Nu
6.1325 6.17

6.13

6.165
6.1275

6.125 6.16
410 415 420 425 430 1000 1005 1010 1015 1020
t t

(c) (d)
6.245 6.38

6.24
6.37
Nu

Nu

6.235

6.36

6.23

6.35
6.225
1000 1005 1010 1015 1020 1000 1005 1010 1015 1020
t t

(e) (f)
6.9 7.3

7.2

6.85
7.1
Nu
Nu

6.8 7

6.9

6.75
6.8

6.7 6.7
1580 1585 1590 1595 1600 860 870 880 890
t t

Fig. 13. Time history of the Nusselt number of two adjacent semi-circular cylinders for (a) ε/D = 10, (b) ε/D = 8, (c) ε/D = 6,
(d) ε/D = 4, (e) ε/D = 2 and (f) ε/D = 1.
Page 20 of 23 Eur. Phys. J. Plus (2017) 132: 532

(a) (b)
12 12

10 10

8 8
L

L
6
Nu

Nu
6
C
4 4
B
2 2
A
0 0
0 0.5 1 1.5 2 2.5 0 0.5 1 1.5 2 2.5
A B C A A B C A
Position around an object Position around an object
(c) (d)
12 12

10 10

8 8
L

6
Nu

6
Nu

4 4

2 2

0 0
0 0.5 1 1.5 2 2.5 0 0.5 1 1.5 2 2.5
B C A B C A
A A
Position around an object Position around an object

(e) (f) 12
12

10 10

8 8
L
Nu L

Nu

6 6

4 4

2 2

0 0
0 0.5 1 1.5 2 2.5 0 0.5 1 1.5 2 2.5
B A B C A
A C A
Position around an object Position around an object
Fig. 14. Variation of the local Nusselt number on curved and rear surfaces of each semi-circular cylinder at gap ratios of (a)
ε/D = 10, (b) ε/D = 8, (c) ε/D = 6, (d) ε/D = 4, (e) ε/D = 2 and (f) ε/D = 1.
Eur. Phys. J. Plus (2017) 132: 532 Page 21 of 23

(a) (b)
5 0.4

0.35
4

0.3
3
CD

St
0.25

2
0.2

1
0.15

0 0.1
2 4 6 8 10 2 4 6 8 10
ε/D ε/D

(c)
7.5

7.25

7
Nu avg

6.75

6.5

6.25

6
2 4 6 8 10
ε/D
Fig. 15. Variation of (a) mean drag coefficient, (b) Strouhal number and (c) average Nusselt number at varying ε/D.

7 Conclusions

A detailed numerical study has been conducted to report the unsteady laminar flow and heat transfer characteristics
across a row of semi-circular cylinders at different spacing to diameter ratios. The optimum number of semi-circular
cylinders needed to simulate a row was found to be seven and this conclusion was made by studying the variation in
the output parameters for 5, 7 and 9 semi-circular cylinders placed in a row. The study revealed that at the higher
range of gap ratios (10 ≥ ε/D ≥ 6) there is almost no interaction found in the flow and the flow patterns showed a
synchronous phase behavior with a phase difference of either 0◦ (in-phase) or 180◦ (anti-phase). The successive gap
ratio studied, i.e. ε/D = 4 showed an anti-phase behavior in the flow patterns with anomaly detected in the objects at
both ends. The drag, lift and Nusselt number signals showed a sinusoidal behavior and the vortex shedding frequency
dominated the flow (absence of secondary frequency). At ε/D = 2, interactions in flow were noted downstream at a
distance of about 5D from the objects. The wakes immediately downstream were distinct and showed a combination
of both in-phase and anti-phase behaviors. An alternating pattern in the narrow and wide wakes was noted behind
the objects placed in the row. This led to the development of a secondary frequency which had its effect on the drag
and lift signals. Though the variation in the signals was sinusoidal, a damping pattern was noted in the lift signal and
the amplitude in case of drag signal seemed to have increased. The flow at further low gap ratios (ε/D = 1) becomes
Page 22 of 23 Eur. Phys. J. Plus (2017) 132: 532

complex due to interactions in the flow immediately downstream of the objects and the jet flow phenomenon that gets
initiated due to the very low gap between the objects. The drag, lift and Nusselt number signals are very noisy with a
lot of fluctuations. A secondary frequency has been found to dominate the flow at this stage whose effects on the drag
coefficient are observed in the signals with high amplitude. The jets of fluid downstream are observed to coalesce due
to the deflection in the direction of the jet discharged from the gaps which is explained by the Coanda effect.
The thermal patterns closely resembled the flow patterns owing to the assurance that the heat transfer is caused
by the fluid flow. The patterns seemed to concentrate on the curved face of the object indicating maximum heat
transfer in this area. Interactions in thermal patterns are seen at lower gap ratios which assures good heat transfer
characteristics well downstream of the object. The variation of the Nusselt number across the surface of the object
revealed that at lower gap ratios, the heat transfer at the edges of the object increases by a significant amount which
can be due to the increased local velocities observed in and around the object at these spacing ratios.
The mean drag coefficient, Strouhal number, and average Nusselt number seemed to vary similarly with variation
in spacing to diameter ratio. A very mild increase was observed in the values as the gap ratio decreased from 10 to
4; a still small but more significant increase was observed in the values when the gap ratio was reduced from 4 to
2, except for the Nusselt number which seemed to have a steep slope at this transition, and the maximum or a very
steep increase in the values was observed when the gap ratio was reduced to a value below 2. It was also found that
the values of the global engineering parameters of each object in the row was always greater than that observed in the
case of a single object placed in the fluid flow field. Investigation at various Re flows will be the scope for future work.

Nomenclature

CD Total drag coefficient (= FD / 12 ρU∞

2
) T Temperature (K)

CD Mean drag coefficient T∞ Fluid temperature at the inlet (K)

CL Lift coefficient (= FL / 12 ρU∞

2
) Tw Surface temperature of the object (K)

Cp Specific heat (J/kg K) u x velocity component

D Diameter (m) v y velocity component

f Frequency of vortex shedding (1/s) U∞ Average velocity (m/s)

FD Drag force (N/m) x Streamwise coordinate

FL Lift force (N/m) Xd Downstream distance (m)

h Local heat transfer coefficient (W/m2 K) Xu Upstream distance (m)

havg Average heat transfer coefficient (W/m2 K) y Crossways coordinate

H Height (m) Greek symbols

k Thermal conductivity (W/mK) ε Distance between the edges of two

N uL Local Nusselt number (= hD/k) adjacent semi-circular cylinders (m)

N uavg Average Nusselt number (= havg D/k) θ Temperature

p Pressure μ Fluid viscosity (Pa s)

Pr Prandtl number (= μCp /k) ρ Fluid density (kg/m3 )

Re Reynolds number (= DU∞ ρ/μ) Superscript

St Strouhal number (= f D/U∞ ) ∗ Dimensional value

t Time
Eur. Phys. J. Plus (2017) 132: 532 Page 23 of 23

References
1. M.M. Zdravkovich, Flow Around Circular Cylinders: Fundamentals, Vol. 1 (Oxford University Press, New York, 1997).
2. M.M. Zdravkovich, Flow Around Circular Cylinders: Applications, Vol. 2 (Oxford University Press, New York, 2003).
3. B.M. Sumer, J. Fredsøe, Hydrodynamics Around Cylindrical Structures (World Scientific, Singapore, 1997).
4. H. Ma, D. Oztekin, S. Bayraktar, S. Yayla, A. Oztekin, J. Heat Transf. 173, 051801 (2015).
5. S. Ishigai, E. Nishikawa, K. Nishimura, K. Cho, Bull. JSME 15, 949 (1972).
6. P. Bearman, A. Wadcock, J. Fluid Mech. 61, 499 (1973).
7. C.H.K. Williamson, J. Fluid Mech. 159, 1 (1985).
8. D. Sumner, S. Wong, S. Price, M. Paı̈doussis, J. Fluids Struct. 13, 309 (1999).
9. S. Kang, Phy. Fluids 15, 2486 (2003).
10. K. Liu, D. Ma, D. Sun, X. Yin, J. Hydrodyn. Ser. B 19, 690 (2007).
11. J. Meneghini, F. Saltara, C. Siqueira, J. Ferrari, J. Fluids Struct. 15, 327 (2001).
12. Z. Wang, Y. Zhou, Int. J. Heat Fluid Flow 26, 362 (2005).
13. Y. Peng, J. Mod. Phys. 4, 89 (2013).
14. C. Ng, N. Ko, J. Wind Eng. Ind. Aerodyn. 54/55, 277 (1995).
15. H. Zhang, Y. Zhou, Phys. Fluids 13, 3675 (2001).
16. N.S.K. Chaitanya, A.K. Dhiman, Int. J. Heat Mass Transfer 55, 5941 (2012).
17. A. Daniel, A. Dhiman, Ind. Eng. Chem. Res. 52, 17294 (2013).
18. K. Tamada, H. Fujikawa, Q. J. Mech. Appl. Math. 10, 425 (1957).
19. T. Miyagi, J. Phys. Soc. Jpn. 13, 493 (1958).
20. B. Fornberg, J. Fluid Mech. 225, 655 (1991).
21. S. Ishigai, E. Nishikawa, Bull. JSME 18, 528 (1975).
22. Z. Huang, J. Olson, R. Kerekes, S. Green, Comp. Fluids 35, 485 (2006).
23. H. Yamamoto, N. Hattori, Heat Transf. Jpn. Res. 25, 192 (1996).
24. N. Boisaubert, M. Coutanceau, P. Ehrmann, J. Fluid Mech. 327, 73 (1996).
25. S.A. Nada, H. El-Batsh, M. Moawed, Heat Mass Transf. 43, 1157 (2007).
26. M. Koide, T. Takahashi, M. Shirakashi, JSME, Ser. B 45, 249 (2002).
27. A. Chandra, R.P. Chhabra, Int. J. Heat Mass Transfer 54, 225 (2011).
28. A. Chandra, R.P. Chhabra, Appl. Math. Model. 35, 5766 (2011).
29. A. Chandra, R.P. Chhabra, Numer. Heat Transf. Part A: Appl. 63, 489 (2013).
30. A.P.S. Bhinder, S. Sarkar, A. Dalal, Int. J. Heat Mass Transfer 55, 5171 (2012).
31. A. Kumar, A. Dhiman, L. Baranyi, Int. J. Heat Mass Transfer 82, 159 (2015).
32. M.K. Sukesan, A.K. Dhiman, Int. Commun. Heat Mass Transf. 58, 25 (2014).
33. A. Kumar, A. Dhiman, Heat Transf. Eng. 36, 1540 (2015).
34. S.R. Kumar, A. Sharma, A. Agrawal, J. Fluid Mech. 606, 369 (2008).
35. D. Chatterjee, G. Biswas, S. Amiroudine, Int. J. Heat Fluid Flow 30, 1114 (2009).