International Journal of Heat and Mass Transfer 128 (2019) 319–334

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer

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

Forced convection heat transfer from a circular cylinder with a

flexible fin
Xu Sun a,⇑, Zehua Ye a, Jiajun Li a, Kai Wen a,b, Hui Tian c
a
Beijing Key Laboratory of Urban Oil and Gas Distribution Technology, China University of Petroleum-Beijing, Beijing 102249, China
b
State Key Laboratory for Manufacturing Systems Engineering, Xi’an Jiaotong University, Xi’an 710049, China
c
Department of Mechanical Engineering, Chengde Petroleum College, Chengde 067000, China

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

Article history: Forced convection heat transfer from a circular cylinder with a flexible fin in laminar flow with Re = 200
Received 6 May 2018 and Pr = 0.7 is investigated numerically. The two-dimensional incompressible Navier-Stokes equations
Received in revised form 24 August 2018 and energy equation are coupled with the Euler-Bernoulli beam equation to describe the flow-induced
Accepted 28 August 2018
vibration (FIV) of the flexible fin considering the convection heat transfer process. The modified
characteristic-based split scheme, Galerkin finite element method, semi-torsional spring analogy method
and loosely coupled partitioned approach are employed irrespectively for the flow and convection heat
Keywords:
transfer, fin vibration, mesh movement and fluid–structure interaction. The accuracy and stability of
Forced convection heat transfer
Circular cylinder
the proposed numerical method are validated using three benchmark models including the forced con-
Flexible fin vection heat transfer from a stationary cylinder, forced convection heat transfer from a transversely oscil-
Flow-induced vibration lating cylinder and FIV of a flexible plate behind a square cylinder. Finally, forced convection heat transfer
Fluid-structure interaction characteristics from a circular cylinder with a flexible fin with fin length l = 0.5D–1.5D (D is the cylinder
diameter) and elastic modulus E = 104 - 5
105 are analyzed in detail. The numerical results show that,
when the vortex shedding frequency approaches the natural frequency of the flexible fin, the FIV fre-
quency is locked on the natural frequency and the fin exhibits large-amplitude vibration. As a result,
the ‘dead water’ region behind the cylinder is reduced and the convection heat transfer is improved. In
the combinations of parameters considered, a maximum of 11.07% enhancement in heat transfer is
obtained by the flexible fin.
Ó 2018 Elsevier Ltd. All rights reserved.

1. Introduction
For the purpose of heating or cooling, heat exchangers are used
widely in the oil, chemical, refrigeration, power and electronic
industries. In most heat exchangers, the convective thermal resis-
tance is much larger than the conductive thermal resistance.
Hence, the convection heat transfer usually plays a decisive role
in the overall heat transfer process. To make the heat exchangers
more efficient and compact, the convection heat transfer enhance-
ment is very important and has attracted many attentions in the
past decades [1,2]. As is well-known, the convection heat transfer
from a solid wall is strongly associated with the near-wall flow fea-
tures. Generally speaking, the boundary layer and the ‘‘dead water”
regions are two major sources of convection thermal resistance. To
enhance convection heat transfer, one way is to introduce pertur-
bations to improve the advection between the boundary layer or
⇑ Corresponding author.
E-mail address: xsun@cup.edu.cn (X. Sun).
‘‘dead water” regions and the main flow. Following this idea, vari-
ous methods of convection heat transfer enhancement have been
proposed. One of them is the surface vibration method [1].
As the name implies, the surface vibration method employs
structure vibration to change the flow structure as well as heat
transfer in the vicinity of the wall. Based on the driving way of
the vibration, the surface vibration method can be divided into
two types: forced vibration method and self-excited vibration
method. For the forced vibration method, the heat transfer struc-
ture is driven mechanically or electrically by external energy input.
This method was first proposed in 1923 and its effectiveness for
forced convection heat transfers has been confirmed both experi-
mentally and numerically. For example, Scanlan [3] investigated
experimentally the effect of normal vibration on the forced convec-
tion heat transfer from a heating surface in water flows with
Re = 360, 720, 1460 and 2170. At certain combination of vibrating
amplitude, frequency and flow speed, up to 300% improvement
in local heat transfer coefficient was obtained. Takahashi and
Endoh [4] studied the forced convection heat transfer from a

https://doi.org/10.1016/j.ijheatmasstransfer.2018.08.123
0017-9310/Ó 2018 Elsevier Ltd. All rights reserved.
320 X. Sun et al. / International Journal of Heat and Mass Transfer 128 (2019) 319–334
Nomenclature
a thermal diffusivity
A cross-section area of the plate/fin normalized by D2
AC amplitude of the oscillating cylinder normalized by D
CD,av time-averaged drag coefficient of the cylinder
CL,max maximum lift coefficient of the cylinder
Dt real time step normalized by D=u1
Ds pseudo time step normalized by D=u1
D diameter of the circular cylinder
d displacement of a grid node in time interval½t n ; t nþ1 
normalized by D
d0 displacement of a grid node in time interval½t n1 ; t nþ1 
normalized by D
Dp pressure difference between the lower and upper sur-
face of the flexible plate/fin normalized by q1 u21
n local coordinate on the plate/fin normalized by D
E elastic modulus of the flexible plate/fin normalized
byq1 u21
f domain frequency of the flow-induced vibration nor-
malized by u1 =D
fC frequency of the oscillating cylinder normalized by
u1 =D
f n1 natural frequency for the first mode of the plate/fin nor-
malized by u1 =D
I second moment of area of the fin’s cross-section nor-
malized by D4
h fin thickness normalized by D
l length of the fin
0
l length of the fin normalized by D
L edge length of the square cylinder
NJ Hermite shape function at the node J of the plate/fin
NE total number of grid elements in the flow domain
NM total number of grid elements on the plate/fin
NP total number of grid nodes in the flow domain
nP total number of grid nodes connected to grid node P
Nu overall Nusselt number on the cylinder or cylinder with
a fin
vibrating sphere, a cylinder and a square-section tube in water
flows with Re = 3000–8000, 1000–2600 and 1500–4500, respec-
tively. In their experiments, convection heat transfer coefficients
from the sphere, cylinder and square-section tube were enhanced
respectively by about 200%, 200% and 180% at very small energy
dissipation ratio. Cheng et al. [5] investigated experimentally the
forced convection heat transfer characteristics of a circular cylinder
oscillating transversely in air flows with Re  4000. A maximum of
34% increase in heat transfer was obtained within the parameter
range considered. Using the transient test technique and smoke
flow visualization method, they also found that the heat transfer
enhancement is attributed to the lock-on effect and the turbulence
effect. Cheng and Hong [6] carried out a numerical study on the
influence of transversely oscillation on the forced convection heat
transfer from a cylinder in laminar flows with Re  300. The
enhancement of the convection heat transfer was achieved only
in the lock-on regime, and an approximate 13% enhancement
was obtained at Re = 200. Bronfenbrener et al. [7] studied theoret-
ically and experimentally the heat transfer from a tube with rings
rotating on the external surface. For the model they proposed,
about 40% enhancement in the heat transfer coefficient was
achieved by the vibration. Gau et al. [8] investigated the forced
convection heat transfer from an inline oscillating circular cylinder
in air flows with Re = 1600–3200. In their experiments, nearly 40%
enhancement in heat transfer was obtained although the oscillat-
ing amplitude was very small. They also found that the greatest
Nuav time-averaged overall Nusselt number on the cylinder
or cylinder with a fin
Nuav time-averaged mean Nusselt number on the cylinder
Nuh instantaneous local Nusselt number on the cylinder
Nuh,av time-averaged local Nusselt number on the cylinder or
cylinder with a fin
p pressure normalized by q1 u21
pr Prandlt number, pr ¼ m=a
Re Reynolds number, Re ¼ u1 D=m
q1 density of the incompressible flow
qS density of the flexible fin normalized byq1
St vortex shedding frequency from the cylinder normal-
ized by u1 =D
t real time normalized by D=u1
s pseudo time normalized by D=u1
sp oscillating period of the cylinder normalized by D=u1
sNu oscillating period of the overall Nusselt number on the
smooth cylinder or cylinder with a fin
T temperature
T1 temperature of the free stream
Tw wall temperature of the circular cylinder and fin
H non-dimensional temperature defined by
H ¼ ðT  T 1 Þ=ðT w  T 1 Þ
u1 velocity component of the free stream in x direction
ui velocity components of the flow field normalized by u1
ui intermediate velocity components of the flow field nor-
malized by u1
m fluid kinematic viscosity
w displacement of the plate/fin normalized by D
xi coordinate components of the flow domain normalized
by D
ymax Maximum vibrating amplitude at the free end of the
plate/fin normalized by D
enhancement in heat transfer can be achieved when the inline
oscillating frequency is closer to twice the natural vortex shedding
frequency. Yang and Fu [9] performed a numerical investigation on
heat transfer from a heated oscillating rectangular cylinder in cross
flows with Re = 250 and 500 using the arbitrary Lagrangian-
Eulerian (ALE) method, and an augment up to 115% was obtained
in heat transfer. Fu and Tong [10] investigated numerically the
flow and thermal fields of laminar flows with Re = 100, 200 and
500 over a heated transversely oscillating cylinder in a channel.
In the parameter range considered, a maximum of 28.7% increase
in convection heat transfer was achieved. Açikalin et al. [11] stud-
ied experimentally and numerically the application of the piezo-
electric fans in the cooling of low-power electronics. In their
study, the periodical vibration of a plate was generated based on
the reverse piezoelectric effect to improve the flow around a heat
source, and an enhancement up to 375% was obtained in heat
transfer. Ghazanfarian and Nobari [12,13] investigated numerically
the convection heat transfer from a rotating cylinder with cross-
flow and inline oscillations at Re  300. Their numerical results
also showed that the heat transfer coefficient of the circular cylin-
der can be increased significantly in the lock-on regime. Léal et al.
[14] studied the effect of dynamic deformation of the wall on the
convection heat transfer in a straight channel. In their numerical
study, the heat transfer coefficient was enhanced by 450% when
one wall of the channel was deformed in the form of a progressive
sinusoidal wave.
 X. Sun et al. / International Journal of Heat and Mass Transfer 128 (2019) 319–334 321

Different from the forced vibration method, the self-excited ing the amplitude of the oscillating fin can significantly enhance
vibration method takes advantage of the flow-induced vibration the Nusselt number.
(FIV) of flexible structures to disturb the flow field and enhance In previous studies such as Go et al. [15], Go [16] and Ali et al.
the convection heat transfer. For the self-excited vibration method, [26,27], the potential of flexible fins in enhancing forced convec-
vibration is sustained by the flow and no external energy input is tion heat transfers from the plain-wall heat sink and in the micro
needed. Therefore, it is easier to implement in practical applica- channel have been confirmed. Compared with heat transfer from
tions compared with the forced vibration method. With rapid a plain wall or in a channel, forced convection heat transfer from
developments of material science and mechanical manufacturing a circular cylinder is more common in engineering. Can the flexible
technology, this method of heat transfer enhancement has fin be used to enhance the forced convection heat transfer from the
attracted more and more attention in recent years, and many circular cylinder? To the best of our knowledge, little work has
new heat transfer structures based on FIV have been proposed. been done on this subject.
For example, Go et al. [15] and Go [16] proposed a microfin heat In this paper, a novel flexible fin model is proposed for circular
sink to improve cooling of highly compacted electronic compo- cylinder and its effectiveness in enhancement of the forced convec-
nents in high performance computers. In their design, the microfin tion heat transfer is investigated numerically using a FSI solver
was fabricated flexible enough to induce self-excited vibration in considering heat transfer process. The paper is organized as fol-
low-speed air flow. Taking advantage of FIV of a microfin array, lows: first, the mechanical model and mathematical formulation
the thermal resistance was decreased by up to 11.5% compared of the forced convection heat transfer from a circular cylinder with
with that of the traditional plain-wall heat sink. Cheng et al. [17], a flexible fin is presented in Section 2. Section 3 discusses in detail
Yan et al. [18], Ge et al. [19], Duan et al. [20] and Su et al. [21] pro- the numerical procedure. In Section 4, a numerical validation is
posed a novel heat exchanger using elastic planar or conical spiral carried out using three benchmark models. In Section 5, effects
tube bundles which could induce vibrations when there are waters of the fin length and rigidity on the forced convection heat transfer
flowing in and cross them. Both of the experimental and numerical from a circular cylinder with a flexible fin are analyzed in detail.
results showed that FIV can significantly enhance the heat transfer Finally, Section 6 is dedicated to the concluding remarks.
from the tube bundles. Additionally, in the experiments of Cheng
et al. [17], it was also found that the fouling resistance of the elastic
2. Mechanical model and mathematical formulation
tube bundle is only one-third of that of the ordinary rigid bare
tube. Park et al. [22] proposed a milli-scale flexible wing shape
Fig. 1 illustrates the mechanical model of the flexible fin. As
with a relatively large body and a narrow connecting leg. The flex-
seen in the figure, a flexible fin is attached to the lee side of a rigid
ible wing could exhibit large-amplitude FIV in a low-speed air flow
circular cylinder. When there is fluid flowing around, vortices will
and increase the heat transfer coefficient of a heated surface by
be generated and shed from the cylinder surface as long as the Rey-
11.3%. Using different fluid-structure interaction (FSI) solvers, Shi
nolds number is not too small. Excited by the vortices on the two
et al. [23] and Soti et al. [24] studied the influence of FIV of an elas-
sides, the flexible fin might exhibit vibration with apparent ampli-
tic thin plate attached to the lee side of a rigid circular cylinder on
tude, which could disturb the ‘‘dead water” behind the cylinder
the fluid mixing and heat transfer of a laminar channel flow. In the
and improve the convection heat transfer. In this model, the flexi-
study of Shi et al. [23], a maximum enhancement of 90.1% in the
ble fin not only increases the heat transfer surface but also works
average heat transfer coefficient of the channel wall was obtained
as a flow disturber.
at Re = 204.8–327.7. In the study of Soti et al. [24], effect of the dis-
In this paper, the forced convection heat transfer at low Rey-
turber on the channel flow and heat transfer at Re = 100, 200 and
nolds number (Re  200) is considered. The flow is assumed to
500 was investigated, and the results showed that the convection
be two-dimensional (2D) [12], incompressible and laminar, and
heat transfer on the channel walls can be increased up to 69% at
the effect of body forces such as gravity are ignored. Under these
Re = 500. Joshi et al. [25] studied the effect of the large-scale
assumptions, the 2D incompressible Navier-Stokes (NS) equations
flow-induced deformation of two flexible fins on the heat transfer
[24] are taken as the governing equations of the flow, which can be
in heated channel with laminar pulsating inflow. Their numerical
written as
results showed that the vortex ring created due to the flow-
induced deformation of the fins can enhance the fluid mixing @ui
and convection heat transfer in the channel. Ali et al. [26,27] also ¼0 ð1Þ
@xi
proposed a flexible vortex generator to enhance fluid mixing and
heat transfer of the laminar channel flow. In their investigations,
@ui @ui @p 1 @ 2 ui
several elastic flaps, which were inclined in the opposite direction þ uj ¼ þ ð2Þ
of the main flow, were mounted on the opposite parallel channel @t @xj @xi Re @xj @xj
walls. When there was fluid flowing in the channel, the elastic flaps
where the summation convention is used. On the cylinder and fin
vibrated and generated large-scale vortices detaching downstream,
surfaces, the non-slip boundary condition is imposed.
which could increase the fluid mixing and heat transfer in the
The fin is supposed of very high conductivity and the tempera-
channel. Arranging suitably the flaps near the channel inlet, up to
ture along the fin is assumed the same with that of the cylinder
275% heat transfer enhancement was obtained in Re = 1000–
surface. The wall temperature of the fin and cylinder (T w ) is higher
2000. Jamesahar et al. [28], Mehryan et al. [29] and Mehryan
than that of the fluid (T 1 ). The unsteady heat transfer process from
et al. [30] investigated numerically the free convection heat trans-
fer in a square cavity partitioned by a flexible membrane at various
conditions considering the diagonal and vertical partitions, mag-
netic field, incline of the cavity and sinusoidal temperature heat-
ing. The results showed that the large-scale deformation of the
membrane appearing at low elastic modulus or high Rayleigh
number is helpful to the heat transfer in the cavity. Ghalambaz
et al. [31] investigated the natural convection heat transfer over
a flexible oscillating fin in a square cavity. They found that increas- Fig. 1. A schematic of a circular cylinder with a flexible fin.
322 X. Sun et al. / International Journal of Heat and Mass Transfer 128 (2019) 319–334

the structure to the fluid and inside the fluid is described by the Hmþ1  Hm .To circumvent the Babuška-Brezzi (BB) restriction, split-
following dimensionless energy equation [24] ting Eq. (7) into three equations following the CBS algorithm yields
@H @H 1 @ H 2
Step 1 : ui
þ uj ¼ ð3Þ " #m
@t @xj Pr  Re @xj @xj  m
@ui 1 @ 2 ui Ds2 m @ @u
¼ um  Ds uj  þ uk uj i ð9Þ
The thermal boundary conditions H ¼ 0 and H ¼ 1 are imposed i
@xj Re @xj @xj 2 @xk @xj
on the free stream and the surfaces of the cylinder and fin,
respectively.    
@ @pmþ1 1 @ @pm
The thickness of the fin is much smaller than its length and only Step 2 : a ¼ ui  Dsð1  aÞ ð10Þ
@xi @xi Ds @xi @xi
the in-plane vibration is considered, so that the flexible fin is mod-
elled as a cantilevered Euler-Bernoulli beam, the governing equa-  m
@pmþa Ds2 m @ @p
tion of which can be written as [32] Step 3 : umþ1 ¼ ui  Ds þ uk
i
@xi 2 @xk @xi
@2w @4w 3um  4u n
þ u n1
qS A þ EI 4 ¼ Dp ð4Þ  Ds i i i
ð11Þ
@t 2
@n 2 Dt
For a cantilevered beam, the boundary conditions at the fixed In Eqs. (8)–(11), ui , um mþ1
i , ui , unþ1
i , pm , pmþ1 , pnþ1 , Hm , Hmþ1 and
and free ends are w ¼ 0; @w=@n ¼ 0 and H nþ1
are unknowns at each grid node at time instant tnþ1 , while uni ,
@ 2 w=@n2 ¼ 0; @ 3 w=@n3 ¼ 0, respectively, and the natural frequency n
pn , H and un1 , Hn1 are unknowns at the same spatial coordinates
i
for the first mode can be expressed by
at time t and t n1 . For flow problems with fixed boundaries, the
n
0 2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
f n1 ¼ ð1:875=l Þ EI=qS A=ð2pÞ. mesh is fixed during computation and uni , pn , Hn , un1 and Hn1
i
are also node values. For the moving boundary problem, however,
3. Numerical methods the mesh is moved during simulation when dynamic mesh is used,
so that uni , pn , Hn , un1
i and Hn1 are no longer node values. Hence,
3.1. Numerical method for the moving boundary flow with heat
uni , pn , Hn , un1 and Hn1 in Eqs. (8)–(11) are approximated by the
transfer i
   

node values uni node , pn jnode , Hn node , un1
i

node
and Hn1  at time t n
node
Because the flexible fin might deform with time under the fluid and t n1 using the Taylor expansions as
load, a flow problem with moving boundaries and heat transfer 8  @uni j
should be solved. To simulate such moving boundary flow prob- >
> n 2 @pn jnode 2
> ui ¼ ui node þdj @xj þoðjdj Þ; p ¼ p jnode þdj @xj þoðjdj Þ;
n node n n
>
>
lem, an implicit solver on moving mesh is proposed based on a >
>
<  
combination of the modified characteristic-based split (CBS) finite @ Hn j @un1 j
Hn ¼ Hn node þdj @xnode 2
þoðjdj Þ; un1 ¼ un1  þd0j i @x node þoðjd0 j Þ;
2

element method (FEM) [33] and the dual-time stepping (DTS) >
> j i i node j
>
>
>
> 
method [34]. >
: Hn1 ¼ Hn1  @ Hn1 j
þd0j i @x node þoðjd0 j2 Þ
First, adding respectively a derivative of ui and a derivative of H, i i
node j

with respect to a pseudo time s, to Eqs. (2) and (3) following the ð12Þ
DTS method [34], we have
The unstructured triangular mesh with linear shape function is
@ui @ui @ui @p 1 @ 2 ui employed for the velocity, pressure and temperature, and then Eqs.
þ þ uj ¼ þ ð5Þ
@s @t @xj @xi Re @xj @xj (8)–(11) are solved by the Galerkin FEM. More details about the
CBS FEM can be found in Zienkiewicz et al. [35] and Sun et al.
@H @H @H 1 @2H [33,36–38] and will not be presented here.
þ þ uj ¼ ð6Þ
@s @t @xj Pr  Re @xj @xj
3.2. Moving mesh method
Then, discretizing Eqs. (5) and (6) in the pseudo time using the
characteristic-Galerkin method [35] yields The semi-torsional spring analogy method proposed by Blom
" #
[39] is employed to adjust the unstructured triangular grids in
m @ui
m
1 @ 2 um @pmþa Ds2 m
mþ1
ui  ui ¼ Ds uj
m
 i
þ þ u the flow domain. Following the idea of the semi-torsional spring
@xj Re @xj @xj @xi 2 k
analogy method, at each real time step, the mesh at t n is regarded
 m
@ @u @p 3um  4uni þ un1 as a network of springs. Since the equilibrium length of each spring

uj i þ  Ds i i
ð7Þ
@xk @xj @xi 2 Dt is defined equal to the distance between the two grid nodes it con-
nects, there is no elastic force on each spring at t n and the whole
" #m spring network is initially in balance. At t nþ1 , the springs attached
mþ1 m @H 1 @2H Ds 2 m
H  H ¼ Ds uj  þ u to the fin are compressed or stretched first due to the structure
@xj Pr  Re @xj @xj 2 k
deformation, and the equilibrium state of the spring system is bro-
 m
@ @H 3Hm  4Hn þ Hn1 ken. Under the elastic forces, grid nodes in the flow domain will be

uj  Ds ð8Þ moved until the spring system reaches a new equilibrium state. In
@xk @xj 2Dt
this process, the displacement of the grid node P is governed by
where a is a control parameter and the superscripts m and n indi-
X
nP
cate the number of the pseudo and real time steps. The last terms kPQ ðdP  dQ Þ ¼ 0 ð13Þ
in Eqs. (7) and (8) are the discretized forms of @ui =@t and @ H=@t j¼1
in Eqs. (5) and (6), respectively. In each real time interval ½tn ; tnþ1 ,
where Q stands for the grid node connecting to P and kPQ is the stiff-
taking u0i ¼ uni , p0 ¼ pn and H0 ¼ Hn as the initial conditions and
ness of the spring between P and Q, which is defined as
solving Eqs. (7) and (8) iteratively, unknowns unþ1
i , pnþ1 and Hnþ1
2w
at t nþ1 can be obtained when umþ1  um pmþ1  pm kPQ ¼ /ðxnP  xnQ Þ =u ð14Þ
i i , and
 X. Sun et al. / International Journal of Heat and Mass Transfer 128 (2019) 319–334 323

where / and w are two control parameters and u is the angle facing mary, the FSI solution procedure at each real time step ½tn ; t nþ1  can
segment PQ. be divided into five steps:
Taking the displacement of each grid node on the fluid-fin inter-
face the same with that of the fin and making the grid nodes on 1. Compute the fluid load Dp on the beam based on the flow pres-
other flow boundaries fixed (i.e. d ¼ 0), displacements of the grid sure at tn ;
nodes inner the flow domain can be obtained by solving Eqs. (13) 2. Solve Eq. (20) to obtain the displacement and velocity of the
and (14). Finally, the new location of each grid node at t nþ1 can beam at t nþ1 ;
be computed by 3. Move the mesh in the flow domain by solving Eqs. (13)–(15);
xnþ1 ¼ xnP þ dP ð15Þ 4. Compute uni , pn , Hn , un1
i and Hn1 using Eq. (12);
P
5. Solve iteratively Eqs. (8)–(11) on the mesh at tnþ1 until the con-
vergence criteria is approached.
3.3. Numerical method for fin vibration
In step 5, the Aitken-D2 method [36] is utilized to accelerate the
The governing equation of the Euler-Bernoulli beam, namely Eq. iteration process, and the convergence criteria is defined as
(4), is solved using the standard Galerkin FEM and the generalized- 8 rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
> h i2
> PNP ðumþ1
a method. First, the beam is divided into many elements with the >
same length, and the displacement w along each element is >
> ÞJ ðum1 ÞJ
>
>
J¼1 1
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi < 106 ;
approximated by the displacements and velocities at its two end >
>
> PNP mþ1 2ffi
>
> ð 1 ÞJ
u
nodes using the Hermite shape function. Using the weighted >
>
J¼1
>
> rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
residual-Galerkin method, Eq. (4) can be discretized spatially as > PNP mþ1 m 2
>
<

pJ pJ
J¼1
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 4 ð21Þ
€ þ Kx þ B ¼ Q
Mx ð16Þ >
> P  2ffi < 10 ;
>
>
>
NP
pmþ1
>
> J¼1 J
where >
> qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
8 >
>
> P
2ffi
    T >
>
NP
Hmþ1 Hm
>
> >
> qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
J J 4
PNP
mþ1 2 < 10
J¼1
>
> x ¼ w ; @w
; w ; @w
   w ; @w
; >
>
>
> 1 @n 2 @n ðNMþ1Þ @n :
>
>
1 2 ðNMþ1Þ H
>
< Rl 0 Rl 2 T 2
0
J¼1 J

M ¼ qS A 0 NT Ndn; K ¼ 0 @@nN2 @@nN2 dn; ð17Þ

>
> l0 2 l0
>
> R l0  T 
> 3
Q ¼ 0 NT Dpdn; B ¼ EIð@@nw3 NT Þ  EI @@nw2 @N
> @n 
> 4. Code validation
>
> 0 0
:
N ¼ ½N1 ; N2 ;    N2ðNMþ1Þ :
Before solving the convection heat transfer from a circular
where B ¼ 0 is applied at the free end and the Dirichlet boundary cylinder with a flexible fin, the proposed numerical method for
condition, namely w1 ¼ 0 and ð@w=@nÞ1 ¼ 0, is imposed on the fixed convection heat transfer and FSI in laminar flows is applied first
end. to three benchmark problems for validation.
Then, applying the generalized-a method to Eq. (16) yields
€nþ1am þ Kxnþ1af þ B ¼ Q ðtnþ1af Þ 4.1. Forced convection heat transfer from a stationary circular cylinder
Mx ð18Þ

where Using the flow solver proposed in Section 3.1, forced convection
8 heat transfer from a stationary circular cylinder is computed first.
> €nþ1am ¼ ð1  am Þx
x €nþ1 þ am x€n ; Fig. 2 displays the computational model of this problem. As seen in
>
>
>
> xnþ1af ¼ ð1  af Þxnþ1 þ af xn ;
>
> the figure, the inlet and side boundaries are located respectively at
>
>
< Q ðt nþ1af Þ ¼ ð1  af ÞQ ðtnþ1 Þ þ af Q ðt n Þ; 10D and 12.5D from the cylinder centre, and the free-stream veloc-
ð19Þ ity and temperature, namely u = 1, v = 0 and H ¼ 0 are imposed on
>
> x_ nþ1 ¼ x_ n þ Dt½ð1  cÞx
€ n þ cx€nþ1 ;
>
> them. The outlet is placed at 20D downstream the cylinder, where
> xnþ1 ¼ xn þ Dtx_ n þ Dt 2 ½ð1  bÞx
> €n þ bx€nþ1 ;
>
> 2 the pressure is supposed to be zero. On the cylinder surface, the
>
: 211 1 2
am ¼ 1þ1 ; b ¼ 4 ð1  am þ af Þ ; af ¼ 1þ1 1
; c ¼ 12  am þ af no-slip condition, namely u = 0 and v = 0, is applied and the non-
dimensional temperature is H ¼ 1.
In Eq. (19) the control parameter 1 is set to be 0.1. Substituting Using three meshes with different grid densities, forced convec-
Eq. (19) into Eq. (18), the fully discretized form of Eq. (4) can be tion heat transfer from a circular cylinder at Re = 200 and Pr = 0.7 is
obtained as computed. For the three meshes, the computational time step is
 
ð1  am Þ
M þ ð1  af ÞK xnþ1
bDt2
 
ð1  am Þ ð1  am Þ ð1  am  2bÞ
¼M x n þ x_ n þ €
x n
b Dt 2 bDt 2b
 af Kxn þ Q ðtnþ1af Þ ð20Þ

which is solved by the Gauss-Seidel method.

3.4. FSI solution procedure

The loosely-coupled partitioned approach is used to couple the

flow and structure solvers. Grid nodes of the fluid and beam are
overlapping each other on the fluid-structure interface to decrease Fig. 2. Computational model of the forced convection heat transfer from a
the error of information transfer between the two solvers. In sum- stationary circular cylinder.
324 X. Sun et al. / International Journal of Heat and Mass Transfer 128 (2019) 319–334

Table 1
Grid information and computed flow and heat transfer parameters of a fixed circular cylinder.

Mesh NP NE CL,max CD,av St Nuav

Mesh_1 4413 7263 0.683 1.346 0.195 7.348
Mesh_2 7594 14,816 0.688 1.356 0.200 7.323
Mesh_3 13,912 27,369 0.685 1.356 0.200 7.316
Soti et al. [24] – – – 7.39
Churchill and Bernstein [40] 7.19

fixed at Dt ¼ 0:01 and the simulation is carried out from t = 0 to

500. The total numbers of grid node and element, maximum lift
coefficient, time-averaged drag coefficient and mean Nusselt num-
ber, and Strouhal number are presented in Table 1. As seen in the
table, the numerical results computed from Mesh_2 and Mesh_3
are almost the same and agree well with those reported by other
researchers [24,40]. Fig. 3 shows the distribution of the time-
averaged local Nusselt number on the cylinder surface. As seen
in the figure, the numerical results computed from Mesh_2 are in
good agreement with those reported by Momose and Kimoto
[41] and Zhang et al. [42]. Using Mesh_2, forced convection heat
transfers from a stationary circular cylinder at Pr = 0.7 and
Re = 20–200 are further computed. Fig. 4 displays the obtained
time-averaged mean Nusselt numbers. Again, the heat transfer
coefficients computed from the CBS method show a good agree-
ment with those reported in other literatures [42–44].

4.2. Forced convection heat transfer from a transversely oscillating

circular cylinder Fig. 4. Time-averaged convection heat transfer coefficients of a fixed circular
cylinder at Re = 20–200 and Pr = 0.7.

To examine the feasibility of the proposed flow solver for the

moving boundary flow with heat transfer, forced convection heat
transfer from a circular cylinder oscillating transversely in laminar
flow is computed. Fig. 5 displays the computational model of this
problem, which is very similar with that reported by Fu and Tong
[10] for the purpose of comparison. As seen in Fig. 5, the inlet of the
flow domain is located at 10D upstream the cylinder centre and the
free stream velocity and temperature are imposed. Two adiabatic
side walls are placed at 5D from the cylinder and the no-slip
boundary condition is applied. The outlet is located at 30D down-
stream, where the boundary condition is set as p = 0 and
@ H=@x ¼ 0. The cylinder is assumed to be isothermal with H ¼ 1.
Different from the case presented in Fig. 2, the cylinder is oscillat-
ing transversely following y ¼ AC sinð2pf C tÞ, which leads to a veloc- Fig. 5. Computational model of the forced convection heat transfer from a
ity boundary condition of u = 0 and v ¼ AC 2pf C cosð2pf C tÞ. transversely oscillating circular cylinder.

Fig. 3. Time-averaged local Nusselt number distribution on a stationary circular Fig. 6. Instantaneous distributions of the local Nusselt number on the oscillating
cylinder at Re = 200 and Pr = 0.7. cylinder at Re = 200, Pr = 0.71, AC = 0.4 and fC = 0.2.
 X. Sun et al. / International Journal of Heat and Mass Transfer 128 (2019) 319–334
Using grid density similar with Mesh_2 presented in Table 1, a
computational mesh with NP = 7173 and NE = 13,934 is generated
for this problem. Then, the forced convection heat transfer from
the oscillating cylinder at Re = 200, Pr = 0.71, AC = 0.4 and fC = 0.2
is computed. The time step is taken as Dt ¼ 0:01 and the unsteady
flow with heat transfer at t ¼ 0—500 is simulated. In Fig. 6, the
computed instantaneous distributions of the local Nusselt number
on the cylinder surface at t = 491.25 (sp =4) and t = 493.75 (3sp =4)
are presented and compared with those reported by Fu and Tong
[10]. As shown in the figure, the numerical results from the pro-
posed flow solver are in good agreement with those computed
from other method [10].
4.3. Flow-induced vibration of a flexible plate behind a square cylinder
The benchmark FSI problem involving laminar flow around a
rigid square cylinder with a flexible plate is computed to check
the accuracy and stability of the proposed FSI solver. Fig. 7 displays
the computational model of this FSI problem. As seen in the figure,
a flexible plate with 4L long and 0.06L thick is attached to the lee
side of a rigid square cylinder with a side length of L, which is taken
as the reference length in this problem. The inlet of the flow
domain is located at 5L upstream the cylinder centre and the
free-stream velocity, namely u = 1 and v = 0, is imposed. The two
side boundaries are placed at 6L up and down the cylinder, where
the velocity component in y direction is supposed to be zero. The
outlet is located at 14.5L downward the cylinder and the pressure
condition p = 0 is imposed on it. When there is fluid flowing
around, the flexible plate could be induced to vibrate under the
perturbation of the vortices shedding from the square cylinder.
Three meshes with different grid densities are used to computed
FIV of the flexible plate at Re = 332.6, qS ¼ 84:75, A = 0.06,
I ¼ 1:8
105 and E = 8.05
105. For the three meshes, the com-
putational time step is taken as Dt ¼ 0:01 and the simulation is
carried out from t = 0 to 500. Table 2 presents the grid information
of the three meshes and the numerical results computed from
Fig. 7. Computational model of laminar flow around a flexible plate attached to a
square cylinder.
Table 2
Grid information and computed domain frequency and maximum amplitude at
flexible plate tip.
Mesh NP NE NM f ymax
Mesh_1 2922 5600 50 0.0586
Mesh_2 6120 11,800 100 0.0576
Mesh_3 9708 18,804 150 0.0591
Wall and Ramm [45] 0.0604
Teixeira and Awruch [46] 0.0584
Dettmer and Peri [47] 0.0634
Bazilevs et al. [48] 0.0591
He and Zhang [49] 0.0586
325
them. As seen in the table, the computed vibrating displacement
and dominant frequency are in good agreement with those
obtained in other literatures [45–49].
5. Forced convection heat transfer from a circular cylinder with
a flexible fin
5.1. Computational model and grid independence tests
Using FSI solver proposed and validated above, forced convec-
tion heat transfer from a circular cylinder with a flexible fin is
investigated. Fig. 8 displays the computational model of this prob-
lem. As seen in the figure, the solution domain is enlarged slightly
while boundary conditions on the inlet, outlet, two side boundaries
and cylinder surface are exactly the same with those of the forced
convection heat transfer from a stationary cylinder as shown in
Fig. 2. A flexible fin with l long and 0.06D thick is attached to the
lee side of the circular cylinder. Under the excitation of the vortices
shedding from the cylinder, the flexible fin is allowed to deform
and vibrate. On the fin surface, the iso-thermal and no-slip bound-
ary conditions are imposed. As mentioned before, the temperature
on the fin surface is identical to that of the cylinder surface, and the
fluid and fin have the same displacement and velocity on the fluid-
structure interface.
Before studying heat transfer characteristics of the circular
cylinder with a flexible fin, the grid, time-step and domain size
independence tests are carried out. Five meshes are utilized to
compute the forced convection heat transfer from the circular
cylinder with a flexible fin at Re = 200, Pr = 0.7, l = 1D, qS ¼ 84:75,
A = 0.06, I ¼ 1:8
105 and E = 4
104. At this combination of
parameters, FIV of the flexible fin is in the lock-on regime and
the structure has very large vibrating amplitude, which will be dis-
cussed in detail later. Among the five meshes, Mesh_1, Mesh_2 and
Mesh_3 have the same domain size as depicted in Fig. 8 but with
different grid densities. Mesh_2, Mesh_4 and Mesh_5 have differ-
ent domain size but with similar grid distribution. For Mesh_4 with
smaller domain size, the inlet and two side boundaries are located
at 10D from the cylinder centre while the outlet is placed at 20D
downstream. For Mesh_5 with larger domain sized, the inlet and
two side boundaries are located at 15D from the cylinder centre
while the outlet is at 30D downstream. For Mesh_2 with medium
grid density, three time steps including Dt ¼ 0:005, 0.01 and 0.02
are employed to conduct a time-step independence test, while
for other meshes the time step is always set as Dt ¼ 0:01. In all
cases, the computation is carried out from t = 0 to 500.
Table 3 presents the grid information of the five meshes and the
numerical results obtained from them. As shown in the table, the
numerical results computed from Mesh_2 with Dt ¼ 0:01 are very
close to those obtained from Mesh_2 with Dt ¼ 0:005, Mesh_3
with Dt ¼ 0:01 and Mesh_5 with Dt ¼ 0:01. As a result, Mesh_2
1.251
1.205
1.228
1.20
1.35
1.25
1.21
1.24 Fig. 8. Computational model of the forced convection heat transfer from a circular
cylinder with a flexible fin.
326 X. Sun et al. / International Journal of Heat and Mass Transfer 128 (2019) 319–334

Table 3
Grid information and computed results of the forced convection heat transfer from a circular cylinder with a flexible fin at Re = 200, E = 4
104 and l = 1D.

Mesh NP NE NM Dt CD,av Nuav ymax f

Mesh_1 5551 10,868 25 0.01 1.182 24.869 0.452 0.210
0.02 1.200 26.080 0.459 0.198
Mesh_2 10,549 20,688 50 0.01 1.203 25.842 0.456 0.210
0.005 1.195 25.769 0.454 0.205
Mesh_3 15,299 29,900 100 0.01 1.203 26.232 0.456 0.215
Mesh_4 8432 16,483 50 0.01 1.209 25.839 0.455 0.226
Mesh_5 12,507 24,577 50 0.01 1.189 25.792 0.456 0.211

Fig. 9. Instantaneous results computed from Mesh_2: (a) total Nusselt number of the cylinder and fin; (b) displacement at the free end of the fin.

with Dt ¼ 0:01 is employed to study the forced convection heat
transfer from the cylinder with a 1D long flexible fin. For the cases
which the fin is not 1D long, computational meshes having the
same domain size and grid density with Mesh_2 are generated.
To show the performance of the FSI solver, Fig. 9 displays the
time histories of the total Nusselt number on the cylinder and fin
and displacement at the free end of the fin computed from Mesh_2
and Dt ¼ 0:01, and Fig. 10 presents the computational mesh at
three time instants corresponding to the maximum negative, zero
and maximum positive fin deformations. As seen in Fig. 9, at early
times in t 2 [0,150], both of the flow flied and fin response are in
the transitional stage from the initial conditions. Near t = 150, the
computation approaches the steady state and both of the Nusselt
number and displacement at the free end of the fin begin to oscil-
late periodically. With increase of time in t 2 [150,500], the oscil-
lating amplitudes of the Nusselt number and fin displacement
are not increased or decreased, which indicates the stability of
the FSI solver. As shown in Fig. 10, grids near and in the wake of
the cylinder and fin are refined to better describe the boundary
layer and vortex structures in these regions. During computation,
only the grids near the cylinder and fin are moved while most of
the grid nodes in the flow domain are almost keeping fixed. Com-
paring Fig. 9(a) with Figs. 9(b) and (c), it can be found that the qual-
ity of the unstructured triangular grids in the flow domain is not
deteriorated under the large-scale deformation of the flexible fin.
This indicates that the semi-torsional spring analogy method used
here is very robust.
5.2. Heat transfer characteristics of a circular cylinder with a flexible
fin
To study the effectiveness of the flexible fin in enhancing the
forced convection heat transfer from a circular cylinder at low Rey-
nolds numbers, heat transfer characteristics of the cylinder-fin
model proposed in Fig. 1 at Re = 200, Pr = 0.7, qS ¼ 84:75,
A = 0.06, I ¼ 1:8
105 , l = 0.5D–1.5D and E = 104–5
105 are com-
puted. Table 4 summarizes the geometrical and thermophysical
properties of the heat transfer model considered here.
Fig. 11 displays the overall time-averaged Nusselt numbers on
the surface of the circular cylinder with a flexible fin at l = 0.5D,
1.0D and 1.5D. In Fig. 11, the overall time-averaged Nusselt num-
bers from a smooth cylinder and the cylinder with a rigid fin are
also presented for comparison. As shown in Fig. 11(a), when a
0.5D long rigid fin is added to the leeside of the circular cylinder,
the overall heat transfer rate of the cylinder is not increased but
decreased by 2.78% from 22.984 to 22.345, although the heat trans-
fer surface is enlarged by 31.8%. This is coincident with that
reported by Abu-Hijleh [50], who found that using multiple short
fins tended to result in a lower Nusselt number than for the case
without any fins especially at low Reynolds numbers, due to the
buffer between the free stream and cylinder surface generated by
the short fins. From Fig. 11(a), it can also be found that replacing
the 0.5D long rigid fin by a flexible fin with the same length has
not improved the overall heat transfer rate. When the fin length
is increased from 0.5D to 1D, the heat transfer characteristics of
the cylinder-fin model is changed a lot. As seen in Fig. 11(b), when
a 1.0D long rigid fin is added to the cylinder, the overall time-
averaged Nusselt number is increased by 2.11% from 22.984 to
23.468, which means that the additional heat transfer from the
fin has overcome the reduction of heat transfer induced by its dis-
ruption to the near-wall flow. Moreover, when the rigid fin is
replaced by the flexible fin with the same length, the overall
time-averaged Nusselt number of the cylinder-fin model is chan-
ged significantly. As shown in Fig. 11(b), at very low elastic modu-
lus near E = 104, the heat transfer rate of the cylinder with a flexible
fin is very close to that from a rigid fin. When E is increased from
104 to 4.6
104, the overall time-averaged Nusselt number is
increased gradually from 23.468 to 26.066, and a maximum of
 X. Sun et al. / International Journal of Heat and Mass Transfer 128 (2019) 319–334
Fig. 10. Movement of Mesh_2 during simulation: (a) maximum negative fin deformation; (b) zero fin deformation; (c) maximum positive fin deformation.
11.07% heat transfer enhancement is achieved. Then, further
increasing E in [4.6
104, 5
105], the overall heat transfer rate
of the cylinder and fin is first dropped to the level of the case with
a rigid fin near 4.6
104 and then not changed any more. For the
cylinder with a 1.5D long fin, the heat transfer characteristics are
similar with the case with a 1.0D long fin. As shown in Fig. 11(c),
when a rigid fin with 1.5D long is attached to the leeside of the
cylinder, the heat transfer surface is further enlarged and the over-
all time-averaged Nusselt number is increased by 7.336% from
22.984 to 24.670. When the rigid fin is replaced by a flexible fin,
heat transfer from the cylinder-fin model is enhanced at E2[105,
2.9
105] while not changed much when E is too large or too
small. Near E = 2
105, a maximum Nusselt number of 26.098 is
obtained, which is increased by 5.79% compared with that from
the cylinder with a 1.5D long rigid fin.
Fig. 12 shows the distributions of the time-averaged local Nus-
selt number along the circular cylinder without fin, with a rigid fin
and with a flexible fin. As displayed in the figure, for the smooth
cylinder the maximum time-averaged local Nusselt number
locates at the front stagnant point while the minimum local Nus-
327
selt number locates at the two separation points. When a rigid
fin is attached to the rear of the cylinder, the distribution of the
local Nusselt number along the cylinder is changed significantly.
On the leeside surface between the two separation points, the
recovery of the heat transfer disappears and the local Nusselt num-
ber is dropped greatly. On the surface between the two separation
points and the front stagnant point, the local Nusselt number is
decreased slightly. On the fin surface, the local Nusselt number is
increased along the fin height. Near the root of the fin, the heat
transfer rate is lower than that of the smooth cylinder. Near the
tip of the fin, the heat transfer rate on the fin surface exceeds that
of the smooth cylinder. With the increase of the length of the rigid
fin, the local Nusselt number on the cylinder surface is not changed
much while heat transfer rate on the fin is increased gradually. As a
result, the overall heat transfer rate is decreased at l = 0.5D but
increased at l = 1.0D and 1.5D, as shown in Fig. 11. For the cylinder
with a 0.5D long flexible fin, the flexibility does not bring any
change to the distribution of the local Nusselt number on the sur-
face of the cylinder and fin, as seen in Fig. 12(a). At l = 1.0D and
1.5D, however, the local Nusselt number on the cylinder surface
328 X. Sun et al. / International Journal of Heat and Mass Transfer 128 (2019) 319–334

Table 4
Non-dimensional geometrical and thermophysical parameters of the cylinder-fin model computed.

Reynolds number of the inflow 200 Area of the fin’s cross-section 0.06
Prandlt number of the fluid 0.7 Second moment of area of the fin’s 1:8
105
Length of the fin 0.5, 1, 1.5 cross-section
Density of the fin 184.75
Height of the fin 0.06 Elastic modulus of the fin 104–5
105

Fig. 11. Overall time-averaged Nusselt numbers of the circular cylinder with a flexible fin: (a) l = 0.5D; (b) l = 1.0D; (c) l = 1.5D.

is increased at some elastic modulus although the heat transfer on
the fin itself is not improved much, as seen in Fig. 12(b) and (c),
which results in the increase of the overall Nusselt number in
Fig. 11(b) and (c).
5.3. Flow-induced vibration of the flexible fin
Fig. 13 shows the maximum vibrating amplitude at the free end
of the flexible fin. As seen in the figure, FIV of the flexible fin is very
weak at l = 0.5D, and the maximum vibrating amplitude in the
computed range of E is only 0.02. When a longer flexible fin with
l = 1D is attached to the cylinder, an apparent FIV is observed. As
shown in Fig. 13, when the elastic modulus of the 1D long flexible
fin is increased from 104 to 3.5
104, the vibrating amplitude at
the free end is increased by about 14 times from 0.032 to 0.445.
Then, FIV amplitude is not changed much further increasing E from
3.5
104 to 4.6
104. When the elastic modulus is increased
slightly from 4.6
104, the vibrating amplitude is dropped greatly
from 0.45 at 4.6
104 to 0.217 at 5
104 and 0.032 at 6
104.
Finally, FIV amplitude is not changed much in E2[6
104,
5.5
105]. At l = 1.5D, the response of the flexible fin is similar with
that at l = 1.0D. As seen in Fig. 13, a peak appears in E2[1.5
105,
2.7
105] on the amplitude curve of l = 1.5D, and a maximum
amplitude of 0.408 is obtained at E = 1.9
105. Comparing Fig. 13
with Fig. 11, it can be found that the overall heat transfer rate of
the cylinder-fin model is strongly related to the vibrating ampli-
tude of the flexible fin. When FIV is weak and its amplitude is
small, the flexible fin behaves like a rigid fin and the heat transfer
is not improved. When large-amplitude FIVs appear in E2[104,
6
104] at l = 1.0D and E2[1.5
105, 2.7
105] at l = 1.5D, the heat
transfer from the cylinder and fin is enhanced significantly.
Fig. 14 displays the dominant FIV frequencies of the three flex-
ible fins with respect to the elastic modulus, and compares them
with the natural frequencies for the first mode of the fin and the
 X. Sun et al. / International Journal of Heat and Mass Transfer 128 (2019) 319–334
Fig. 12. Time-averaged local Nusselt number on the cylinder and fin: (a) l = 0.5D; (b) l = 1.0D; (c) l = 1.5D.
Fig. 13. Maximum FIV amplitude at the free end of the flexible fin.
vortex shedding frequencies of the cylinder with the rigid fins. As
seen in Fig. 14(a), the vortex shedding frequency corresponding
to the 0.5D long fin is far from the natural frequency in the whole
range of elastic modulus considered, and the flexible fin is vibrat-
ing under the vortex shedding frequency. When the 1.0D long flex-
ible fin is attached to the cylinder, however, the FIV frequency is
changed a lot. As seen in Fig. 14(b), when the vortex shedding fre-
329
quency approaches to the natural frequency of the flexible fin in E2
[2
104, 6
104], FIV of the flexible fin is locked on the natural fre-
quency. In other regions, the vortex shedding frequency is far from
the natural frequency, the fin is still vibrating under the vortex
shedding frequency. Comparing Fig. 13 with Fig. 14 (b), it can be
found that it is the lock-on between the FIV frequency and the nat-
ural frequency that results in the large-amplitude vibration of the
fin. At l = 1.5D, the lock-on phenomenon also happens. As seen in
Fig. 14(c), FIV frequency of the flexible fin is locked on the natural
frequency in E2[1.4
105, 2.7
105], which also results in large-
amplitude vibration as shown in Fig. 13. In other regions, FIV fre-
quency of the 1.5D long flexible fin is almost the same with the
vortex shedding frequency.
5.4. Instantaneous flow and thermal fields
To understand above heat transfer and FIV characteristics of a
circular cylinder with a flexible fin in laminar flow, the instanta-
neous vortex structures and temperature distribution around the
cylinder and fin are further analyzed. Figs. 15 and 16 display the
instantaneous streamlines and thermal fields around a smooth
cylinder and a cylinder with a 0.5D long fin. As shown in Figs. 15
(a) and 16(a), for a smooth cylinder in laminar flow with
Re = 200, two vortices are generated and shed alternatively from
the two sides of the cylinder, and the hot fluid near the cylinder
surface is transported into the wake of the cylinder by the shed-
330 X. Sun et al. / International Journal of Heat and Mass Transfer 128 (2019) 319–334

Fig. 14. FIV frequency at the free end of the flexible fin: (a) l = 0.5D; (b) l = 1.0D; (c) l = 1.5D.

(a)

(b)

(c)

(d)

t=0 t = τ Nu 4 t = τ Nu 2 t = 3τ Nu 4 t = τ Nu

Fig. 15. Instantaneous streamlines around a circular cylinder with a 0.5D long fin: (a) smooth cylinder; (b) rigid fin; (c) flexible fin at E = 104; (d) flexible fin at E = 5
105.
 X. Sun et al. / International Journal of Heat and Mass Transfer 128 (2019) 319–334 331

(a)

(b)

(c)

(d)

t=0 t = τ Nu 4 t = τ Nu 2 t = 3τ Nu 4 t = τ Nu

Fig. 16. Instantaneous temperature distribution around a circular cylinder with a 0.5D long fin: (a) smooth cylinder; (b) rigid fin; (c) flexible fin at E = 104; (d) flexible fin at
E = 5
105.

(a)

(b)

(c)

(d)

t=0 t = τ Nu 4 t = τ Nu 2 t = 3τ Nu 4 t = τ Nu

Fig. 17. Instantaneous streamlines around a circular cylinder with a 1D long fin: (a) smooth cylinder; (b) rigid fin; (c) flexible fin at E = 104; (d) flexible fin at E = 4.6
104.

ding vortices. Since the regions near the two separation points are
always covered by the vortices, the thermal boundary layer near
there is thicker and the local Nusselt number is minimum, as seen
in Figs. 16(a) and 12. When a 0.5D long rigid fin is added to the
cylinder, the two vortices are enlarged and the thickness of ther-
mal boundary layer in the lee side region is increased, as seen in
Fig. 15(b) and 16(b). As a result, the local Nusselt number on the
leeside of the cylinder is decreased in Fig. 12(a). When a flexible
fin with 0.5D long is attached to the cylinder, because the FIV is
too weak, the vortex structure and thermal field around the cylin-
der and fin are almost unchanged, as seen in Figs. 15(c), (d), 16(c)
and (d). Hence, the cylinder with a 0.5D long flexible fin shows
nearly the same heat transfer features with the cylinder with a
0.5D long rigid fin, as shown in Figs. 11(a) and 12(a).
Figs. 17 and 18 show the instantaneous streamlines and ther-
mal fields around a smooth circular cylinder and a cylinder with
a 1D long fin. Comparing Fig. 17(b) with Fig. 15(b), it can be seen
that, when the length of the rigid fin is increased from 0.5D to
1D, the vortices shed from the cylinder are further enlarged and
divided into smaller ones during the shedding process. With
increase of the fin length, more part of the fin is extended out of
the ‘dead water’ region behind the cylinder and contacts with the
colder fluid as shown in Fig. 18(b), which results in the enhance-
ment of the heat transfer from the cylinder as seen in Figs. 11(b)
332 X. Sun et al. / International Journal of Heat and Mass Transfer 128 (2019) 319–334

(a)

(b)

(c)

(d)

t=0 t = τ Nu 4 t = τ Nu 2 t = 3τ Nu 4 t = τ Nu

Fig. 18. Instantaneous temperature distribution around a circular cylinder with a 1D long fin: (a) smooth cylinder; (b) rigid fin; (c) flexible fin at E = 104; (d) flexible fin at
E = 4.6
104.

(a)

(b)

(c)

(d)

t=0 t = τ Nu 4 t = τ Nu 2 t = 3τ Nu 4 t = τ Nu

Fig. 19. Instantaneous streamlines around a circular cylinder with a 1.5D long fin: (a) smooth cylinder; (b) rigid fin; (c) flexible fin at E = 2
104; (d) flexible fin at E = 2
105.

and 12(b). When the 1D long rigid fin is replaced by a flexible one
with the same length, the situation is changed a lot. For a flexible
fin out of the lock-on regime, FIV amplitude is small and the vortex
structure and temperature distribution around the cylinder and fin
are similar with those of the cylinder with a rigid fin, as seen in
Fig. 17(c) and 18(c). When the flexible fin is in the lock-on regime,
the large-amplitude FIV of the flexible fin changes significantly the
flow and thermal fields. As shown in Fig. 17(d), under the influence
of the large deformation of the flexible fin, the vortices generated
on the cylinder surface are shed before growing up. As a result,
the ‘dead water’ region behind the cylinder is reduced a lot as
shown in Fig. 18(d) and the heat transfer is enhanced significantly
compared with that from a cylinder with a 1D long rigid fin, as
shown in Fig. 11(b) and 12(b).
Figs. 19 and 20 display the instantaneous streamlines and ther-
mal fields around a smooth circular cylinder and a cylinder with a
1.5D long fin. As shown in the figures, the flow and heat transfer
from a cylinder with a 1.5D long fin are similar with those from
a cylinder with 1D long fin as shown in Figs. 17 and 18. When
the flexible fin is not in the ‘lock-on’ regime, FIV amplitude of the
fin is very small that it cannot affect much the streamlines and
thermal field around the cylinder and fin, as seen in Figs. 19(b),
(c), 20(b) and (c). When the flexible fin is in the ‘lock-on’ regime,
however, the large-amplitude FIV of the flexible fin advances the
 X. Sun et al. / International Journal of Heat and Mass Transfer 128 (2019) 319–334
(a)
(b)
(c)
(d)
t=0 t = τ Nu 4
Fig. 20. Instantaneous temperature distribution around a circular cylinder with a 1.5D long fin: (a) smooth cylinder; (b) rigid fin; (c) flexible fin at E = 2
104; (d) flexible fin
at E = 2
105.
vortex shedding process, which again decreases the ‘dead water’
region behind the cylinder and improve the convection heat trans-
fer from the cylinder and fin, as shown in Figs. 19(d) and 20(d).
6. Conclusions
Application of FIV of the flexible fin in convection heat transfer
enhancement of a circular cylinder in laminar flow is investigated
using a CBS-based FSI solver considering heat transfer. Based on
the numerical results obtained, several conclusions can be drawn
as follows: (1) Combining the modified CBS scheme for moving
boundary flows proposed in [33] with the deal-time stepping
method, semi-torsional spring analogy method, Galerkin finite ele-
ment method, generalized-a method and the loosely coupled par-
titioned method, accurate and stable numerical solution
procedures can be developed for the moving boundary flow prob-
lems with heat transfer and fluid-structure interaction problems
with heat transfer in laminar flows. (2) When the frequency of vor-
tex shedding from a circular cylinder with a flexible fin approaches
the natural frequency, the frequency of FIV will be locked on the
natural frequency. At the same time, the vibrating amplitude of
the flexible fin will be amplified greatly. (3) When the flexible fin
is out of the lock-on regime, the amplitude of the flow-induced
vibration is very small, and the circular cylinder with a flexible
fin has almost the same flow and heat transfer characteristics with
the circular cylinder with a rigid fin. When the flexible fin is in the
lock-on regime, however, the large-amplitude vibration of the flex-
ible fin will advance the vortex shedding process from the cylinder
and fin and reduce the ‘dead water’ region behind the cylinder. As a
result, the forced convection heat transfer from the cylinder and fin
can be enhanced significantly.
Conflict of interest
None.
Acknowledgements
This work is supported by the National Natural Science Founda-
tion of China (No. 51506224) and Natural Science Foundation of
333
t = τ Nu 2 t = 3τ Nu 4 t = τ Nu
Hebei province China (No. E201641100). The author would like
to thank for the kindly support of these foundations.
References
[1] A.E. Bergles, R.M. Manglik, Current progress and new developments in
enhanced heat and mass transfer, J. Enhanc. Heat Transf. 20 (1) (2013) 1–15.
[2] S. Liu, M. Sakr, A comprehensive review on passive heat transfer enhancements
in pipe exchangers, Renew. Sustain. Energ. Rew. 19 (2013) 64–81.
[3] J.A. Scanlan, Effects of normal surface vibration on laminar forced convection
heat transfer, Ind. Eng. Chem. 50 (10) (1958) 1565–1568.
[4] K. Takahashi, K. Endoh, A new correlation method for the effect of vibration on
forced-convection heat transfer, J. Chem. Eng. Japan 23 (1) (1999) 45–50.
[5] C.H. Cheng, H.N. Chen, W. Aung, Experimental study of the effect of transverse
oscillation on convection heat transfer from a circular cylinder, J. Heat. Transf.,
Trans. ASME 119 (1997) 474–482.
[6] C.-H. Cheng, J.L. Hong, Numerical prediction of lock-on effect on convection
heat transfer from a transversely oscillating circular cylinder, Int. J. Heat Mass
Transf. 40 (8) (1997) 1825–1834.
[7] L. Bronfenbrener, L. Grinis, E. Korin, Experimental study of heat transfer
intensification under vibration condition, Chem. Eng. Technol. 24 (4) (2004)
367–371.
[8] C. Gau, S.X. Wu, H.S. Su, Synchronization of vortex shedding and heat transfer
enhancement over a heated cylinder oscillating with small amplitude in
streamwise direction, J. Heat. Transf., Trans. ASME 123 (2001) 1139–1148.
[9] S.-J. Yang, W.-S. Fu, Numerical investigation of heat transfer from a heated
oscillating rectangular cylinder in a cross flow, Numer. Heat Transf. A-Appl. 39
(2001) 569–591.
[10] W.-S. Fu, B.-H. Tong, Numerical investigation of heat transfer from a heated
oscillating cylinder in a cross flow, Int. J. Heat Mass Transf. 45 (14) (2002)
3033–3043.
[11] T. Açikalin, S.V. Garimella, A. Raman, J. Petroski, Characterization and
optimization of the thermal performance of miniature piezoelectric fans, Int.
J. Heat Fluid Flow 28 (4) (2007) 3033–3043.
[12] J. Ghazanfarian, M.R.H. Nobari, A numerical study of convection heat transfer
from a rotating cylinder with cross-flow oscillation, Int. J. Heat Mass Transf. 52
(2009) 5402–5411.
[13] M.R.H. Nobari, J. Ghazanfarian, Convection heat transfer from a rotating
cylinder with inline oscillation, Int. J. Therm Sci. 49 (2010) 2026–2036.
[14] L. Léal, F. Topin, P. Lavieille, L. Tadrist, M. Miscevic, Simultaneous integration,
control and enhancement of both fluid flow and heat transfer in small scale
heat exchangers: a numerical study, Int. Commun. Heat Mass Transf. 49 (2013)
36–40.
[15] J.S. Go, S.J. Kim, G. Lim, et al., Heat transfer enhancement using flow-induced
vibration of a microfin array, Sens. Actuat. A 90 (3) (2001) 232–239.
[16] J.S. Go, Design of a microfin array heat sink using fiow-induced vibration to
enhance the heat transfer in the laminar flow regime, Sens. Actuat. A 105 (2)
(2003) 201–239.
[17] L. Cheng, T. Luan, W. Du, M. Xu, Heat transfer enhancement by flow-induced
vibration in heat exchangers, Int. J. Heat Fluid Flow 52 (3–4) (2009) 1053–
1057.
334 X. Sun et al. / International Journal of Heat and Mass Transfer 128 (2019) 319–334
[18] K. Yan, P.Q. Ge, W.B. Bi, Y.C. Su, R.R. Hu, Vibration characteristics of fluid-
structure interaction of conical spiral tube bundle, J. Hydrodyn. 22 (1) (2010)
121–128.
[19] M.R. Ge, K. Yan, P.Q. Ge, J. Gao, Study on the conical spiral tube bundle heat
exchangers with pulsation flow generator structure, Appl. Mech. Mater. 143–
144 (2012) 698–702.
[20] D.R. Duan, P.Q. Ge, W.B. Bi, Numerical investigation on heat transfer
performance of planar elastic tube bundle by flow-induced vibration in heat
exchanger, Int. J. Heat Mass Transf. 103 (2016) 868–878.
[21] Y.C. Su, M.L. Li, M.L. Liu, G.D. Ma, A study of the enhanced heat transfer of flow-
induced vibration of a new type of heat transfer tube bundle-the planar
bending elastic tube bundle, Nucl. Eng. Des. 309 (2016) 294–302.
[22] K.H. Park, J.K. Min, J.K. Kim, S.H. Park, M.Y. Ha, A study on a flexible wing with
up-down vibration in a pulsating flow of cooling air to improve heat transfer
efficiency, Heat Mass Transf. 49 (10) (2013) 1459–1470.
[23] J.X. Shi, J.W. Hu, S.R. Schafer, S.R. Schafer, C.L. Chen, Numerical study of heat
transfer enhancement of channel via vortex-induced vibration, Appl. Therm.
Eng. 70 (2014) 838–845.
[24] A.K. Soti, R. Bhardwaj, J. Sheridan, Flow-induced deformation of a flexible thin
structure as manifestation of heat transfer enhancement, Int. J. Heat Mass
Transf. 84 (2015) 1070–1081.
[25] R.U. Joshi, A.K. Soti, R. Bhardwaj, Numerical study of heat transfer
enhancement by deformable twin plates in laminar heated channel flow,
Comput. Therm. Sci. 7 (2015) 467–476.
[26] S. Ali, C. Habchi, S. Menanteau, T. Lemenand, J.L. Harion, Heat transfer and
mixing enhancement by free elastic flaps oscillation, Int. J. Heat Mass Transf.
85 (2015) 250–264.
[27] S. Ali, S. Menanteau, C. Habchi, T. Lemenand, J.L. Harion, Heat transfer and
mixing enhancement by using multiple freely oscillating flexible vortex
generators, Appl. Therm. Eng. 105 (2016) 276–289.
[28] E. Jamesahar, M. Ghalambaz, A.J. Chamkha, Fluid–solid interaction in natural
convection heat transfer in a square cavity with a perfectly thermal-
conductive flexible diagonal partition, Int. J. Heat Mass Transf. 100 (2016)
303–319.
[29] S.A.M. Mehryan, A.J. Chamkha, M.A. Ismael, M. Ghalambaz, Fluid–structure
interaction analysis of free convection in an inclined square cavity partitioned
by a flexible impermeable membrane with sinusoidal temperature heating,
Meccanica 52 (2017) 2685–2703.
[30] S.A.M. Mehryan, M. Ghalambaz, M.A. Ismael, A.J. Chamkha, Analysis of fluid-
solid interaction in MHD natural convection in a square cavity equally
partitioned by a vertical flexible membrane, J. Magn. Magn. Mater. 424 (2017)
161–173.
[31] M. Ghalambaz, E. Jamesahar, M.A. Ismael, A.J. Chamkha, Fluid-structure
interaction study of natural convection heat transfer over a flexible
oscillating fin in a square cavity, Int. J. Therm Sci. 111 (2017) 256–273.
[32] W.T. Thomson, M.D. Dahleh, Theory of Vibration with Applications, fifth ed.,
Prentice-Hall Inc., New Jersey, 1997.
[33] X. Sun, J.Z. Zhang, X.L. Ren, Characteristic-based split (CBS) finite element
method for incompressible viscous flow with moving boundaries, Eng. Appl.
Comp. Fluid 6 (3) (2012) 461–474.
[34] J. Jameson, Time dependent calculations using multigrid with application to
unsteady flows past airfoils and wings, AIAA paper (1991) 91–1596.
[35] O.C. Zienkiewicz, R.L. Taylor, P. Nithiarasu, The Finite Element Method for Fluid
Dynamics, sixth ed., Butterworth-Heinemann, Elsevier, 2005.
[36] X. Sun, J. Z. Zhang, Finite-element analysis of nonlinear fluid-membrane
interactions using a modified characteristic-based split (CBS) scheme, in: V.
Afraimovich, J.A. Tenreiro Machado, J.Z. Zhang, Complex Motions and Chaos,
Springer, 2016, pp. 75–105 (Chap. 3).
[37] X. Sun, J.Z. Zhang, Effect of the reinforced leading or trailing edge on the
aerodynamic performance of a perimeter-reinforced membrane wing, J. Fluid
Struct. 68 (2017) 90–112.
[38] X. Sun, X.L. Ren, J.Z. Zhang, Nonlinear dynamic responses of a perimeter-
reinforced membrane wing in laminar flows, Nonlinear Dyn. 88 (1) (2017)
749–776.
[39] F.J. Blom, Considerations on the spring analogy, Int. J. Numer. Meth. Fl. 32 (6)
(2000) 647–668.
[40] S.W. Churchill, M.J. Bernstein, A correlating equation for forced convection
from gases and liquids to a circular cylinder in crossflow, J. Heat Transf. 99
(1977) 300–306.
[41] K. Momose, H. Kimoto, Forced convection heat transfer from a heated circular
cylinder with arbitrary surface temperature distributions, Heat Transf.-Asian
Res. 28 (6) (1999) 484–499.
[42] N. Zhang, Z.C. Zhang, S. Eckels, Study of heat-transfer on the surface of a
circular cylinder in flow using an immersed-boundary method, Int. J. Heat
Fluid Flow 29 (2008) 1558–1566.
[43] R.P. Bharti, R.P. Chhabra, A numerical study of the steady forced convectional
heat transfer from an unconfined circular cylinder, Heat Mass Transf. 43
(2007) 639–648.
[44] F. Baratch M. Saghafian B. Baratchi Numerical investigation on lock-on
condition and convection heat transfer from an elastically supported
cylinder in a cross flow J. Fluid Eng.-T. ASME 135 2013 031103.
[45] W.A. Wall, E. Ramm, Fluid–structure interaction based upon a stabilized (ALE)
finite element method, In: S.R. Idelsohn, E. Onate, E.N. Dvorkin (Eds.),
Proceedings of the 4th world congress on computational mechanics: new
trends and applications. CIMNE, Barcelona, 1998, pp. 1–20.
[46] P.R.F. Teixeira, A.M. Awruch, Numerical simulation of fluid–structure
interaction using the finite element method, Comput. Fluid 34 (2) (2005)
249–273.
[47] W. Dettmer, D. Peric, A computational framework for fluid–structure
interaction: finite element formulation and applications, Comput. Methods
Appl. Mech. Eng. 195 (41) (2006) 5754–5779.
[48] Y. Bazilevs, V.M. Calo, T.J.R. Hughes, Y. Zhang, Isogeometric fluid–structure
interaction: theory, algorithms, and computations, Comput. Mech. 43 (1)
(2008) 3–37.
[49] T. He, K. Zhang, An overview of the combined interface boundary condition
method for fluid–structure interaction, Arch. Comput. Methods Eng. 24 (4)
(2017) 891–934.
[50] A.K. Abu-Hijleh, Fin placement for optimal forced convection heat transfer
from a cylinder in cross-flow, Int. J. Numer. Method H. 15 (3) (2005) 277–295.