International Journal of Heat and Mass Transfer 77 (2014) 501–518

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer

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

Natural convection in a square enclosure with two inner circular

cylinders positioned at different vertical locations
Yong Gap Park 1, Man Yeong Ha ⇑, Changyoung Choi 1, Jaehyun Park 1
School of Mechanical Engineering, Pusan National University, Jang Jeon 2-Dong, Geum Jeong Gu, Busan 609-735, Republic of Korea

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

Article history: The present study investigates the natural convection induced by a temperature difference between a
Received 14 March 2014 cold outer square enclosure and two hot inner circular cylinders. A two-dimensional solution for natural
Received in revised form 20 May 2014 convection in an enclosure with inner cylinders is obtained using an accurate and efficient immersed
Accepted 20 May 2014
boundary method. The immersed boundary method based on the finite volume method is used to handle
inner cylinders located at different vertical centerline positions of the enclosure for different Rayleigh
numbers in the range 103 6 Ra 6 106 . The results in the case of two cylinders are compared with those
Keywords:
in the case of the single cylinder in order to see the effect of the interaction between the adjacent two
Natural convection
Numerical study
inner hot cylinders in addition to the interaction between two hot inner cylinders and the cold walls
Two inner cylinders of the enclosure. The distribution of isotherms and streamlines eventually reaches a steady state or
Vertical location changes its state from steady to unsteady, depending on the values of the Rayleigh number and the cyl-
Square enclosure inder position in the enclosure. The distribution of the local and surface-averaged Nusselt numbers is
obtained for different Rayleigh numbers and the cylinder positions.
Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction Nusselt number was investigated. As a result, the average heat

The heat transfer and flow characteristics of natural convection
in an enclosure have many industrial and environmental applica-
tions, such as heat exchangers, solar collectors, nuclear safety sys-
tems, chemical reactors, electronic equipment cooling, and
stratified atmosphere boundary layers. Many engineers in indus-
trial settings wish to avoid the use of active control equipment
(e.g., fans) for cooling or heating because of their additional cost,
noise, and vibration problems. Therefore, it is important to fully
understand the mechanism of natural convection. In this study,
we investigate the effect of two cylinders at different vertical loca-
tions in the enclosure on the heat transfer and fluid flow.
For several decades, many investigations have dealt with natu-
ral convection from a heated body placed concentrically [1–6] or
eccentrically [7–11] inside a cooled enclosure.
Cesini et al. [1] performed a numerical and experimental analy-
sis of natural convection from a horizontal cylinder enclosed in a
rectangular cavity. The influence of the cavity aspect ratio and
the Rayleigh number on the distribution of temperature and
⇑ Corresponding author. Tel.: +82 51 510 2440; fax: +82 51 515 3101.
E-mail addresses: pyg777@pusan.ac.kr (Y.G. Park), myha@pusan.ac.kr (M.Y. Ha),
ccy21c@gmail.com (C. Choi), jhpark@naver.com (J. Park).
1
Fax: +82 51 515 3101.
transfer coefficients increase with the increasing Rayleigh number.
Asan [2] numerically studied two-dimensional natural convec-
tion in an annulus between two isothermal concentric square
ducts and obtained solutions up to a Rayleigh number (Ra) of
106. The results showed that the dimension ratio and Rayleigh
number have a profound influence on the temperature and flow
field.
Moukalled and Acharya [3] and Shu and Zhu [4] studied the
change of the thermo-flow field between a low-temperature outer
square enclosure and a high-temperature inner circular cylinder
according to the radius of the inner circular cylinder. Moukalled
and Acharya [3] considered three different aspect ratios, r/L, of
the cylinder radius, r, to the enclosure height, L, in the range of
104 6 Ra 6 107. They showed that, at a constant enclosure aspect
ratio, the total heat transfer increases with increasing Rayleigh
number. When the Rayleigh number is constant, the convection
contribution to the total heat transfer decreases with the increas-
ing value of the aspect ratio. Shu and Zhu [4] obtained numerical
results for Rayleigh numbers ranging from 104 to 106 and aspect
ratios between 1.67 and 5.0. It was found that both the aspect ratio
and the Rayleigh number are critical to the patterns of the flow and
thermal fields. They also suggested that a critical aspect ratio may
exist at high Rayleigh numbers to distinguish between the flow
and thermal patterns.

http://dx.doi.org/10.1016/j.ijheatmasstransfer.2014.05.041
0017-9310/Ó 2014 Elsevier Ltd. All rights reserved.
502 Y.G. Park et al. / International Journal of Heat and Mass Transfer 77 (2014) 501–518
Nomenclature
fi momentum forcing
g acceleration of gravity [m/s2]
L length of square enclosure [m]
n normal direction to the wall
Nucyl local Nusselt number along the inner single circular
cylinder
Nucyl surface-averaged Nusselt number along the inner single
circular cylinder
Nuupper cyl local Nusselt number along the upper inner circular
cylinder
Nulower cyl local Nusselt number along the lower inner circular
cylinder
Nuen local Nusselt number along the walls of the enclosure
Nuupper cyl surface-averaged Nusselt number along the upper inner
circular cylinder
Nulower cyl surface-averaged Nusselt number along the lower inner
circular cylinder
P pressure [Pa]   2
P dimensionless pressure ¼ PqaL2
Pr Prandtl number (= m/a)
r dimensionless radius of the cylinder (= R/L)
R radius of circular cylinder [m]
 
Ra
3
Rayleigh number ¼ gbL ðTmah T c Þ
t time [s]   
t dimensionless time ¼ tL2a
T dimensional temperature [K]
Kim et al. [7] and Yoon et al. [8] conducted numerical investiga-
tions on the natural convection induced by the temperature differ-
ence between a cold outer square enclosure and a hot inner
circular cylinder for different Rayleigh numbers varying over the
range 103 6 Ra 6 107. The location of the inner circular cylinder
(d) was changed vertically along the centerline of the square enclo-
sure. The number, size, and formation of the cells strongly
depended on the Rayleigh number and position of the inner circu-
lar cylinder within the enclosure. At Ra = 107, the bifurcation of
natural convection from the unsteady to the steady state depended
on d. They showed that the flow and heat transfer at Ra = 107
became unsteady at d 6 dC,L and d P dC,U, where dC,L = 0.05 and
dC,U = 0.18.
However, natural convection in the presence of an array of cyl-
inders inside an enclosure is quite different from that in the pres-
ence of a single cylinder in the enclosure. This is because of the
mutual interaction among the buoyant plumes generated by an
array of cylinders, in addition to the interaction between the array
and the enclosure. Several researchers have considered the effect
of the presence of an array of cylinders on natural convection
numerically [12–17] or experimentally [18–20].
Lacroix [12] numerically studied the natural convection heat
transfer from two vertically separated heated cylinders to a rectan-
gular cavity cooled from above. Two cavity widths and three top
cylinder positions were investigated over Rayleigh numbers in
the range 104 6 Ra 6 106 . The results showed the complexity of
convective motion in different geometric configurations and the
effects of the cylinder positions and Rayleigh number on the local
and overall heat transfer rates.
Lacroix and Joyeux [13] conducted a numerical study of natural
convection heat transfer from two horizontal heated cylinders con-
fined to a rectangular enclosure having conducting vertical walls of
finite thicknesses and horizontal walls at the heat sink temperature.
They focused on investigating the interaction between convection
Th hot temperature [K]
Tc cold temperature [K]
ui velocity [m/s]  
u L
ui dimensionless velocity ¼ ai
xi Cartesian coordinates [m]  
x
xi dimensionless Cartesian coordinates ¼ Li
yc distance between the square enclosure center to the cir-
cular cylinder center [m]
yc dimensionless distance between the square enclosure
center to the circular cylinder center [m]
Greek symbols
a thermal diffusivity [m2/s]
b thermal expansion coefficient [K1]
d⁄ vertical distance between the centers of two circular
cylinders [m]
d dimensionless vertical distance between the centers of
two circular cylinders
q density [kg/m3]
m kinematic viscosity [m2/s]
u angle from the top of the circular
 cylinder

h dimensionless temperature ¼ TTT c
h T c
Subscripts/superscripts

dimensional value

surface-averaged quantity
in the fluid-filled cavity and conduction in the vertical walls, indi-
cating that heat transfer is strongly influenced by the coupling
effect between solid wall conduction and fluid convection.
Chae and Chung [18] conducted an experimental investigation
of the natural convection heat transfer for two parallel horizontal
cylinders. They considered various pitch-to-diameter ratios (P/D)
from 1.02 to 9, Prandtl numbers from 2014 to 8334, and Rayleigh
numbers varying over the range 7:3  107 6 Ra 6 4:5  1010 . They
measured the mass transfer rate from the cylinders and obtained
the heat transfer rate (Nusselt number) on the basis of the analogy
concept. Their results showed that the Nusselt number ratios of the
upper to lower cylinders, which increased with P/D, were less than
1 at P/D values of less than about 1.5 for laminar flows, and were
almost 1 at P/D values very close to 1 for turbulent flows. The Nus-
selt number ratios of the upper to lower cylinders also depended
on the Prandtl number, showing a steep variation with P/D at
higher Prandtl numbers.
Although many researchers have studied on the natural convec-
tion, there is little information about natural convection processes
within a cooled square enclosure containing two hot circular cylin-
ders located at different vertical positions along the center of the
enclosure. In this situation, the flow and heat transfer characteristics
are largely affected by the location of two cylinders and the buoy-
ancy-induced convection at different Rayleigh numbers. The aim
of this study is to examine the effects of the locations of two hot
inner circular cylinders in the enclosure and the buoyancy-induced
convection on heat transfer and fluid flow in the enclosure when two
circular cylinders are located at different vertical positions along the
center of the enclosure at different Rayleigh numbers. The results in
the case of two cylinders are compared with those in the case of the
single cylinder in order to investigate the effect of the interaction
between two adjacent hot cylinders on the fluid flow and heat trans-
fer in the enclosure, in addition to the interaction between the two
hot cylinders and the cold enclosure walls.
 Y.G. Park et al. / International Journal of Heat and Mass Transfer 77 (2014) 501–518 503

Fig. 1. Computational domain and coordinate system along with boundary conditions.

2. Numerical methodology corresponding dimensionless velocity components, t is the dimen-

sionless time, p is the dimensionless pressure, and h is the dimen-
Fig. 1 shows the computational domain and coordinate system sionless temperature.
along with the boundary conditions considered in the present The above non-dimensionalization results in two dimensionless
3
study. The system consisted of a square enclosure with side lengths parameters: Pr ¼ m=a and Ra ¼ gbL ðTmah T c Þ, where m, g, and b are the
L, within which one and two circular cylinders of radius R (= 0.1L) kinematic viscosity, gravitational acceleration, and volume expan-
were located and moved along the vertical centerline in the range sion coefficient, respectively. In the simulations reported here, Pr
of xc ¼ 0 and 0:35L 6 yc 6 0:35L, where xc and yc represent the x and r (= R/L) were taken to be 0.7 (corresponding to that of air)
and y coordinates of the center of the circular cylinder center, and 0.1, respectively. The Rayleigh number was varied in the range
respectively. The walls of the square enclosure were maintained of 103–106. The dimensionless position of the one and two cylin-
at a constant low temperature of Tc, whereas the cylinder was ders varied in the range of xc ¼ 0 and 0:35 6 yc 6 0:35. For the
maintained at a constant high temperature of Th. The dimension- velocity field, no-slip and no-penetration boundary conditions
less vertical distance d, which represents the distance between were imposed on the walls. For the thermal boundary conditions,
the centers of the two inner cylinders along the vertical centerline, h = 0 was imposed at the wall of the cold enclosure, whereas
varied in the range of 0.3 6 d 6 0.7. h = 1 was used for the wall of the hot inner cylinder.
In this study, we assumed that radiation effects were negligible. The mass source/sink q in Eq. (1) and momentum forcing fi in
The fluid properties were assumed to be constant except for the Eq. (2) were applied on the body surface or inside the body to sat-
density in the buoyancy term, which follows the Boussinesq isfy the no-slip condition and mass conservation in the cell con-
approximation. The gravitational acceleration acted in the negative taining the immersed boundary. In Eq. (3), the heat source/sink h
y-direction. The immersed boundary method (IBM) was used to was applied to satisfy the isothermal boundary condition at the
handle the one and two cylinders, which were located at the cen- immersed boundary.
ters of the enclosure. The governing equations describing unsteady A two-step time-split scheme was used to advance the flow
incompressible viscous flow and thermal fields used in the present field. This scheme was based on the previous work of Kim and
study were the continuity, momentum, and energy equations in Moin [21] and Zang et al. [22]. The temperature field was advanced
their non-dimensional forms defined as in a similar manner with a second-order Adams–Bashforth scheme
@ui for the advection term and a Crank–Nicolson scheme for the diffu-
q¼0 ð1Þ sion term. The central difference scheme with second-order accu-
@xi
racy based on the finite volume method was used for spatial
discretization. Additionally, a second-order linear or bilinear inter-
@ui @ui @P @ 2 ui polation scheme was applied to satisfy the no-slip and isothermal
þ uj ¼ þ Pr þ RaPrhdi2 þ fi ð2Þ
@t @xj @xi @xj @xj conditions on the immersed boundary. Further details of the IBM
are given in the papers of Kim et al. [7], Yoon et al. [8], Kim et al.
@h @h @2h [23] and Kim and Choi [24].
þ uj ¼ þh ð3Þ
@t @xj @xj @xj Once the velocity and temperature fields were obtained, the
local and surface-averaged Nusselt numbers were defined as
The dimensionless variables in the above equations were
defined as  Z
@h  1 W
Nu ¼ ; Nu ¼ Nu dS ð5Þ
t a x ui L P  L2 T  Tc @n wall W 0
t¼ ; xi ¼ i ; ui ¼ ; P¼ ; h¼ ð4Þ
L2 L a qa2 Th  Tc
where n is the normal direction with respect to the walls and W is
In the above equations, q, T, and a represent the density, dimen- the surface area of the walls.
sional temperature, and thermal diffusivity, respectively. The Fig. 2 shows the computational geometry in the x–y plane with
superscript ⁄ in Eq. (4) represents the dimensional variables. xi a uniform grid distribution. A grid resolution of 501  501 along
indicates dimensionless Cartesian coordinates, ui indicates the the horizontal (x) and vertical (y) directions was employed in our
504 Y.G. Park et al. / International Journal of Heat and Mass Transfer 77 (2014) 501–518

Fig. 2. A typical grid distribution for xc = 0 and yc = 0.

computations. The grid point was uniformly distributed in the Table 2

computational domain. Grid independence of the solution has been Comparison of surface-averaged Nusselt numbers from present study with those of a
tested with additional simulations on much finer grids with up to previous numerical and experimental studies.
601  601 points as shown in Table 1. The differences in the results Ra Mean Nusselt number at hot wall
of the Nusselt numbers obtained using the coarse and fine grids are
Present Kim et al. [7] Moukalled and Warrington
less than 1%. The time-dependency test has been also conducted study Acharya [3] and Powe [20]
with respect to the various time step as shown in Table 1. Based
103 5.107 5.093 – –
on the results of the time-dependency test, the time interval to sat- 104 5.128 5.108 5.49 5.286
isfy the CFL number less than 0.2 was selected for the temporal 105 7.836 7.767 8.377 8.670
integration in the time-marching procedure. 106 14.462 14.11 15.4 14.229
To validate the code, a natural convection problem in a low-
temperature outer square enclosure with a high-temperature inner
circular cylinder located at the center was tested. For comparison
flow and thermal fields in the case of two cylinders become
purposes, we considered the computational and experimental
unsteady and asymmetric about the vertical centerline at x = 0,
studies reported by Moukalled and Acharya [3] and Warrington
when d = 0.5 at Ra = 106, because of the increasing effect of convec-
and Powe [20], in addition to the computational study reported
tion. This is different from previous results with a single cylinder,
by Kim et al. [7]. We compared the surface-averaged Nusselt num-
because of the interaction between the two adjacent inner hot cyl-
bers obtained from our computational results with those from the
inders in addition to the interaction between the inner hot cylin-
previous works, as shown in Table 2, and found good agreement
ders and the enclosure walls.
between them.

3. Results and discussion 3.1. Ra = 103

When the single cylinder moves vertically in the enclosure, the Fig. 3 shows the distribution of isotherms and streamlines for
flow and thermal fields eventually reach a steady state regardless different yc and d values when an inner single cylinder and two
of cylinder position and Rayleigh number considered in this study. inner cylinders move downward and upward from yc = 0, respec-
However, when two cylinders move vertically in the enclosure, the tively, at Ra = 103. The figures in the left and right columns, as
natural convection in this system eventually reaches a steady state shown in Fig. 3(a)–(k), represent the distribution of isotherms
or changes its state from steady to unsteady, depending on the val- and streamlines, respectively.
ues of the Rayleigh number and the cylinder position. The flow and In the case of the single cylinder, the distributions of isotherms
thermal fields in the case of two cylinders also reach a steady state and streamlines in the enclosure for different yc values show sym-
eventually and are symmetric about the vertical centerline at x = 0 metric shapes with respect to the vertical centerline at x = 0, as
regardless of d when 103 6 Ra 6 105. However, the states of the shown in Fig. 3(a)–(f). However, in the case of the single cylinder,
the distribution of isotherms and streamlines at y = yc in upper part
of the enclosure is not symmetric with respect to the horizontal
Table 1
centerline at y = 0, as compared to that at y = yc in lower part of
Results of grid- and time-dependency tests through the surface average Nusselt the enclosure, as a result of the presence of the very weak convec-
number of the upper cylinder when Ra = 106 and d ¼ 0:3 . tive flow in the enclosure. When the single cylinder moves down-
Grid Nusselt number Percent difference
ward, the gradient of the isotherms becomes larger at the lower
part of the enclosure because of the decreasing distance between
Grid-dependency test
401  401 21.7 –
the lower part of the inner single cylinder and the bottom wall,
501  501 22.1 1.68 whereas they becomes smaller in the opposite region. When the
601  601 22.2 0.697 single cylinder moves upward, the variations in the patterns of
CFL number Time step Nusselt number the isotherms are opposite to those when yc < 0.
Time-dependency test In the case of two cylinders, the distributions of isotherms and
0.1 0.0000005 22.1 streamlines for all d values show symmetric shapes with respect to
0.2 0.000001 22.1
the vertical and horizontal centerlines at x = 0 and y = 0, respec-
0.3 0.0000015 22.1
tively, because of the dominance of conduction, unlike to the
 Y.G. Park et al. / International Journal of Heat and Mass Transfer 77 (2014) 501–518
(a) yc = −0.15
(b) yc = −0.25
(c) yc = −0.35
(d) yc = 0.15
(e) yc = 0.25
(f) yc = 0.35
Fig. 3. Isotherms and streamlines for different yc and d values in the case of the single cylinder (left column) and in the case of two cylinders (right column) at Ra = 103
(isotherm contour values range from 0 to 1 with 10 levels).
symmetric shape only with respect to the vertical centerline at
x = 0 in the case of the single cylinder. As the vertical distance
between the centers of two inner cylinders (d) increases, the dis-
tances between the top surface of the upper cylinder and the top
wall of the enclosure and between the bottom surface of the lower
cylinder and the bottom wall of the enclosure decrease. This causes
the gradients of the isotherms in these regions to increase, as
shown in Fig. 3(g)–(k). The thermal boundary layers between the
two cylinders merge at d ¼ 0:3 and d = 0.4, because the upper
and lower hot cylinders are in close proximity. The circulation of
the flow shows two overall rotating symmetric eddies.
505
(g) δ = 0.3( yc = ±0.15)
(h) δ = 0.4( yc = ±0.20)
(i) δ = 0.5( yc = ±0.25)
(j) δ = 0.6( yc = ±0.30)
(k) δ = 0.7 ( yc = ±0.35)
Fig. 4 shows the distribution of local Nusselt numbers along the
surface of the inner cylinders and the walls of the enclosure for dif-
ferent yc and d values when an inner single cylinder and two inner
cylinders move downward and upward from yc = 0, respectively, at
Ra = 103. Here, Nucyl represents the local Nusselt number along the
cylinder surface in the case of the single cylinder. Nuupper cyl and
Nulower cyl represent the local Nusselt number along the upper and
lower cylinder surfaces in the case of two cylinders. The variation
of local Nusselt number in both cases of the single cylinder and two
cylinders at Ra = 103 depends on the distributions of isotherms
shown in Fig. 3. Because the dominant heat transfer mode is
506 Y.G. Park et al. / International Journal of Heat and Mass Transfer 77 (2014) 501–518

Fig. 4. Distribution of local Nusselt numbers along the surface of the inner cylinders (Nucyl, Nuupper cyl and Nulower cyl ) and the walls of the enclosure (Nuen) for different yc and d
values in the case of the single cylinder (left column) and in the case of two cylinders (right column) at Ra = 103.

conduction under the very weak convection at Ra = 103, the effect
of the gap distance between the inner cylinders and the walls of
the enclosure is the main factor in determining the distributions
of Nucyl, Nuupper cyl , Nulower cyl and Nuen.
When an inner single cylinder moves upward with yc P 0 at
Ra = 103, the variation in the values of Nucyl around u ¼ 0 (or
u ¼ 360 ) according to yc is very large because the effect of the
gap distance between the inner cylinder and the top wall of the
enclosure on Nucyl becomes larger with increasing yc. However,
when an inner single cylinder moves downward at Ra = 103, the
variation in the values of Nucyl around u ¼ 0 (or u ¼ 360 ) accord-
ing to yc is small, because the relatively large space for heat trans-
fer is secured between the inner cylinder and the top wall of the
enclosure. As a result, the effect of the gap distance between the
inner cylinder and the top wall of the enclosure on Nucyl becomes
small, as shown in Fig. 4(a). The variation in the values of Nucyl
 Y.G. Park et al. / International Journal of Heat and Mass Transfer 77 (2014) 501–518
(a) yc = −0.15
(b) yc = −0.25
(c) yc = −0.35
(d) yc = 0.15
(e) yc = 0.25
(f) yc = 0.35
Fig. 5. Isotherms and streamlines for different yc and d values in the case of the single cylinder (left column) and in the case of two cylinders (right column) at Ra = 104
(isotherm contour values range from 0 to 1 with 10 levels).
around u ¼ 180 according to yc for the case of the single cylinder
at Ra = 103 is opposite to that around u ¼ 0 (or u ¼ 360 ), result-
ing in small variations when yc P 0 and large variations when
yc 6 0, as shown in Fig. 4(a).
When an inner single cylinder moves vertically in the enclosure
at Ra = 103, the values of Nuen on the walls of the enclosure are gen-
erally maximized when the distance between the inner single cyl-
inder and the enclosure walls is at its smallest at Ra = 103, as
shown in Fig. 4(b). As we move from point A to point B along the
top wall of the enclosure in the case of the single cylinder, the
value of Nuen increases with increasing yc because the space
507
(g) δ = 0.3( yc = ±0.15)
(h) δ = 0.4( yc = ±0.20)
(i) δ = 0.5( yc = ±0.25)
(j) δ = 0.6( yc = ±0.30)
(k) δ = 0.7 ( yc = ±0.35)
between the inner cylinder and the top wall decreases. When we
move from point B to point C along the right wall of the enclosure,
the location at which Nuen reaches the highest peak value moves
from point C to point B as the inner single cylinder moves upward
with increasing yc. The pattern for the variation in Nuen along the
bottom wall of the enclosure (corresponding to the region from
point C to point D) according to yc in the case of the single cylinder
is similar to that along the top wall of the enclosure (corresponding
to the region from point A to point B). However, a careful observa-
tion indicates that the peak value of Nuen along the top wall of the
enclosure is slightly larger than that along the bottom wall of the
508 Y.G. Park et al. / International Journal of Heat and Mass Transfer 77 (2014) 501–518
enclosure due to the presence of the weak convection flow. The
pattern for the variation in Nuen along the left and right walls of
the enclosure (corresponding to the regions from point D to point
E and from point B to point C) according to yc in the case of the sin-
gle cylinder is symmetric about the vertical centerline at x = 0.
For all values of d when two cylinders move vertically at
Ra = 103, the values of Nuupper cyl and Nulower cyl have minimum values
around u ¼ 108 and u = 0° (or 360 ), respectively, because the
upper and lower hot cylinders have the same temperature, as
shown in Fig. 4(c) and (d). In addition, the values of Nuupper cyl and
Nulower cyl have maximum values around u ¼ 0 or ð360 Þ and
u ¼ 108 , respectively, because the distance between the upper
hot cylinder and the top cold wall of the enclosure and the distance
between the lower hot cylinder and the bottom cold wall of the
enclosure are small under the dominant conduction heat transfer
mode. As d increases, the values of Nuupper cyl and Nulower cyl increase,
because of the increasing gap distance effect as a result of the
decreasing distance between the two hot inner cylinders and the
top and bottom walls of the enclosure.
The distribution of Nuen in the case of two cylinders also
depends strongly on the positions of the cylinders in the enclosure
in the conduction-dominant regime of Ra = 103, as shown in
Fig. 4(e). As we move from point A to point B along the top wall
of the enclosure in the case of two cylinders, the value of Nuen
increases with increasing d because the space between the upper
inner cylinder and the top wall decreases. When we move from
point B to point C along the right wall of the enclosure for the case
of two cylinders, the value of Nuen is almost the same for different d
values because of the same distance between the two hot inner
cylinders and the right wall of the enclosure, unlike to the case
of the single cylinder. The patterns of variation in Nuen along the
bottom wall of the enclosure according to d for the case of two cyl-
inders are similar to those along the top wall of the enclosure due
to the dominance of conduction. The pattern for the variation in
Nuen along the left and right walls of the enclosure according to
yc in the case of two cylinders is symmetric about the vertical cen-
terline at x = 0.
3.2. Ra = 104
Fig. 5 shows the distribution of isotherms and streamlines for
different yc and d values when an inner single cylinder and two
inner cylinders move downward and upward from yc = 0, respec-
tively, at Ra = 104. As Ra increased to 104, the effect of convection
on the fluid flow and heat transfer becomes slightly larger than
that at Ra = 103. As a result, for all yc and d values, the vortex cores
of the two rotating eddies in both cases of the single cylinder and
two cylinders at Ra = 104 are located at a slightly higher position
than those at Ra = 103.
In the case that an inner single cylinder moves vertically, the
distributions of isotherms according to yc at Ra = 104 are different
from those for Ra = 103 because of the effect of convection, espe-
cially when the single cylinder moves downward with yc < 0. As
yc decreases according to the downward movement of the single
cylinder, the isotherms are gradually distorted and the size of the
thermal plumes formed on the top of the inner cylinder becomes
larger.
In the case that two inner cylinders move vertically at Ra = 104,
the isotherms and streamlines are not symmetric about the hori-
zontal centerline at y = 0, unlike their symmetric shape at
Ra = 103, as shown in Figs. 3(g)–(k) and 5(g)–(k), because the
effect of convection on the flow and heat transfer in the enclosure
increases slightly with the increasing Rayleigh number. The distri-
bution of isotherms and streamlines in the case of two cylinders
at Ra = 104 is also different from that in the case of the single
cylinder, as shown in Fig. 3(a)–(f) and (g)–(k), because of the
interaction of the fluid flow and heat transfer between two adja-
cent inner cylinders, as compared to the fluid flow and heat trans-
fer around the inner single cylinder. In the range of 0.3 6 d 6 0.6
in the case of two inner cylinders, the isotherms near the lower
part of the lower inner cylinder are denser than those near the
upper inner cylinder, owing to the increase in the convective flow.
However, at d = 0.7, the isotherms near the lower part of the lower
inner cylinder are almost the same as those near the upper inner
cylinder because the distances between the two inner cylinders
and the top and bottom walls of the enclosure are small. As d
increases, less space is secured between the two inner cylinders
and the top and bottom walls of the enclosure, reducing the
buoyancy-induced convection. However, more spaces between
the two inner cylinders are secured, enhancing the effect of
convection.
Fig. 6 shows the distribution of local Nusselt numbers along the
surface of the inner cylinders and the walls of the enclosure for dif-
ferent yc and d values when an inner single cylinder and two inner
cylinders move downward and upward from yc = 0, respectively, at
Ra = 104. When Ra = 104, the effect of the convection heat transfer
is still weak, and, as a result, the gap distance between the inner
cylinder and the walls of the enclosure is the main factor in deter-
mining the distributions of Nucyl, Nuupper cyl , Nulower cyl and Nuen in
both cases of the single cylinder and two cylinders. Therefore,
the variation pattern in the distribution of Nucyl, Nuupper cyl ,
Nulower cyl , and Nuen according to yc and d when Ra = 104 in both
cases of the single cylinder and two cylinders, as shown in
Fig. 6(a)–(e), are generally similar to those when Ra = 103, as
shown in Fig. 4(a)–(e).
However, when the single cylinder moves vertically with yc 6 0,
the values of Nucyl around u ¼ 108 and the peak value of Nuen
along the top wall of the enclosure for Ra = 104 are larger than
those for Ra = 103 because of the presence of the ascending plume,
as shown in Fig. 5(a)–(c). In the case of two cylinders, the values of
Nuupper cyl and Nulower cyl at around u ¼ 108 for Ra = 104 are also
slightly larger than those for Ra = 103 because of the slightly
increasing effect of convection with increasing Rayleigh number.
In the case of two cylinders at Ra = 104, the value of Nuen on the
top wall of the enclosure is slightly larger than that on the bottom
wall of the enclosure. In addition, for the case of two cylinders at
Ra = 104, the location at which Nuen on the right wall of the enclo-
sure has the highest peak value is around point B, due to the
slightly increasing effect of convection. The pattern for the varia-
tion in Nuen along the left and right walls of the enclosure accord-
ing to yc and d in both cases of the single cylinder and two cylinders
is symmetric about the vertical centerline at x = 0, because of the
symmetric distribution of isotherms with respect to the vertical
centerline at x = 0, as shown in Fig. 5(a)–(f).
3.3. Ra = 105
Fig. 7 shows the distribution of isotherms and streamlines for
different yc and d values when an inner single cylinder and two
inner cylinders move downward and upward from yc = 0, respec-
tively, at Ra = 105. When Ra = 105, the effect of convection on the
fluid flow and heat transfer is increasingly predominant with
increasing Rayleigh numbers. As a result, the distributions of iso-
therms and streamlines in both cases of the single cylinder and
two cylinders at Ra = 105 are different from those at Ra = 103 and
104.
In case that an inner single cylinder moves vertically in the
enclosure with yc < 0, the magnitude and intensity of the ascending
mushroom-shaped plume at Ra = 105 are larger than those at
Ra = 104. Thus, as the value of yc decreases with yc < 0 at Ra = 105
in the case of the single cylinder, the size of the ascending mush-
room-shaped plume increases because more space is created
 Y.G. Park et al. / International Journal of Heat and Mass Transfer 77 (2014) 501–518 509

Fig. 6. Distribution of local Nusselt numbers along the surface of the inner cylinders (Nucyl, Nuupper cyl and Nulower cyl ) and the walls of the enclosure (Nuen) for different yc and d
values in the case of the single cylinder (left column) and in the case of two cylinders (right column) at Ra = 104.

between the inner cylinder and the top wall of the enclosure,
allowing for upward movement of the isotherms. When the inner
single cylinder moves upward with yc > 0 at Ra = 105, the pattern
of isotherms is much different from that when the inner single cyl-
inder moves downward. When yc = 0.25 in the case of the single
cylinder, two ascending plumes and one weak descending plume
are formed between the inner cylinder and the top wall of the
enclosure, as shown in Fig. 7(e), because of the decreasing space
between the inner cylinder and the top wall of the enclosure. As
yc increases further in the case of the single cylinder, no descending
plume is found at yc = 0.35 because of the limited space available
for the upward movement of the isotherms between the inner cyl-
inder and the top wall of the enclosure. As the Rayleigh number
increased to 105 in the case of the single cylinder, the thermal gra-
dients on the top wall of the enclosure and the surface of the inner
cylinder becomes larger, and the isotherms at the lower part of the
510 Y.G. Park et al. / International Journal of Heat and Mass Transfer 77 (2014) 501–518

(a) yc = −0.15 (g) δ = 0.3( yc = ±0.15)

(b) yc = −0.25 (h) δ = 0.4( yc = ±0.20)

(c) yc = −0.35 (i) δ = 0.5( yc = ±0.25)

(d) yc = 0.15 (j) δ = 0.6( yc = ±0.30)

(e) yc = 0.25 (k) δ = 0.7 ( yc = ±0.35)

(f) yc = 0.35
Fig. 7. Isotherms and streamlines for different yc and d values in the case of the single cylinder (left column) and in the case of two cylinders (right column) at Ra = 105
(isotherm contour values range from 0 to 1 with 10 levels).

enclosure becomes more stratified with an increasing stagnation
region because of the increasing effect of convection with increas-
ing Rayleigh numbers, as compared to those at Ra = 104.
The distribution of isotherms and streamlines in the case of two
cylinders, as shown in Fig. 7(g)–(k), is different from that in the
case of the single cylinder, as shown in Fig. 7(a)–(f), because of
the mutual interaction between two adjacent inner hot cylinders
in addition to the interaction between the two inner hot cylinders
and the cold enclosure walls, as compared to the interaction
between the inner single hot cylinder and the cold enclosure walls.
When d ¼ 0:3 and 0.4 in the case of two cylinders, two upwelling
plumes appear on the top surface of the upper inner cylinder
and, as a result, two secondary vortices are newly generated
between the upper inner cylinder and the top wall of the enclosure,
as shown in Fig. 7(g) and (h), unlike the case of the single cylinder
in which two upwelling plumes and two secondary vortices
formed on the top surface of the upper inner cylinder are not
observed at yc = 0.15, as shown in Fig. 7(d). As d increases further
in the case of two cylinders, the reduced space above the top of
the upper inner cylinder confines the vertical motion of flow and,
 Y.G. Park et al. / International Journal of Heat and Mass Transfer 77 (2014) 501–518
consequently, the heat conduction is locally predominant over the
convective heat transfer in this space. Thus, two secondary vortices
below the top wall of the enclosure disappear, as shown in
Fig. 7(i)–(k).
Fig. 8 shows the distribution of local Nusselt numbers along the
surface of the inner cylinders and the walls of the enclosure for dif-
ferent yc and d values when an inner single cylinder and two inner
cylinders move downward and upward from yc = 0, respectively, at
Ra = 105. The distribution of local Nusselt numbers for Ra = 105 in
both cases of the single cylinder and two cylinders, as shown in
Fig. 8, is different from that for Ra = 103 and 104, as shown in Figs. 4
and 6, owing to the increasing effect of convection induced by the
ascending plume from the single hot cylinder and two hot cylin-
ders, respectively, with increasing Rayleigh number, in addition
to the effect of yc and d. The distribution of local Nusselt numbers
in the case of two cylinders at Ra = 105, as shown in Fig. 8(c)–(e), is
also different from that in the case of the single cylinder, as shown
in Fig. 8(a) and (b), owing to the mutual interaction between two
adjacent cylinders in addition to the interaction between two hot
inner cylinders and the cold enclosure walls in the case of two cyl-
inders, as compared to the interaction just between the single hot
cylinder and the cold enclosure wall in the case of the single cylin-
der. The values of local Nusselt numbers for Ra = 105 in both cases
of the single cylinder and two cylinders are higher than those for
Ra = 103 and 104 as a result of increasing convection effect with
increasing Rayleigh number.
When the inner single cylinder moves vertically at yc 6 0.15, a
minimum peak of Nucyl appears at approximately u ¼ 0 (or
u ¼ 360 ), as shown in Fig. 8(a), due to the presence of the rising
plume on the top of the inner cylinder. However, when yc = 0.25
and 0.35 in the case of the single cylinder, Nucyl has a minimum
peak at approximately u ¼ 41 (or u ¼ 319 ) and u ¼ 74 (or
u ¼ 286 ), respectively, because the rising plume on the top sur-
face of the inner single cylinder moves in the direction of increas-
ing or decreasing u. In the case of the single cylinder, the values of
Nucyl around u ¼ 180 at yc 6 0.25 for Ra = 105 are much larger
than those for Ra = 104, owing to the increasing convection effect.
When two cylinders move vertically at Ra = 105, Nuupper cyl has
two local low peaks at around u ¼ 41 —90 or u ¼ 270 —321 ,
as shown in Fig. 8(c), because the position of the thermal plume
on the surface of the upper inner cylinder depends on d. When
d = 0.3 in the case of two cylinders at Ra = 105, the value of
Nuupper cyl only has a third local low peak at around u ¼ 180 ,
because the distance between the upper and lower hot cylinders
is small. As we increase d in the case of two cylinders at Ra = 105,
the value of Nuupper cyl at u ¼ 180 increases and, as a result,
Nuupper cyl has a high peak at u ¼ 180 when d P 0:5. The value of
Nulower cyl in the case of two cylinders has three local high peaks
at around u ¼ 100 —105 , u ¼ 180 , and u ¼ 255 —260 , and
increases with increasing d, as shown in Fig. 8(d). However, when
d = 0.3 and d = 0.4 in the case of two cylinders, the value of Nulower cyl
at around u ¼ 180 is larger than that when d = 0.5, because of the
reduced buoyancy-induced convection as a result of less space
secured between the lower cylinder and the bottom wall of the
enclosure at d = 0.5.
As we move from point A to point B along the top wall of the
enclosure in the case of the single cylinder at Ra = 105, Nuen has a
single high peak at the center of the top wall of the enclosure in
the presence of a single ascending plume from the single hot cylin-
der, as shown in Fig. 8(b), except at yc = 0.25 where Nuen has two
high peaks and a single low peak, as a result of two ascending
plumes and one weak descending plume formed between the inner
cylinder and the top wall of the enclosure. However, as we move
from point A to point B along the top wall of the enclosure in the
case of two cylinders, Nuen for Ra = 105 has a single high peak at
d = 0.6 and d = 0.7, while it has two high peaks and a single low
511
peak at d = 0.3 and d = 0.4 because of the presence of the secondary
vortices that form on the top wall of the enclosure, as shown in
Fig. 7(g) and (h). When d = 0.5 at Ra = 105, Nuen along the top wall
of the enclosure in the case of two cylinders has an almost uniform
value around the center of the top wall.
As we move from point B to point C along the right wall of the
enclosure in both cases of the single cylinder and two cylinders at
Ra = 105, the location at which Nuen has its highest peak value gets
closer to point B, as compared to that at Ra = 104 as a result of the
increasing convection effect with increasing Rayleigh number. The
variation in the value of Nuen on the right wall of the enclosure
according to different d values at Ra = 105 in the case of two cylin-
ders is smaller than that in the case of the single cylinder, because
of the interaction at the same distance between two hot inner
cylinders and the right cold wall of the enclosure, as compared to
the interaction between the single hot inner cylinder and the right
cold wall of the enclosure.
As we move from point C to point D along the bottom wall of
the enclosure in both cases of the single cylinder and two cylinders,
the variation in the value of Nuen according to different yc and d val-
ues at Ra = 105 is generally similar to that at Ra = 104. However, the
values of Nuen on the bottom wall of the enclosure at Ra = 105 is
smaller than that at Ra = 104, because of the increasing stagnant
region around the bottom wall of the enclosure caused by the
increasing convection effect with increasing Rayleigh number,
except the values of Nuen at yc = 0.35 in the case of the single cylin-
der and d = 0.7 in the case of two cylinders where the distance
effect between the lower cylinder and the bottom wall of the
enclosure becomes dominant.
The pattern for the variation in Nuen along the left and right
walls of the enclosure according to yc and d in both cases of the
single cylinder and two cylinders is symmetric about the vertical
centerline at x = 0.
3.4. Ra = 106
Fig. 9 shows the distribution of isotherms and streamlines for
different yc and d values when an inner single cylinder and two
inner cylinders move downward and upward from yc = 0, respec-
tively, at Ra = 106. When the Rayleigh number increases to 106,
the magnitude of the velocity circulating in the enclosure increases
and the isotherms in both cases of the single cylinder and two cyl-
inders are more distorted as a result of the stronger convection
effects, leading to the stable stratification of the isotherms. Thus,
in both cases of the single cylinder and two cylinders, the thickness
of the thermal boundary formed on the surfaces of the inner cylin-
ders and the enclosure becomes thinner and the thermal gradients
on the walls become larger when Ra = 106, as compared to those
when Ra = 105.
When an inner single cylinder moves vertically in the enclosure
at Ra = 106, as shown in Fig. 9(a)–(f), the rising thermal plume is
present on the top of the inner cylinder. As we increase yc in the
case of the single cylinder at Ra = 106, this rising plume on the
top surface of the inner cylinder moves to the increasing or
decreasing u direction as a result of the decreasing space between
the inner cylinder and the top wall of the enclosure. When
yc = 0.15 and yc = 0.15 at Ra = 106 in the case of the single cylin-
der, we can observe the formation of the secondary vortices on
the bottom wall of the enclosure, which are not present at Ra = 105.
When two inner cylinders are located at d = 0.5 at Ra = 106, the
flow and thermal fields become time-dependent, as mentioned
before. Therefore, the isotherms and streamlines at d = 0.5 in the
case of two cylinders, as shown in Fig. 9(i), represent the time-
averaged temperature and velocity fields. Because the effect of
convection on the fluid flow and heat transfer at Ra = 106 becomes
much stronger than that at Ra = 105, the distributions of the
512 Y.G. Park et al. / International Journal of Heat and Mass Transfer 77 (2014) 501–518

Fig. 8. Distribution of local Nusselt numbers along the surface of the inner cylinders (Nucyl, Nuupper cyl and Nulower cyl ) and the walls of the enclosure (Nuen) for different yc and d
values in the case of the single cylinder (left column) and in the case of two cylinders (right column) at Ra = 105.

isotherms and streamlines in the case of two cylinders at Ra = 106
are different from those at Ra = 105, as shown in Figs. 9(g)–(k) and
7(g)–(k), respectively. The distributions of isotherms between the
two adjacent inner cylinders in the case of two cylinders at
Ra = 106 are denser and directed more upward, as compared to
those at Ra = 105. The intensity of the mushroom-shaped plume
rising from the two inner cylinders and impinging on the top wall
of the enclosure in the case of two cylinders at Ra = 106 is also lar-
ger than that at Ra = 105, because of increasing convection with
increasing Rayleigh numbers. As a result, when Ra = 106 in the case
of two cylinders, the secondary vortices that form on the top wall
of the enclosure at d ¼ 0:3 and d = 0.4 when Ra = 105, as shown in
Fig. 7(g) and (h), disappear and other secondary vortices appear
on the bottom wall of the enclosure at d = 0.3, as shown in
Fig. 9(g). When d = 0.5 in the case of two cylinders at Ra = 106, a
plume at d = 0.4 is divided into two ascending plumes on the upper
inner cylinder and a descending plume on the top wall of the
enclosure, corresponding to the secondary vortices formed on the
 Y.G. Park et al. / International Journal of Heat and Mass Transfer 77 (2014) 501–518 513

(a) yc = −0.15 (g) δ = 0.3( yc = ±0.15)

(b) yc = −0.25 (h) δ = 0.4( yc = ±0.20)

(c) yc = −0.35 (i) δ = 0.5( yc = ±0.25)

(d) yc = 0.15 (j) δ = 0.6( yc = ±0.30)

(e) yc = 0.25 (k) δ = 0.7 ( yc = ±0.35)

(f) yc = 0.35
Fig. 9. Isotherms and streamlines for different yc and d values in the case of the single cylinder (left column) and in the case of two cylinders (right column) at Ra = 106
(contour values range from 0 to 1 with 10 levels).

upper inner cylinder, as shown in Fig. 9(i). When d = 0.6 in the case
of two cylinders at Ra = 106, three plumes at d = 0.5 are divided into
three ascending plumes on the upper inner cylinder and two
descending plumes on the top wall of the enclosure, corresponding
to the tertiary vortices formed on the upper inner cylinder, as
shown in Fig. 9(j). When d = 0.7 in the case of two cylinders at
Ra = 106, the distance between the upper inner cylinder and the
top wall of the enclosure is small. As a result, the tertiary vortices
formed on the upper inner cylinder disappear, as shown in
Fig. 9(k).
Fig. 10 shows the time histories and power spectra of the
surface-averaged Nusselt numbers of the upper cylinder and the
distribution of the instantaneous isotherms and streamlines at
selected time instances at d = 0.5 in the case of two cylinders when
Ra = 106. The time histories of the surface-averaged Nusselt num-
bers around the lower cylinder, Nulower cyl , at d = 0.5 for Ra = 106
have small variations. Therefore, we show only the time histories
of Nuupper cyl for brevity. The time histories of Nuupper cyl at d = 0.5
for Ra = 106 in the case of two cylinders oscillate in a periodic fash-
ion as a function of time, as shown in Fig. 10(a), resulting in the
514 Y.G. Park et al. / International Journal of Heat and Mass Transfer 77 (2014) 501–518
(a)
(c) t = A
(e) t = C
Fig. 10. (a) Time histories, (b) power spectrum of the surface-averaged Nusselt number of the upper cylinder, (c)–(f) instantaneous isotherms and streamlines at selected
time instances at Ra = 106 and d = 0.5 (yc = ±0.25).
dominant primary frequency of 88 with secondary frequency of
173 appearing in the power spectrum, as shown in Fig. 10(b).
The instantaneous isotherms and streamlines at four different
times, which are denoted by A; B; C and D in Fig. 10(a), are shown
in Fig. 10(c)–(f). At t = A, at which Nuupper cyl has a maximum value,
one downwelling plume from the top wall of the enclosure and
two upwelling plumes from the upper inner cylinder are formed,
resulting in the formation of two secondary vortices on the top
wall of the enclosure. During the time interval between t = A and
t = B, the downwelling plume and the two secondary vortices from
the top wall of the enclosure move rightward. During the time
interval between t = B and t = C, the downwelling plume moves fur-
ther to the right and disappears and, as a result, two secondary vor-
tices on the top wall of the enclosure disappear. Because of the
presence of the upwelling plume around the upper inner cylinder,
Nuupper cyl has a minimum peak value at t = C. During the time inter-
val between t = C and t = D, one newly formed downwelling plume
from the top wall of the enclosure appears, resulting in the forma-
tion of two secondary vortices on the top wall of the enclosure.
These distributions of isotherms and streamlines repeat their
shapes periodically. Therefore, at d = 0.5 for Ra = 106 in the case
of two cylinders, the time-averaged isotherm and streamline are
asymmetric about the vertical centerline at x = 0, as shown in
Fig. 9(i).
Fig. 11 the distribution of local Nusselt numbers along the sur-
face of the inner cylinders and the walls of the enclosure for differ-
ent yc and d values when an inner single cylinder and two inner
(b)
(d) t = B
(f) t = D
cylinders move downward and upward from yc = 0, respectively,
at Ra = 106. The values of the local Nusselt number at Ra = 106 in
both cases of the single cylinder and two cylinders are larger than
those at Ra ¼ 103 ; 104 and 105 , due to much larger effect of con-
vection with increasing Rayleigh number.
When the single cylinder moves downward with yc < 0, the pat-
tern of variation of the local Nusselt number along the surface of
the inner cylinder is different from that of Ra = 105. The value of
Nucyl for Ra = 106 in the case of the single cylinder has a local low
peak at around u ¼ 180 , as shown in Fig. 11(a), because the con-
vection effect is stronger than the effect of the gap distance
between the inner cylinder and the bottom walls of the enclosure.
When yc = 0.35 at Ra = 106 in the case of the single cylinder, Nucyl
has a minimum peak at approximately u ¼ 52 (or u ¼ 308 ),
due to the presence of two ascending plumes.
In the case of two cylinders, the distributions of Nuupper cyl ,
Nulower cyl and Nuen at d = 0.5 for Ra = 106 show the time-averaged
distributions, since the flow and thermal fields are time-depen-
dent. When two cylinders move vertically in the enclosure, the var-
iation in the distribution of Nuupper cyl according to d at Ra = 106 is
similar to that at Ra = 105. In the case of two cylinders, Nuupper cyl
for Ra = 106 has local low peaks at around u ¼ 0 —65 , as com-
pared to u ¼ 90 —41 for Ra = 105, or u ¼ 275 —360 , as compared
to u ¼ 270 —321 for Ra = 105, depending on d. When d = 0.3 in the
case of two cylinders at Ra = 106, Nuupper cyl has a low peak at
u ¼ 180 . However, as we increase d in the case of two cylinders
at Ra = 106, the value of Nuupper cyl around u ¼ 180 increases and,
 Y.G. Park et al. / International Journal of Heat and Mass Transfer 77 (2014) 501–518
as a result, Nuupper cyl has a high peak around u ¼ 180 . The value of
Nuupper cyl in the case of two cylinders at Ra = 106 is much higher
than that at Ra = 103, 104 and 105, because of the strong effect of
the convection caused by the rising thermal plume from the lower
hot cylinder.
The pattern of variation of Nulower cyl . according to d in the case of
two cylinders at Ra = 106 is also similar to that at Ra = 105. How-
ever, the values of Nulower cyl in the case of two cylinders at
Ra = 106 are larger than those at Ra = 105, due to the strong effect
of convection with increasing Rayleigh number. The value of
Nulower cyl in the case of two cylinders at Ra = 106 has three local
high peaks at around u ¼ 93 —103 , u ¼ 180 , and
u ¼ 257 —267 , depending on d. The values of Nulower cyl in the case
of two cylinders at Ra = 106 increase with increasing d in the region
of u ¼ 93 —103 and u ¼ 257 —267 , while they decrease with
increasing d in the region of u ¼ 108 except that for d = 0.7.
As we move from point A to point B along the top wall of the
enclosure at yc 6 0:25 in the case of the single cylinder at
Ra = 106, Nuen has a single high peak at the center of the top wall
of the enclosure in the presence of a single ascending plume from
the single hot cylinder, as shown in Fig. 10(b). However, Nuen has
two high peaks and a single low peak at yc = 0.35 in the case of
the single cylinder at Ra = 106, because of the presence of the
two ascending plumes and one weak descending plume which is
formed between the inner cylinder and the top wall of the enclo-
sure, unlike Ra = 105 at which Nuen has two high peaks and a single
low peak at yc = 0.25.
As we move from point A to point B along the top wall of the
enclosure in the case of two cylinders, the distribution of Nuen for
Ra = 106 is much different from that for Ra = 106. When d = 0.3 in
the case of two cylinders for Ra = 106, the value of Nuen on the
top wall has only single high peak at the center of the top wall
because of the presence of a single ascending plume from two
hot cylinders, unlike two high peaks and one low peaks of Nuen
in the presence of the secondary vortices that form on the top wall
of the enclosure for Ra = 105. When d = 0.4 in the case of two cylin-
ders for Ra = 106, the value of Nuen on the top wall has two high
peaks and a single low peak because of the presence of the two
ascending plumes and one weak descending plume which is
formed between the inner cylinder and the top wall of the enclo-
sure. When d = 0.5 in the case of two cylinders for Ra = 106, the
time-averaged value of Nuen on the top wall has also two high
peaks and a single low peak. However, the peak value of Nuen at
the left side of the top wall is larger than that at the right side of
the top wall, because of the asymmetric distribution of the time-
averaged isotherms about the vertical centerline at x = 0, as shown
in Fig. 9(i). When d = 0.6 in the case of two cylinders for Ra = 106,
the value of Nuen on the top wall has three high peaks and two
low peaks, because of the presence of three ascending plumes on
the upper inner cylinder and two descending plumes on the top
wall of the enclosure corresponding to the tertiary vortices formed
on the upper inner cylinder, as shown in Fig. 9(j). When d = 0.7 in
the case of two cylinders for Ra = 106, the value of Nuen on the
top wall has still three high peaks and two low peaks, even though
the tertiary vortices formed on the upper inner cylinder disappear,
as shown in Fig. 9(k). The values of of Nuen along the top wall of the
enclosure at yc 6 0:25 in both cases of the single cylinder and two
cylinders for Ra = 106 are much higher than those for Ra = 105,
because of increasing effect of convection with increasing Rayleigh
number.
As we move from point B to point C along the right wall of the
enclosure in both cases of the single cylinder and two cylinders at
Ra = 106, the location at which Nuen has its highest peak value gets
closer to point B, similar to that at Ra = 105, as a result of the large
effect of convection. The values of Nuen along the right wall of the
enclosure in the case of two cylinders for Ra = 106 are larger than
515
those in the case of the single cylinder for Ra = 105, because of
increasing effect of convection with increasing Rayleigh number.
When the single inner cylinder is located at the lower part of
the enclosure, the variation in the value of Nuen along the right wall
of the enclosure is small because of the strong effect of the ascend-
ing plume from the hot inner single cylinder. However, when the
single inner cylinder is located at the upper part of the enclosure,
the variation in the value of Nuen along the right wall of the enclo-
sure according to yc is large, resulting in decreasing value of Nuen
because of the increasing stagnation region between the lower sur-
face of the hot inner cylinder and the bottom wall of the enclosure
as yc increases.
The variation in the values of Nuen along the right wall of the
enclosure in the case of two cylinders is relatively lower than that
in the case of the single cylinder, because two hot cylinders are
located at both positions of the lower and upper parts of the enclo-
sure. The values of Nuen along the right wall of the enclosure in the
case of two cylinders are larger than those in the case of the single
cylinder, because of heat transfer from two hot cylinders at the
lower and upper parts of the enclosure to the right cold wall, as
compared to heat transfer from the single hot cylinder.
As we move from point C to point D along the bottom wall of
the enclosure in both cases of the single cylinder and two cylinders,
the values of Nuen according to yc and d values at Ra = 106 are smal-
ler than those at Ra = 105, except at yc = 0.35 and d = 0.7, because
of the increasing effect of the stagnation region between the cylin-
der and the lower wall of the enclosure with increasing Rayleigh
number.
When Ra = 105, Nuen along the bottom wall in both cases of the
single cylinder and two cylinders has only one high peak around
the center of the bottom wall. However, the distribution of Nuen
along the bottom wall in both cases of the single cylinder and
two cylinders at Ra = 106 is different from that at Ra = 105, because
of much stronger effect of convection at Ra = 106 than Ra = 105.
When yc P 0:15 in the case of the single cylinder at Ra = 106, the
values of Nuen along the bottom wall are very small without any
peak values because of the larger space between the inner single
cylinder and the bottom wall. When yc = 0.15 and yc = 0 in the
case of the single cylinder at Ra = 106, Nuen along the bottom wall
has two high peaks and the variation in the values of Nuen around
the center of the bottom wall between these two high peaks is
small. When yc = 0.35 and yc = 0.25 in the case of the single cyl-
inder at Ra = 106, Nuen along the bottom wall has three high peaks
and two low peaks, because of the interaction between the inner
hot cylinder and the bottom wall in the presence of the strong
convection.
When d = 0.3 and d = 0.4 in the case of two cylinders at Ra = 106,
Nuen along the bottom wall has two high peaks and the variation in
the values of Nuen around the center of the bottom wall between
these two high peaks is small, similar to that when yc = 0.15
and yc = 0 in the case of the single cylinder. When d = 0.5, d = 0.6
and d = 0.7 in the case of two cylinders at Ra = 106, Nuen along
the bottom wall has three high peaks and two low peaks, similar
to that when yc = 0.25 and yc = 0.35 in the case of the single cyl-
inder, because of the interaction between the inner hot cylinders
and the bottom wall in the presence of the strong convection.
The pattern for the variation in Nuen along the left and right
walls of the enclosure according to yc in both cases of the single
cylinder and two cylinders is symmetric about the vertical center-
line at x = 0.
3.5. Surface-averaged Nusselt numbers
Fig. 12 shows the surface-averaged Nusselt numbers of the
enclosure and the cylinder as a function of yc for different Rayleigh
numbers for both cases of the single cylinder and two cylinders.
516 Y.G. Park et al. / International Journal of Heat and Mass Transfer 77 (2014) 501–518

Fig. 11. Distribution of local Nusselt numbers along the surface of the inner cylinders (Nucyl, Nuupper cyl and Nulower cyl ) and the walls of the enclosure (Nuen) for different yc and d
values in the case of the single cylinder (left column) and in the case of two cylinders (right column) at Ra = 106.

In the case of the single cylinder, when Ra = 103 and 104, Nuen
has an almost parabolic profile with a minimum value at d = 0
because of the dominant conduction heat transfer mode in the
presence of weak convection. However, when Ra = 105 and 106,
the symmetry of Nuen is broken because of the increasing convec-
tion with increasing Rayleigh number. The pattern of the variation
of Nucyl as a function of yc is generally similar to that of Nuen . How-
ever, the magnitude of Nucyl is larger than that of Nuen for all the
different cylinder positions because the isotherms form more den-
sely on the surface of the inner cylinder surface than on the walls
of the enclosure.
In the case of two cylinders, the values of Nuen and Nulower cyl
(which corresponds to Nucyl at yc < 0 in Fig. 12) generally increase
with increasing Rayleigh number as a result of the increasing effect
of convection, similar to the case of the single cylinder. However,
the variation of Nuupper cyl (which corresponds to Nucyl at yc > 0 in
 Y.G. Park et al. / International Journal of Heat and Mass Transfer 77 (2014) 501–518
Fig. 12. Surface-averaged Nusselt numbers as function of yc for the single cylinder
and two cylinders at different Rayleigh numbers: (a) enclosure (Nuen ) and (b)
cylinder (Nucyl ).
Fig. 12) according to the Rayleigh number in the case of two cylin-
ders is different from that in the case of the single cylinder. The val-
ues of Nuupper cyl in the case of two cylinders at Ra = 103 are larger
than those at Ra = 104, because the effect of the convection from
the lower cylinder on the distribution of isotherms around the
upper cylinder at Ra = 104 is larger than that at Ra = 103, as shown
in Figs. 3 and 5, resulting in a reduction in the value of Nuupper cyl at
Ra = 104 compared to that at Ra = 103. As we increase the Rayleigh
number to 105, the values of Nuupper cyl are slightly larger than those
at Ra = 103 and 104, unlike large increase in the values of Nuen , Nucyl
(single cylinder) and Nulower cyl with increasing Rayleigh number
from 104 to 105 at yc > 0. As we increase the Rayleigh number to
106, the values of Nuupper cyl become much larger than those at
Ra = 103, 104 and 105, because of much stronger effect of convec-
tion. Because the lower hot cylinder reduces the heat transfer per-
formance of the upper hot cylinder as a result of the effect of the
convection from the lower cylinder on the distribution of iso-
therms around the upper cylinder, the values of Nulower cyl in the
cases of two cylinders are larger than those of Nuupper cyl , as shown
in Fig. 12(b).
The values of Nuen in the case of two cylinders are larger those
that in the case of the single cylinder for all Rayleigh numbers and
cylinder positions considered because more heat is transferred to
the enclosure walls in the case of two cylinders than that in the
517
case of the single cylinder. However, the difference in the values
between Nucyl in the case of the single cylinder and Nulower cyl and
Nuupper cyl in the case of two adjacent cylinders depends on the Ray-
leigh number and cylinder positions, because of the interaction
between two hot cylinders in addition to the interaction between
the two hot cylinders and the cold enclosure wall, as compared
to the interaction between the single hot cylinder and the cold
enclosure walls. When Ra = 103, the values of Nulower cyl and
Nuupper cyl in the case of two cylinders are lower at 0:2 6 yc 6 0:3
and higher at 0:35 6 yc 6 0:25 and yc = 0.35, respectively, as
compared to those in the case of the single cylinder. When we
increase the Rayleigh number to 104, the values of Nulower cyl and
Nuupper cyl in the case of two cylinders are lower at
0:15 6 yc 6 0:3 and higher at 0:35 6 yc 6 0:2, respectively,
as compared to those in the case of the single cylinder. When we
increase the Rayleigh number further to 105 and 106, the values
of Nulower cyl in the case of two cylinders are higher at
0:35 6 yc 6 0:15, while the values of Nuupper cyl in the case of
two cylinders are much lower at 0:15 6 yc 6 0:35, as compared
to those for the case of the single cylinder, because heat transfer
from the lower cylinder diminishes the heat transfer performance
of the upper cylinder.
4. Conclusions
The present study investigates the natural convection induced
by a temperature difference between a cold outer square enclosure
and two hot inner circular cylinders for different Rayleigh numbers
in the range of 103 6 Ra 6 106 . The immersed boundary method
based on the finite volume method is used to handle inner cylin-
ders located at the vertical centerline of the enclosure in the range
of 0:3 6 d 6 0:7. The results for the case of two cylinders are com-
pared with those for the case of the single cylinder.
The flow and thermal fields eventually reach the steady state
and are symmetric about the vertical centerline at x = 0 for all dif-
ferent values of Ra, yc, and d considered in the present study in both
cases of the single cylinder and two cylinders, except those at
d = 0.5at Ra = 106 in the case of two cylinders. The time histories
of isotherms, streamlines and surface-averaged Nusselt numbers
at d = 0.5 for Ra = 106 in the case of two cylinders oscillate in a peri-
odic fashion as a function of time, resulting in the asymmetric dis-
tribution of the time-averaged isotherms and streamlines about
the vertical centerline at x = 0.
The distribution of isotherms and streamlines in the enclosure
in the case of two cylinders depends on the mutual interaction
between the adjacent two hot cylinders in addition to the interac-
tion between the two inner hot cylinders and the cold enclosure
walls, unlike the inner single cylinder case that the distribution
of the flow and thermal fields in the enclosure depends on only
the interaction between the inner hot single cylinder and the cold
enclosure walls. As a result, the number, size, and formation of the
cells and the distribution of Nusselt numbers on the cylinder sur-
faces and the walls of the enclosure in the case of two cylinders
are different from those in the case of the single cylinder.
Because the lower hot cylinder reduces the heat transfer perfor-
mance of the upper hot cylinder as a result of the effect of the con-
vection from the lower cylinder on the distribution of isotherms
around the upper cylinder, the values of Nulower cyl are larger than
those of Nuupper cyl in the case of two cylinders for all Rayleigh num-
bers considered. The values of Nulower cyl increase with increasing
Rayleigh number because of the increasing effect of convection.
However, the values of Nuupper cyl in the case of two cylinders at
Ra = 104 are smaller than those at Ra = 103 because the effect of the
518 Y.G. Park et al. / International Journal of Heat and Mass Transfer 77 (2014) 501–518
convection from the lower cylinder on the distribution of iso-
therms around the upper cylinder at Ra = 104 is larger than that
at Ra = 103. The values of Nuupper cyl in the case of two cylinders at
Ra = 105 are slightly larger than those at Ra = 103 and 104, unlike
large increase in the values of Nuen and Nulower cyl with increasing
Rayleigh number from 104 to 105. The values of Nuupper cyl in the
case of two cylinders at Ra = 106 become much larger than those
at Ra = 103, 104 and 105, because of much stronger effect of
convection.
The values of Nuen in the case of two cylinders are larger those
that in the case of the single cylinder for all Rayleigh numbers and
cylinder positions considered because more heat is transferred to
the enclosure walls for the case of two cylinders than the case of
the single cylinder. However, the difference in the values between
Nucyl in the case of the single cylinder and Nulower cyl and Nuupper cyl in
the case of two cylinders depends on the Rayleigh number and
cylinder positions.
Conflict of interest
None declared.
Acknowledgments
This research was supported by Leading Foreign Research Insti-
tute Recruitment Program through the National Research Founda-
tion of Korea (NRF) funded by the Ministry of Science, ICT & Future
Planning (No. 2009-00495). This work was supported by the
National Research Foundation of Korea (NRF) grant funded by the
Korea government (MSIP) (No. NRF-2013R1A2A2A01067251).
References
[1] G. Cesini, M. Paroncini, G. Cortella, M. Manzan, Natural convection from a
horizontal cylinder in a rectangular cavity, Int. J. Heat Mass Transfer 42 (1999)
1801–1811.
[2] H. Asan, Natural convection in an annulus between two isothermal concentric
square ducts, Int. Commun. Heat Mass Transfer 27 (2000) 367–376.
[3] F. Moukalled, S. Acharya, Natural convection in the annulus between
concentric horizontal circular and square cylinders, J. Thermophys. Heat
Transfer 10 (1996) 524–531.
[4] C. Shu, Y.D. Zhu, Efficient computation of natural convection in a concentric
annulus between an outer square cylinder and an inner circular cylinder, Int. J.
Numer. Methods Fluids 38 (2002) 429–445.
[5] N.K. Ghaddar, Natural convection heat transfer between a uniformly heated
cylindrical element and its rectangular enclosure, Int. J. Heat Mass Transfer 35
(1992) 2327–2334.
[6] A. Kumar De, A. Dalal, A numerical study of natural convection around a
square, horizontal, heated cylinder placed in an enclosure, Int. J. Heat Mass
Transfer 49 (2006) 4608–4623.
[7] B.S. Kim, D.S. Lee, M.Y. Ha, H.S. Yoon, A numerical study of natural convection
in a square enclosure with a circular cylinder at different vertical locations, Int.
J. Heat Mass Transfer 51 (2008) 1888–1906.
[8] H.S. Yoon, M.Y. Ha, B.S. Kim, D.H. Yu, Effect of the position of a circular cylinder
in a square enclosure on natural convection at Rayleigh number of 107, Phys.
Fluids 21 (2009) 1–11.
[9] J.M. Lee, M.Y. Ha, H.S. Yoon, Natural convection in a square enclosure with a
circular cylinder at different horizontal and diagonal locations, Int. J. Heat Mass
Transfer 53 (2010) 5905–5919.
[10] C. Shu, H. Xue, Y.D. Zhu, Numerical study of natural convection in an eccentric
annulus between a square outer cylinder and a circular inner cylinder using
DQ method, Int. J. Heat Mass Transfer 44 (2000) 3321–3333.
[11] D.H. Kang, M.Y. Ha, H.S. Yoon, C. Choi, Bifurcation to unsteady natural
convection in square enclosure with a circular cylinder at Rayleigh number of
107, Int. J. Heat Mass Transfer 64 (2013) 926–944.
[12] M. Lacroix, Natural convection heat transfer around two heated horizontal
cylinders inside a rectangular cavity cooled from above, Numer. Heat Transfer
A 21 (1992) 37–54.
[13] M. Lacroix, A. Joyeux, Coupling of wall conduction with natural convection
from heated cylinder in a rectangular enclosure, Int. Commun. Heat Mass
Transfer 23 (1996) 143–151.
[14] M. Corcione, Interactive free convection from a pair of vertical tube-arrays at
moderate Rayleigh numbers, Int. J. Heat Mass Transfer 50 (2007) 1061–1074.
[15] R. Chouikh, A. Guizani, M. Maalej, A. Belghith, Numerical study of the laminar
natural convection flow around an array of two horizontal isothermal
cylinders, Int. Commun. Heat Mass Transfer 26 (1999) 329–338.
[16] Y.G. Park, H.S. Yoon, M.Y. Ha, Natural convection in square enclosure with hot
and cold cylinders at different vertical locations, Int. J. Heat Mass Transfer 55
(2012) 7911–7925.
[17] Y.G. Park, H.S. Yoon, M.Y. Ha, Study on natural convection in a cold square
enclosure with a pair of hot horizontal cylinders positioned at different
vertical locations, Int. J. Heat Mass Transfer 65 (2013) 696–712.
[18] M. Chae, B. Chung, Effect of pitch-to-diameter ratio on the natural convection
heat transfer of two vertically aligned horizontal cylinders, Chem. Eng. Sci. 66
(2011) 5321–5329.
[19] M.S. Sadeghipour, M. Asheghi, Free convection heat transfer from array of
vertically separated horizontal cylinders at low Rayleigh numbers, Int. J. Heat
Mass Transfer 37 (1994) 103–109.
[20] R.O. Warrington, R.E. Powe, The heat transfer of heat by natural convection
between bodies and their enclosures, Int. J. Heat Mass Transfer 28 (1985) 319–
330.
[21] J. Kim, P. Moin, Application of a fractional step method to incompressible
Navier–Stokes equations, J. Comput. Phys. 59 (1985) 308–323.
[22] Y. Zang, R.L. Street, J.R. Koseff, A non-staggered grid, fractional step method for
time-dependent incompressible Navier–Stokes equations in curvilinear
coordinates, J. Comput. Phys. 114 (1994) 18–33.
[23] J. Kim, D. Kim, H. Choi, An immersed-boundary finite volume method for
simulations of flow in complex geometries, J. Comput. Phys. 171 (2001) 132–
150.
[24] J. Kim, H. Choi, An immersed-boundary finite-volume method for simulation
of heat transfer in complex geometries, KSME Int. J. 18 (2004) 1026–1035.