International Journal of Heat and Mass Transfer 55 (2012) 7911–7925

Contents lists available at SciVerse ScienceDirect

International Journal of Heat and Mass Transfer

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

Natural convection in square enclosure with hot and cold cylinders

at different vertical locations
Yong Gap Park a,1, Hyun Sik Yoon b,2, Man Yeong Ha a,⇑
a
School of Mechanical Engineering, Pusan National University, Jang Jeon 2-Dong, Geum Jeong Gu, Busan 609-735, Republic of Korea
b
Global Core Research Center for Ships and Offshore Plants, 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: This study investigates the natural convection in a square enclosure with hot and cold cylinders, induced
Received 29 May 2012 by the temperature difference between a cold enclosure and the hot and cold circular cylinders contained
Received in revised form 1 August 2012 within it. The immersed boundary method is used to model the two enclosed cylinders, based on the
Accepted 7 August 2012
finite volume method, and two-dimensional natural convection was studied for different Rayleigh num-
Available online 30 August 2012
bers in the range 103 6 Ra 6 106 . Additionally, the study investigates the effect the presence of the hot
and cold cylinders at different locations within the enclosure have on the heat transfer and fluid flow.
Keywords:
The locations of the hot and cold circular cylinders are varied vertically along the centerline at the left
Natural convection
Hot and cold inner cylinders
and right parts of the enclosure, respectively. The existence of local peaks for the Nusselt numbers along
Square enclosure the surfaces of the cylinders and enclosure is determined by the gap between the cylinders and the enclo-
Nusselt number sure and thermal plume governed by the convection, respectively. Detailed analysis results for the distri-
butions of the streamlines, isotherms, and Nusselt numbers are presented in this paper.
Ó 2012 Elsevier Ltd. All rights reserved.

1. Introduction
The heat transfer and flow characteristics of the natural convec-
tion in an enclosure has many industrial and environmental appli-
cations, including heat exchangers, nuclear and chemical reactors,
electronic equipment cooling, and stratified atmosphere boundary
layers. The geometries that arise in engineering applications are
more complicated than a simple enclosure filled with a convective
fluid. The geometric configuration of interest is coupled with the
presence of bodies embedded within the enclosure. This study
investigated the effect the presence of the hot and cold cylinders
at different locations in the enclosure had on the heat transfer
and fluid flow.
For several decades, research has been conducted on the effect
the presence of a body under different thermal conditions within
an enclosure has on natural convection. In these studies, the ther-
mal boundary condition applied to the square enclosure has been
either a horizontally imposed temperature difference or a heat flux
condition [1–5], or a vertically imposed temperature difference or
a heat flux condition [6–11].
⇑ Corresponding author. Tel.: +82 51 510 2440; fax: +82 51 515 3101.
E-mail addresses: pyg777@pusan.ac.kr (Y.G. Park), lesmodel@pusan.ac.kr
(H.S. Yoon), myha@pusan.ac.kr (M.Y. Ha).
1 Reymond et al. [15] carried out an experimental investigation
Fax: +82 51 515 3101.
2
Fax: +82 51 581 3718.
Kim et al. [12] and Yoon et al. [13] carried out numerical inves-
tigations on the natural convection induced by the temperature
difference 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 cell strongly depended
on the Rayleigh number and position of the inner circular cylinder
within the enclosure. At Ra = 107, the bifurcation of natural convec-
tion from the unsteady to steady state depended on d. They
showed that the flow and heat transfer at Ra = 107 became unstea-
dy at d 6 dC;L and d P dC;U , where dC,L = 0.05 and dC,U = 0.18.
Lee et al. [14] conducted a numerical investigation on the natu-
ral convection within a square enclosure containing a circular cyl-
inder placed at different horizontal and diagonal locations for
different Rayleigh numbers varying in the range 103 6 Ra 6 106 .
Regardless of the Rayleigh number, when the inner cylinder was
placed closer to the side wall or corner of the enclosure by a hori-
zontal or diagonal movement, respectively, the two inner eddies
were separated and located at the upper and lower parts of the
enclosure in the narrow region facing the cylinder movement.
Otherwise, one large circulation was formed in the opposite region.
The natural convection around an array of cylinders is different
from those around a single cylinder because of the mutual interac-
tion of the buoyant plumes generated by the cylinders.
on the natural convection heat transfer from a pair of vertically

0017-9310/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ijheatmasstransfer.2012.08.012
7912 Y.G. Park et al. / International Journal of Heat and Mass Transfer 55 (2012) 7911–7925

Nomenclature
  
@h
Nu local Nusselt number Nu ¼ @n wall
Tc cold temperature
Nuhotcyl local Nusselt number along the hot inner circular cylin- ui dimensionless velocity
der xi dimensionless Cartesian coordinates
Nucoldcyl local Nusselt number along the cold inner circular cylin- fi momentum forcing
der g gravity
Nuen local Nusselt number along the walls  of theR enclosure
 L length of square enclosure
1 W
Nu surface-averaged Nusselt number Nu ¼ W 0 NudS n normal direction to the wall
NuT surface-averaged Nusselt number along the top wall of W surface area of the wall
the enclosure
NuB surface-averaged Nusselt number along the bottom wall Greek symbols
of the enclosure a thermal diffusivity
NuR surface-averaged Nusselt number along the right wall of b thermal expansion coefficient
the enclosure d vertical distance from center of square cylinder to circu-
NuL surface-averaged Nusselt number along the left wall of lar cylinder center
the enclosure di2 Kronecker delta
Nuen surface-averaged Nusselt number along the square q density
enclosure m kinematic viscosity
P dimensionless pressure u angle from the top of circular cylinder
Pr Prandtl number (Pr ¼ ma) h dimensionless temperature
R radius of circular cylinder
 
3
Ra Rayleigh number Ra ¼ gbL ðTmah T c Þ Sub/Superscripts
⁄ dimensional value
t dimensionless time
- surface-averaged quantity
T dimensional temperature
Th hot temperature

aligned horizontal cylinders for different Rayleigh numbers tion caused by the temperature difference has on the fluid flow
(2  106, 4  106, and 6  106) and for different lengths of cylinder and heat transfer in the enclosure in the presence of the cylinders.
spacing (1.5, 2, and 3 cylinder diameters). Their results showed
that the fluid flow and heat transfer around the lower cylinder
were unaffected by the presence of an unheated lower cylinder, 2. Numerical methodology
and those around the heated upper cylinder were likewise unaf-
fected by the presence of the unheated lower cylinder. However, Fig. 1 shows the computational domain and coordinate system
when both the lower and upper cylinders were heated, the fluid along with the boundary conditions considered in the present
flow and heat transfer around the upper heated cylinder was study. The system consisted of a square enclosure with sides of
strongly affected by the presence of the heated lower cylinder be- length L, within which hot and cold cylinders with radius R
cause a plume rising from the heated lower cylinder interacted
with the heated upper cylinder.
Chae and Chung [16] carried out an experimental investigation
on 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 from 7.3  107 to 4.5  1010. They measured the mass
transfer rate from the cylinders and obtained the heat transfer rate
(Nusselt number) based upon the analogy concept. Their results
showed that the Nusselt number ratios of the upper to lower cyl-
inders that 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 Nusselt 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 natural convection in
the absence or presence of a body within an enclosure, few studies
have considered natural convection in the presence of a pair of cyl-
inders within an enclosure. In the present study, hot and cold cyl-
inders were placed at different locations within a cold enclosure to
investigate the effect the locations of the hot and cold cylinders
within the enclosure has on the fluid flow and heat transfer be-
cause a plume rising from the hot cylinder interacts with the cold
cylinder and enclosure. We also considered different Rayleigh Fig. 1. Computational domain and coordinate system along with boundary
numbers to investigate the effect the buoyancy-induced convec- conditions.
 Y.G. Park et al. / International Journal of Heat and Mass Transfer 55 (2012) 7911–7925 7913

(=0.2L) were located at x = 0.25L and x = 0.25L, respectively, and Once the velocity and temperature fields were obtained, we cal-
moved along the vertical centerline from y = 0.25L to y = 0.25L. culated the local and surface-averaged Nusselt numbers, which are
The walls of the square enclosure were kept at a constant low tem- defined as
perature of Tc, and the left hot and right cold cylinders within the
enclosure were kept at constant high temperatures of Th and Tc,
respectively.  Z
@h  1 W
In this study, we assumed that the radiation effects were negli- Nu ¼ ; Nu ¼ Nu dS ð5Þ
gible. The fluid properties were assumed to be constant, except for @n wall W 0

the density in the buoyancy term, which followed the Boussinesq

approximation. The gravitational acceleration acted in the negative
y-direction. The immersed boundary method was used to effi- where n is the normal direction with respect to the walls and W is
ciently handle the hot and cold circular cylinders within the enclo- the surface area of the walls.
sure, as shown in Fig. 1. The governing equations describing the Fig. 2 shows the computational geometry in the x–y plane with
unsteady incompressible viscous flow and thermal fields in the a uniform grid distribution. A grid resolution of 301  301 in the
present study were the continuity, momentum, and energy equa- horizontal (x) and vertical (y) directions was employed in our com-
tions in their non-dimensional forms, which can be defined as putations. The grid independence of the solution was tested with
@ui additional simulations on much finer grids up to 401ðxÞ  401ðyÞ
q¼0 ð1Þ points. The difference in the Nusselt number results obtained using
@xi
the coarse and fine grids was less than 0.7%.
@ui @ui @P @ 2 ui For the purpose of code validation, we solved the natural con-
þ uj ¼ þ Pr þ RaPrhdi2 þ fi ð2Þ
@t @xj @xi @xj @xj vection problem, in which a hot cylinder is located at the center
of the cold square enclosure, and compared the surface-averaged
@h @h @2h Nusselt number of the hot cylinder with that obtained by Kim
þ uj ¼ þh ð3Þ
@t @xj @xj @xj et al. [12]. As shown in Table 1, the surface-averaged Nusselt num-
The dimensionless variables in the above equations are defined as bers obtained from the present calculation are in good agreement
with those obtained from the numerical simulation carried out by
t a xi ui L P  L2 T  Tc Kim et al. [12].
t¼ ; xi ¼ ; ui ¼ ; P¼ ; h¼ ð4Þ
L 2 L a qa2 Th  Tc
In the above equations, q, T, and a represent the density, dimen-
sional temperature, and thermal diffusivity, respectively. The super-
script ⁄ in Eq. (4) represents the dimensional variables. xi are
Cartesian coordinates, ui are the corresponding velocity compo-
nents, t is the time, p is the pressure, and h is the temperature.
The above non-dimensionalization results in two dimensionless
3
parameters: Pr ¼ ma and Ra ¼ gbL ðTmah T c Þ, where m, g, and b are the
kinematic viscosity, gravitational acceleration, and volume expan-
sion coefficient, respectively. In the simulations to be reported
here, the Prandtl number Pr, and r(=R/L) have been taken to be
0.7 (corresponding to air) and 0.2, respectively. The Rayleigh num-
ber, Ra, varies in the range 103–106. The dimensionless vertical
distance d, which represents the position of the two circular cylin-
ders along the vertical centerline, varies in the range 0.25–0.25.
The mass source/sink q in Eq. (1) and momentum forcing fi in
Eqs. (1) and (2) are applied on the body surface or inside the body
to satisfy the no-slip condition and mass conservation in the cell
containing the immersed boundary. In Eq. (3), the heat source/sink
h is applied to satisfy the isothermal boundary condition on the
immersed boundary.
A two-step time-split scheme was used to advance the flow
field. This scheme was based on the previous works of Kim and
Moin [17] and Zang et al. [18]. The temperature field was advanced
in a similar manner by using a second-order Adams–Bashforth Fig. 2. Typical grid distribution for d = 0.
scheme for the advection term and a Crank–Nicolson scheme for
the diffusion term. The central difference scheme with second-
order accuracy based on the finite volume method was used for
the spatial discretization. Additionally, a second-order linear or
Table 1
bilinear interpolation scheme was applied to satisfy the no-slip Comparison of surface-averaged Nusselt numbers obtained from this study with
and isothermal conditions on the immersed boundary. Further those obtained from a previous numerical study.
details on the immersed-boundary method are provided in Kim
Ra Mean Nusselt number at hot wall
et al. [19] and Kim and Choi [20].
For the velocity boundary conditions, no-slip and no- Present study Kim et al. [12] Difference (%)

penetration boundary conditions were imposed at the walls. For 103 5.107 5.093 1.492
the thermal boundary conditions, h = 0 was imposed at the walls 104 5.128 5.108 0.392
105 7.836 7.767 0.888
of the enclosure and cold inner cylinder, whereas h = 1 was used
106 14.462 14.11 2.495
for the wall of the hot inner cylinder.
7914 Y.G. Park et al. / International Journal of Heat and Mass Transfer 55 (2012) 7911–7925
3. Results and discussion
Fig. 3 shows the distributions of isotherms and streamlines for
four different Rayleigh numbers (103, 104, 105, and 106) when
d = 0. For all Rayleigh numbers considered in this study, the flow
and thermal fields eventually reach steady state when d = 0. Be-
cause the hot and cold cylinders are located at the left and right
sides of the enclosure, respectively, a plume rising from the hot
cylinder interacts with the cold cylinder and square enclosure. As
a result, small counterclockwise eddies form at the left side,
whereas large clockwise eddies form at the right side of the
enclosure.
When Ra = 103, the heat transfer in the enclosure is mainly
dominated by the conduction mode. Thus, the isotherms and
streamlines are almost symmetric about the horizontal centerline
at y = 0, as shown in Fig. 3(a). Because the hot and cold cylinders
are located at the left and right sides of the enclosure, respectively,
the isotherm distribution only surrounds the hot cylinder. This
gives a larger thermal gradient at the left and right sides of the
hot cylinder because of the relatively smaller spaces available
and a smaller thermal gradient at the top and bottom sides of
the hot cylinder because of relatively larger spaces available. As a
result, counterclockwise and clockwise eddies are formed at the
left and right parts of the enclosure, respectively. The size of the
counterclockwise eddies at the left side of the enclosure is much
smaller than that of the clockwise eddies at the right side of the
enclosure because of the presence of the hot cylinder at the left
part of the enclosure. Because the gap between the hot cylinder
and left wall of the enclosure is small, two counterclockwise eddies
formed at the left upper and lower parts of the enclosure. Because
the cold cylinder is located at the right part of the enclosure and
the distance between the hot cylinder and right wall of the enclo-
sure is large, the primary clockwise eddy circulates around the cold
cylinder and two secondary clockwise eddies embedded in the pri-
mary clockwise eddy are formed between the hot and cold
cylinders.
When Ra = 104, the conduction is still a dominant heat transfer
mode in the enclosure because of the small effect of convection on
Fig. 3. Isotherms and streamlines at d = 0 for four different Rayleigh numbers: (a) 103, (b) 104, (c) 105, and (d) 106.
the fluid flow and heat transfer. Thus, the distributions of the iso-
therms and streamlines for Ra = 104, as shown in Fig. 3(b), are gen-
erally similar to those for Ra = 103, as shown in Fig. 3(a).
When Ra = 105, a plume rising from the hot cylinder interacts
with the cold cylinder and cold square enclosure. As a result, con-
vection plays an important role in the distributions of the fluid
flow and heat transfer in the enclosure, in addition to conduction.
A thermal plume is formed at the top of the hot cylinder and slants
to the right because of the greater amount of space on the right
side of the enclosure. As a result, the sizes of the counterclockwise
eddy and secondary clockwise eddy embedded in the primary
clockwise eddy at the upper part of the enclosure become larger
than those at the lower part of the enclosure, as shown in
Fig. 3(c). The thermal gradient at the bottom part of the hot cylin-
der and top wall of the enclosure for Ra = 105 is larger than that for
Ra = 103 and 104 because of the increasing effect of convection
with the increasing Rayleigh number.
When Ra = 106, the strong plume rising from the hot cylinder
impinges on the top of the enclosure and slants more toward the
upper right part of the enclosure, with a strong increase in the ef-
fect of convection on the fluid flow and heat transfer in the enclo-
sure, compared to that when Ra = 105. As a result, the sizes of the
counterclockwise eddy and secondary clockwise eddy embedded
in the primary clockwise eddy at the upper part of the enclosure
become much larger than those at the lower part of the enclosure,
as shown in Fig. 3(d). The thermal gradient around the hot cylinder
and top wall of the enclosure for Ra = 106 becomes larger than
those for Ra = 103, 104, and 105 because of the increasing effect of
convection with the increasing Rayleigh number. However, the
thermal gradient along the bottom wall for Ra = 106 is smaller than
that for Ra = 103, 104, and 105 because of the increase in the stag-
nant region at the lower part of the enclosure with the increasing
Rayleigh number.
3.1. Ra = 103 and Ra = 104
Fig. 4 shows the distributions of the isotherms and streamlines
when the hot and cold cylinders are located at different vertical
 Y.G. Park et al. / International Journal of Heat and Mass Transfer 55 (2012) 7911–7925
Fig. 4. Isotherms and streamlines for different d values at Ra = 103 and Ra = 104 (contour values range from 0 to 1 with 10 levels).
positions at the lower and upper parts of the enclosure, along with
the corresponding negative and positive values of d, respectively,
when Ra = 103 and Ra = 104. Because the heat transfer in the enclo-
sure is mainly dominated by the conduction mode at Ra = 103, the
distributions of the flow and thermal fields for different values of d
are generally similar to those for d = 0, corresponding to the case
where two cylinders are located at the horizontal centerline of
the enclosure, as shown in Fig. 3(a). However, when the two cylin-
ders move downward with d < 0, the size of the two vortices at the
lower part of the enclosure decreases with the decreasing space
between the two cylinders and the bottom wall of the enclosure,
whereas that at the upper part increases with the increasing space
between the two cylinders and the top wall of the enclosure, com-
pared to the equal sizes of the vortices at the lower and upper parts
of the enclosure when d = 0. When the two cylinders move upward
with d > 0, the variation in the sizes of the lower and upper vortices
is the opposite to that when d < 0, resulting in a decreasing size of
the upper vortices and increasing size of the lower vortices. As a re-
sult, when d < 0, the value of the thermal gradient around the hot
cylinder is high at the bottom side of the cylinder and low at the
top side of the cylinder. However, when d > 0, the value of the ther-
mal gradient around the hot cylinder is opposite to that when d < 0,
resulting in a high value at the top side of the hot cylinder and low
value at the bottom side of the hot cylinder. Because of the conduc-
tion dominance in the heat transfer at Ra = 103, the distributions of
the thermal and flow fields for d > 0 shown in Fig. 4(b) are symmet-
ric to those for d < 0 shown in Fig. 4(a), with respect to the horizon-
tal center line at y = 0.
When Ra = 104, the effect of convection on the fluid flow and
heat transfer is still small, and as a result, the distributions of the
isotherms and streamlines for Ra = 104 are generally similar to
those for Ra = 103. However, because of the slightly increasing ef-
fect of convection on the fluid flow and heat transfer as the Ray-
leigh number increases to 104 from 103, the distributions of the
flow and thermal fields for d < 0 shown in Fig. 4(c) are not symmet-
ric to those for d > 0 shown in Fig. 4(d), with respect to the horizon-
7915
tal center line at y = 0 when Ra = 104, unlike to the symmetric ones
when Ra = 103 .
Fig. 5 shows the distribution of the local Nusselt numbers along
the surface of the left hot cylinder (Nuhot cyl ), right cold cylinder
(Nucold cyl ) and walls of the enclosure (Nuen) for different vertical
positions of the two cylinders when Ra = 103. The distributions of
Nuhot cyl , Nucold cyl , and Nuen depend on the distributions of the iso-
therms shown in Figs. 3 and 4. When Ra = 103, the heat transfer
in the enclosure is mainly dominated by the conduction mode.
For all values of d at Ra = 103, the values of Nuhotcyl have two local
high peaks at around / = 90° and / = 270° for the hot cylinder be-
cause the distance between the left hot and right cold cylinders
and the distance between the left wall of the enclosure and left
hot cylinder are small under the dominant conduction heat trans-
fer mode, forming denser distributions of isotherms at around /
= 90° and / = 270° for the hot cylinder, as shown in Figs. 3 and 4.
Thus, when Ra = 103, the values of Nuhot cyl at around / = 90° and
/ = 270° for the hot cylinder do not depend on d but have almost
the same values for different values of d. However, the values of
Nuhot cyl at around / = 0° and / = 180° strongly depend on d because
the space available for heat transfer between the hot cylinder and
bottom and top walls of the enclosure varies according to d. Thus,
as d decreases at d 6 0, the distance between the hot cylinder and
top wall of the enclosure increases, whereas the distance between
the hot cylinder and bottom wall of the enclosure decreases,
resulting in a significant increase in the value of Nuhot cyl at around
/ = 180° and a slight decrease in Nuhot cyl at around / = 0°. When
d = 0.25, another high peak for Nuhot cyl at around / = 180° is
formed, in addition to two local high peaks at around / = 90° and
/ = 270° where the value of Nuhot cyl at around / = 180° is similar
to that at / = 270° However, as d increases at d P 0, the distance
between the hot cylinder and top wall of the enclosure decreases,
whereas the distance between the hot cylinder and bottom wall of
the enclosure increases, resulting in a slight decrease in the value
of Nuhot cyl at around / = 180° and a significant increase in Nuhot cyl
at around / = 0°. When d = 0.25, another high peak for Nuhot cyl at
7916 Y.G. Park et al. / International Journal of Heat and Mass Transfer 55 (2012) 7911–7925

Fig. 5. Distributions of local Nusselt numbers along (a) surface of hot inner cylinder ðNuhot cyl Þ, (b) surface of cold inner cylinder ðNucold cyl Þ, and (c) walls of enclosure (Nuen) for
different d values at Ra = 103.

around / = 0° also forms, in addition to two local high peaks at
around / = 90° and / = 270°, where the value of Nuhot cyl at around
/ = 0° is also similar to that at / = 270°.
Unlike the distribution of Nuhot cyl , the distribution of Nucold cyl has
only one local high peak value at around / = 270°, regardless of the
value of d, because the distance between the hot and cold cylinders
is small at around / = 270° under the dominant conduction heat
transfer mode at Ra = 103, and as a result, a relatively denser distri-
bution of isotherms forms at around / = 270°. The variation in
Nucold cyl according to d is much smaller than that in Nuhot cyl because
the temperature of the cold cylinder is the same as that of the cold
enclosure. The values of Nucold cyl decrease slightly in the region of
180° < / < 270° with decreasing d at d 6 0 or in the region of
270° < / < 360° with increasing d at d P 0 because of the decreas-
ing distance between the cylinders and bottom wall of the enclo-
sure or the decreasing distance between the cylinders and top
wall of the enclosure. However, the values of Nucold cyl in the region
of 270° < / < 360° at d 6 0 and in the region of 180° < / < 270° at
d P 0:0 do not depend on the variation in d. The values of
Nucold cyl at around 0° < / < 180° for all the values of d are almost
zero because the left side of the cold cylinder corresponding to
0° < / < 180° is exposed to the right cold wall of the enclosure.
The distribution of Nuen also depends strongly on the position of
the cylinders in the enclosure in the conduction dominant regime
of Ra = 103. In the region from point A to point B, along the top wall
of the enclosure, as d increases from d = 0.25 to d = 0.25, the
 Y.G. Park et al. / International Journal of Heat and Mass Transfer 55 (2012) 7911–7925 7917

Fig. 6. Distributions of local Nusselt numbers along (a) surface of hot inner cylinder ðNuhot cyl Þ, (b) surface of cold inner cylinder ðNucold cyl Þ, and (c) walls of enclosure (Nuen) for
different d values at Ra = 104.

distance between the two cylinders and top wall decreases, and as
a result, the values of Nuen increase. In the region from point B to
point C along the right wall of the enclosure, the distribution of
Nuen is almost zero because the cold cylinder and right wall of
the enclosure have the same cold temperature. In the region from
point C to point D, along the bottom wall of the enclosure, as d in-
creases from d = 0.25 to d = 0.25, the distance between the two
cylinders and bottom wall decreases, and as a result, the values
of Nuen decrease. This is the opposite to the distribution of Nuen
in the region from point A to point B along the top wall of the
enclosure. In the region from point D to point E along the left wall
of the enclosure, the high peak values of Nuen for different d values
are almost the same because of the small distance between the hot
cylinder and left wall of the enclosure. However, the location at
which Nuen has a high peak value in the region from point D to
point E along the left wall of the enclosure depends on d and moves
from point D to point E as the hot cylinder moves upward with
increasing d.
Fig. 6 shows the distributions of the local Nusselt numbers
along the surface of the left hot cylinder (Nuhot cyl ), right cold cylin-
der (Nucold cyl ), and walls of the enclosure (Nuen) for different vertical
positions of the two cylinders when Ra = 104. When Ra = 104, the
fluid flow and heat transfer in the enclosure are still governed
mainly by the conduction mode. As a result, the distributions of
7918 Y.G. Park et al. / International Journal of Heat and Mass Transfer 55 (2012) 7911–7925
Fig. 7. Isotherms and streamlines for different d values at Ra = 105 (contour values range from 0 to 1 with 10 levels).
the isotherms and streamlines for Ra = 104 are generally similar to
those for Ra = 103, resulting in similar distributions of Nuhot cyl ,
Nucold cyl , and Nuen according to / for different values of d .
3.2. Ra = 105
Fig. 7 shows the distributions of the isotherms and streamlines
when the hot and cold cylinders are located at different vertical
positions (d values) in the enclosure when Ra = 105. The distribu-
tion of the isotherms in the enclosure at Ra = 105 is significantly
different from those at Ra = 103 and Ra = 104 because the buoy-
ancy-induced convection from the hot cylinder becomes more pre-
dominant than conduction with the increasing Rayleigh number.
When the two cylinders move downward with d < 0, more space
is secured between the two cylinders and the top wall of the enclo-
sure, which enhances the buoyancy-induced convection. Thus, the
ascending plume moves upward along the hot cylinder, giving rise
to a stronger thermal gradient at the location where the rising
plume impinges on the top wall of the enclosure. When the two in-
ner cylinders move upward with d > 0, an ascending plume rising
from the top of the hot cylinder disappears because of the space
constraint between the two cylinders and the top wall of the enclo-
sure, which becomes smaller with increasing d. When Ra = 105, the
variation in the streamlines circulating in the clockwise or counter-
clockwise direction in the enclosure according to d is generally
similar to that when Ra = 103 and Ra = 104. However, when
Ra = 105, the centers of the lower counterclockwise and clockwise
eddies at the left and middle sides of the enclosure, respectively,
move upward, compared to those when Ra = 103 and Ra = 104, be-
cause of the increasing effect of convection with the increasing
Rayleigh number. On the other hand, the centers of the upper
counterclockwise and clockwise eddies at the left and middle sides
of the enclosure, respectively, move slightly to the right when
Ra = 105, compared to those when Ra = 103 and Ra = 104, because
the ascending plume from the hot cylinder moves to the upper
right with the increasing Rayleigh number.
Fig. 8 shows the distributions of the local Nusselt numbers
along the surface of the left hot inner cylinder (Nuhot cyl ), right cold
inner cylinder (Nucold cyl ), and walls of the enclosure (Nuen) for dif-
ferent vertical positions of the two cylinders when Ra = 105. The
distributions of the local Nusselt numbers for Ra = 105 are different
from that those for Ra = 103 and 104 because of the increasing ef-
fect of convection induced by the ascending plume from the hot
cylinder with the increasing Rayleigh number, in addition to the ef-
fect of the cylinder’s position within the enclosure. For the case of
natural convection around a single hot single cylinder in the a cold
 Y.G. Park et al. / International Journal of Heat and Mass Transfer 55 (2012) 7911–7925 7919

Fig. 8. Distributions of local Nusselt numbers along (a) surface of hot inner cylinder ðNuhot cyl ), (b) surface of cold inner cylinder ðNucold cyl Þ, and (c) walls of enclosure (Nuen) for
different d values at Ra = 105.

environment, without a cold cylinder and cold enclosure, the dis-
tribution of the local Nusselt numbers with / is characterized by
a maximum at / = 180° and a monotonic decay with decreasing
/ from 180° to 0° or increasing / from 180° to 360° because of
boundary layer development, which begins at / = 180° and con-
cludes at around / = 0° (or / = 360°) with the formation of a plume
ascending from the cylinder. However, for the case of the present
natural convection around a hot cylinder located in a cold enclo-
sure in the presence of a cold cylinder, the ascending plume that
begins at around / = 180° for the hot cylinder interacts with both
the right cold cylinder and the cold enclosure. As a result, Nuhot cyl
has a local high peak at around / = 180°, decreases with decreasing
or increasing / from / = 180°, reaches local low peaks at around /
= 115–124° or / = 217–225°, depending on d, and then increases
again with decreasing / from / = 115–124° or increasing / from
/ = 217–225° until it reaches high peak values at around / = 75°
or / = 270°. At around / = 270°, the dependency of the local high
peak of Nuhot cyl on the distance between the left wall of the enclo-
sure and the hot cylinder is very strong because of the small dis-
tance, whereas its dependency on the Rayleigh number is very
weak, resulting in similar high peak values of Nuhot cyl at around
/ = 270° for different Rayleigh numbers of Ra = 103, 104, and 105.
When Ra = 103 and 104, another local high peak appears at around
/ = 90° because of the small distance between the left hot and right
cold cylinders in the dominant conduction heat transfer regime.
However, if the Rayleigh number increases to 105, the local high
7920 Y.G. Park et al. / International Journal of Heat and Mass Transfer 55 (2012) 7911–7925
peak of Nuhot cyl moves to / = 75° from / = 90° when Ra = 103 and
104 because of the increasing effect of the ascending plume along
the hot cylinder with the increasing Rayleigh number, in addition
to the gap effect between the hot and cold cylinders. The distribu-
tion of Nuhot cyl at 75° < / < 270° for Ra = 105 is similar to that for
Ra = 103 and 104, resulting in increasing values of Nuhot cyl with
decreasing d at d 6 0 because of the decreasing distance between
the hot cylinder and bottom wall of the enclosure, along with a
small variation in Nuhot cyl with increasing d at d P 0 because of
the increasing stagnant region with the increasing distance be-
tween the hot cylinder and bottom wall of the enclosure. The val-
ues of Nuhot cyl at d 6 0 for Ra = 105 keep decreasing as / decreases
from / = 75° to 0° or increases from / = 270° to 360°. As d increases
at d 6 0 for Ra = 105, the values of Nuhot cyl decrease slightly in the
region 0° < / < 75° and show a small variation in the region
270° < / < 360° because the effects of the ascending plume from
the hot cylinder on Nuhot cyl in these regions are stronger than the
effect of the gap between the hot cylinder and top wall of the
enclosure, as shown in the distribution of the isotherms in Fig. 7.
When d = 0.1 for Ra = 105, the effect of the ascending plume from
the hot cylinder on Nuhot cyl is still strong, resulting in a decreasing
Nuhot cyl with decreasing / from / = 75° to / = 0° or increasing /
from / = 270° to / = 360°. However, when d = 0.2 or 0.25 for
Ra = 105, the ascending plume from the hot cylinder disappears
and the effect of the gap between the cylinders and the top wall
of the enclosure becomes much stronger, resulting in a significant
increase in the value of Nuhot cyl with decreasing / from / = 42–51°
to / = 0° or increasing / from / = 317–327° to / = 360°.
The distribution of Nucold cyl according to / for different values of
d at Ra = 105 is generally similar to that at Ra = 103 and 104, except
for the fact that the variation in Nucold cyl according to d at Ra = 105 is
slightly larger than that at Ra = 103 and 104 because of the increas-
ing effect of convection.
The distributions of Nuen along the enclosure wall for different
values of d at Ra = 105 are different from those at Ra = 103 and
104 because of increasing convection effect with increasing Ray-
leigh number. When we move from point A to point B along the
top wall of the enclosure at Ra = 104, the values of Nuen increase
with the decreasing distance between the cylinders and top wall
of the enclosure as the cylinders move upward with increasing d.
However, when the Rayleigh number increases to 105, the values
of Nuen at the top wall of the enclosure at d 6 0 for Ra = 105 are
higher than those for Ra = 104 because of increasing convection in-
duced by the ascending plume from the hot cylinder and increase
with increasing d at d 6 0, similar to that at Ra = 104. When d = 0.1
at Ra = 105, the peak value of Nuen at the top wall of the enclosure is
slightly smaller than that at d = 0.0 because of the combined effects
of convection induced by the ascending plume from the hot
cylinder and the distance between the cylinders and top wall of
the enclosure. However, as d increases to 0.2 and 0.25 at
Ra = 105, the effect of the distance between the cylinders and top
wall of the enclosure becomes greater than the effect of convec-
tion, resulting in higher values of Nuen compared to those at
d 6 0:1, which result in variations in Nuen along the top wall of
the enclosure at d = 0.2 and 0.25 for Ra = 105 that are similar to
those for Ra = 104 .
When we move from point B to point C along the right wall of
the enclosure at Ra = 105, the values of Nuen along the right wall
of the enclosure are very small because the hot cylinder is located
at the left part of the enclosure, similar to those when Ra = 104.
However, the values of Nuen at the upper part of the right wall of
the enclosure at Ra = 105 are slightly higher than those at
Ra = 104 because of the convection effect induced by the ascending
plume from the hot cylinder.
When d = 0.2 and d = 0.25 in the region from point C to point
D along the bottom wall of the enclosure, the effect of the gap
between the cylinders and bottom wall of the enclosure becomes
more dominant than the convection effect induced by the ascend-
ing plume. As a result, the distribution of Nuen for Ra = 105 is
similar to that for Ra = 104. However, when d P 0:1 for Ra = 105,
the stagnant region between the hot cylinder and bottom wall of
the enclosure increases as a result of the increasing effect of con-
vection with the increasing Rayleigh number, as shown in the dis-
tributions of the isotherms in Figs. 4 and 7. Thus, the values of Nuen
along the bottom wall of the enclosure at d = 0.1 and d = 0 for
Ra = 105 are lower than those for Ra = 104 and those at d P 0:1
for Ra = 105 are almost zero.
When we move from point D to point E along the left wall of the
enclosure, the effect of the gap between the hot cylinder and left
wall of the enclosure on the distribution of Nuen along the left wall
of the enclosure becomes dominant because of the small distance
between the hot cylinder and left wall of the enclosure. Thus, the
distribution of Nuen along the left wall of the enclosure for
Ra = 105 is similar to that for Ra = 103 and 104; the peak values of
Nuen at the left wall of the enclosure for different values of d for
Ra = 105 are almost the same as those for Ra = 103 and 104, and
the location at which Nuen at the left wall of the enclosure has a
peak value moves upward from point D to point E along the left
wall of the enclosure as d increases. However, when d 6 0 at
Ra = 105, we can observe another peak for Nuen at the upper part
of the left wall of the enclosure because of the presence of the
ascending plume from the hot cylinder, unlike the monotonic de-
cay of Nuen from the high peak to the very low value at point E
when Ra = 103 and 104 .
3.3. Ra = 106
Fig. 9 shows the distributions of the isotherms and streamlines
when the hot and cold cylinders are located at different vertical
positions (d values) in the enclosure when Ra = 106. When the Ray-
leigh number increases to Ra = 106, the magnitude of the velocity
circulating in the enclosure increases and the isotherms are dis-
torted more as a result of the stronger convection effects. As a re-
sult, the thickness of the thermal boundary formed at the hot inner
cylinder and the enclosure wall becomes thinner, giving a larger
thermal gradient. The distributions of the isotherms and stream-
lines at Ra = 106 are significantly different from those at Ra = 105
because of the greater increase in the effect of convection with
the increasing Rayleigh number. When d 6 0:1 for Ra = 106, a single
ascending plume is formed around the hot cylinder. However,
when d = 0.2 and 0.25 for Ra = 106, two ascending plumes and
one descending plume are formed between the hot cylinder and
top wall of the enclosure, as shown in Figs. 9(e) and (f), because
of the decreasing space between the two cylinders and top wall
of the enclosure with increasing d. When Ra = 106, the isothermal
lines extend more to the upper right corner because of stronger
convection at all d values, compared to those at the lower Rayleigh
numbers of 103, 104, and 105. At d ¼ 0:2 for Ra = 106, the lower
counterclockwise and clockwise vortices at the left and middle of
the enclosure, are each separated into two vortices. However, the
divided lower eddies at the left of the enclosure merge again into
a single eddy at d = 0.25 for Ra = 106 .
Fig. 10 shows the distributions of the local Nusselt numbers
along the surfaces of the left hot inner cylinder (Nuhot cyl ), right cold
inner cylinder (Nucold cyl ), and walls of the enclosure (Nuen) for dif-
ferent vertical positions of the two cylinders at Ra = 106. As the
Rayleigh number increases to 106, the effect of convection becomes
much larger than that at the lower Rayleigh numbers of Ra = 103,
104, and 105. When Ra = 105, the hot cylinder has a stagnation
point at around / = 180°, and isotherms are separated from the
hot cylinder at around / = 0°. However, if the Rayleigh number in-
creases to 106 from 105, the lower stagnation point moves to
 Y.G. Park et al. / International Journal of Heat and Mass Transfer 55 (2012) 7911–7925
Fig. 9. Isotherms and streamlines for different d values at Ra = 106 (contour values range from 0 to 1 with 10 levels).
around / = 158–168°, depending on d, as a result of the interaction
of the ascending plume with the cold cylinder and cold enclosure,
and the ascending plume is separated from the hot cylinder at
around / = 10–12° because the isotherms are slanted to the right
and extended to the upper right corner as a result of strong convec-
tion. As a result, when Ra = 106, Nuhot cyl has a local high peak at
around / = 158–168°, depending on d .
When d 6 0 at Ra = 106, as / decreases or increases along the
right or left side of the hot cylinder from / = 158–168° (lower stag-
nation point), Nuhot cyl decreases from / = 158–168° to / = 90–93°
or to / = 235–238°, depending on d, because of the increasing ther-
mal boundary layer thickness as the plume moves upward along
the hot cylinder surface. Similarly, when d P 0 at Ra = 106, as / de-
creases or increases from / = 168° (lower stagnation point), Nuhot cyl
decreases from / = 168° to / = 91–94° or to / = 235° because of the
increasing thermal boundary layer thickness. Because the effect of
convection on Nuhot cyl is stronger than the effect of the cylinder po-
sition in the enclosure at Ra = 106, in the region from / = 90–94° to
/ = 235–238°, depending on d, the value of Nuhot cyl for Ra = 106 is
much higher than that for Ra = 105 at 0:25 6 d 6 0:2, and the var-
iation in the peak value of Nuhot cyl for Ra = 106, according to d, is
smaller than that for Ra = 105.
When d 6 0 at Ra = 106, Nuhot cyl increases from / = 90–93° to /
= 71–73° because of the interaction between the hot and cold
7921
cylinders or from / = 235–238° to / = 280° because of the interac-
tion between the hot cylinder and left wall of the enclosure, show-
ing a small variation in the values of Nuhot cyl in this region
according to d. For a similar reason, when d P 0 at Ra = 106,
Nuhot cyl increases from / = 91–94° to / = 70–76° or from / = 235°
to / = 280°, showing a small variation in the values of Nuhot cyl
according to d .
When 0:25 6 d 6 0:1, Nuhot cyl decreases again from / = 71–73°
to /=10–12° (corresponding to the separation point of the ascend-
ing plume from the hot cylinder) as a result of the further develop-
ment of the thermal boundary layer of the ascending thermal
plume along the right side of the hot cylinder, showing a small var-
iation in the values of Nuhot cyl according to d because of the strong
effect of convection on Nuhot cyl , followed by a small increase in
Nuhot cyl from / = 10–12° to / = 0°. However, when d = 0.2 or 0.25
at Ra = 106, two ascending plumes and one descending plume exist
on the top of the hot cylinder, as shown in Fig. 9. As a result, when
d = 0.2 or 0.25, Nuhot cyl decreases from / = 76° to / = 35–44° (corre-
sponding to the separation point of the ascending plume from the
hot cylinder), followed by a large increase in Nuhot cyl as a result of
the strong effect of the distance between the hot cylinder and top
wall of the enclosure, especially at d = 0.25.
When 0:25 6 d 6 0:1, Nuhot cyl decreases from / = 280° to /
= 317–323° owing to further development of the thermal boundary
7922 Y.G. Park et al. / International Journal of Heat and Mass Transfer 55 (2012) 7911–7925

Fig. 10. Distributions of local Nusselt numbers along (a) surface of hot inner cylinder ðNuhot cyl Þ, (b) surface of cold inner cylinder ðNucold cyl Þ, and (c) walls of enclosure (Nuen) for
different d values at Ra = 106.

layer of the ascending thermal plume along the left side of the hot
cylinder, increases from / = 317–323° to / = 331–335° because the
flow is accelerated in this region by isotherms slanted toward the
upper right part, and decreases again from / = 331–335° to /
= 360° because the isotherms are separated from the hot cylinder
with the increasing boundary layer thickness. However, when
d = 0.2 or 0.25, the effect of the distance between the hot cylinder
and top wall of the enclosure becomes much larger, and as a result,
Nuhot cyl decreases from / = 280° to / = 326–347°, which is followed
by a large increase in Nuhot cyl , especially at d = 0.25.
When Ra = 106, the isotherms slanting toward the right are ex-
tended significantly to the upper right corner because of the strong
effect of convection, as shown in Fig. 9. As a result, the distribution
of Nucold cyl according to / for different values of d at Ra = 106 is
significantly different from that at Ra = 105 because of the strong
effect of the convection on the distributions of the isotherms and
on Nucold cyl around the cold cylinder, as shown in Fig. 8(b) and
Fig. 10(b), respectively. When d 6 0 for Ra = 106, the value of
Nucold cyl has a high peak at around / = 260° and decreases with
increasing / from / = 260° to / = 305–317° and with decreasing
/ from / = 260° to / = 180°, similar to the variation in Nucold cyl in
the range from / = 180° to / = 318°/325° for Ra = 105. Nucold cyl
increases as / increases from / = 305–317° to / = 360° at d 6 0
for Ra = 106 because the ascending plume from the hot cylinder
 Y.G. Park et al. / International Journal of Heat and Mass Transfer 55 (2012) 7911–7925
is slanted more toward the upper right part of the enclosure as a
result of stronger effect of convection, compared with the almost
constant small value of Nucold cyl in the range from / = 318–325°
to / = 360° at d 6 0 for Ra = 105. Because the isotherms are slanted
more toward the upper right part of the enclosure with the
increasing Rayleigh number, Nucold cyl in the region of / = 180–
360° at d 6 0 for Ra = 106 increases with increasing d from
d = 0.25 to d = 0, compared with Nucold cyl in the region from /
= 260–265° to / = 318–325° at Ra = 105, which does not depend
on the variation in d owing to the distribution of the isotherms
shown in Fig. 7. Nucold cyl in the region from / = 120° to / = 180°
at d 6 0 for Ra = 106 is almost zero because the cold cylinder is ex-
posed to the cold enclosure in this region, similar to Nucold cyl in the
region from / = 60° to / = 180° at d 6 0 for Ra = 105. Nucold cyl at
d 6 0 for Ra = 106 decreases with increasing / from / = 0° to /
= 120° and with increasing d from d = 0.25 to d = 0.0, for a reason
similar to that for the region from / = 305–317° to / = 360° at
d 6 0 for Ra = 106, compared with the small value of Nucold cyl , which
is independent of the variation in d, in the region from / = 0° to /
= 60° at d 6 0 for Ra = 105 .
When d P 0 for Ra = 105, Nucold cyl has a peak at / = 260°, de-
creases with increasing or decreasing / from / = 260°, and be-
comes almost zero from / = 60 to / = 120. The variation in
Nucold cyl according to d at d P 0 for Ra = 105 is small, except for
the top part of the cold cylinder at around / = 360°. When d P 0
for Ra = 106, Nucold cyl has a peak at / = 258–261° and decreases with
decreasing / from / = 258–261° to / = 120° or increasing / from /
= 258–261° to / = 287–305°, with a small variation in Nucold cyl
according to d, similar to that for Ra = 105. However, when d P 0
for Ra = 106, Nucold cyl decreases with increasing / from / = 0° to /
= 120° and decreases with increasing d from d = 0 to d = 0.25 be-
cause the smaller gap between the cold cylinder and top wall of
the enclosure with increasing d blocks the movement of the
ascending plume through this gap. In the region from / = 287–
305° to / = 360° at d P 0 for Ra = 106, Nucold cyl has a large variation
according to d because of the strong interaction between the cold
cylinder and ascending plume from the hot cylinder. As a result,
Nucold cyl at d P 0 for Ra = 106 increases in the region from /
= 287–305° to / = 313–351°, which is followed by a decrease in
Nucold cyl in the region from / = 313–351° to / = 360°.
When we move from point A and B along the top wall of the
enclosure at d 6 0, the distributions of Nuen along the top wall of
the enclosure for Ra = 106 are similar to those for Ra = 105 for dif-
ferent d values. However, the values of Nuen along the top wall of
the enclosure at d 6 0 for Ra = 106 are higher than those for
Ra = 105 because of increasing convection with the increasing Ray-
leigh number. When we move from point B and C along the right
wall of the enclosure at d 6 0, the distribution of Nuen for
Ra = 106 is also similar to that for Ra = 105, except the values of
Nuen on the right wall of the enclosure at d 6 0 for Ra = 106 are
higher than those for Ra = 105. When we move from point C and
D along the bottom wall of the enclosure, the values of Nuen along
the bottom wall of the enclosure at d = 0, 0.1, and 0.2 for
Ra = 106 are lower than those for Ra = 105 because of the increasing
stagnant region between the cylinders and bottom wall of the
enclosure with the increasing Rayleigh number, which is especially
close to a zero value for Nuen at d = 0 and 0.1. However, the distri-
bution of Nuen at d = 0.25 for Ra = 106 is similar to that for Ra = 105
because of the stronger effect of the gap between the cylinders and
bottom wall of the enclosure compared to the convection effect.
When the cylinders move upward, the distribution of Nuen for
Ra = 106 is different from that for Ra = 105, especially at the top
and left walls of the enclosure. When we move from point A and
B along the top wall of the enclosure at d P 0, the values of Nuen
for different values of d at Ra = 106 are higher than those at
Ra = 105 because of the strong effect of the convection with the
7923
increasing Rayleigh number, except for that at d = 0.25, which gives
almost the same distribution and value of Nuen for Ra = 106 as for
Ra = 105 because of the dominant effect of the gap between the cyl-
inders and top wall of the enclosure. When Ra = 106, at d = 0 or 0.1,
Nuen only has one high peak at the position where the ascending
plume impinges on the top wall of the enclosure, whereas at
d = 0.2 and 0.25, it has two high peaks because of the presence of
two ascending plumes from the top part of the hot cylinder, as
shown in Fig. 9. When we move from point B and C along the right
wall of the enclosure at d P 0, the values of Nuen for Ra = 106 are
higher than those for Ra = 105, which are almost zero, and decrease
with increasing d because the decreasing gap distance between the
cold cylinder and top wall of the enclosure blocks the movement of
isotherms through this gap. When we move from point C and D
along the bottom wall of the enclosure at d P 0, the values of Nuen
for Ra = 106 are almost zero because of the increasing stagnant re-
gion between the cylinders and bottom wall of the enclosure with
increasing d, similar to that for Ra = 105.
When we move from point D and E along the left wall of the
enclosure at d P 0, the effect of the gap distance between the
hot cylinder and left wall of the enclosure becomes stronger. As a
result, the distribution of Nuen for Ra = 106 is generally similar to
that for Ra = 105, except for slightly decreasing peak values of Nuen
for Ra = 106 with increasing d, compared to the almost constant
peak values of Nuen for Ra = 105. We can also observe a slight differ-
ence in Nuen at around the top part of the left wall of the enclosure
between Ra = 105 and Ra = 106.
3.4. Surface-averaged Nusselt number
Fig. 11 shows the surface-averaged Nusselt numbers at the top
wall (NuT ), bottom wall (NuB ), right wall (NuR ), and left wall (NuL )
of the enclosure as a function of d for different Rayleigh numbers.
NuT generally increases with an increasing Rayleigh number, as
shown in Fig. 11(a), because of the increasing effect of convection.
When Ra = 103 or 104, conduction is the dominant heat transfer
mode, and the distribution of the isotherms strongly depends on
the locations of the cylinders in the enclosure, resulting in small
differences in the values of NuT according to d between Ra = 103
and 104. When Ra = 103 or 104, NuT increases gradually as d in-
creases from 0.25 to 0.1, whereas it increases very rapidly as
d increases from 0.1 to 0.25 because of the much stronger effect
of the distance between the cylinders and top wall of the enclo-
sure. As the Rayleigh number increases to 105, the values of NuT be-
come higher than those for Ra = 103 and 104 as a result of the
increasing effect of convection. When Ra = 105, the values of NuT
increase gradually as d increases from - 0.25 to 0, similar to that
for Ra = 103 and 104. However, as d increases from 0 to 0.1, the va-
lue of NuT decreases because the space between the hot cylinder
and top wall of the enclosure for the movement of the ascending
plume from the hot cylinder is not secured. As a result, the ascend-
ing plume from the hot cylinder, which exists at d 6 0 and im-
pinges on the top wall of the enclosure, disappears. As d
increases from 0.1 to 0.25 for Ra = 105, the value of NuT increases
again because of the increasing effect of the gap distance between
the cylinders and top wall of the enclosure. The values of NuT at
d = 0.2 and 0.25 for Ra = 105 are similar to those for Ra = 103 and
104 because of the strong effect of the gap distance between the
cylinders and top wall of the enclosure. As the Rayleigh number in-
creases to 106, the value of NuT becomes much higher than those
for Ra = 103, 104, and 105 because of the much larger effect of con-
vection. As d increases from 0.25 to 0.1 at Ra = 106, the value of
NuT increases gradually. However, as d increases from 0.1 to 0.2
at Ra = 106, the value of NuT decreases slightly because of the pres-
ence of two ascending plumes and a descending plume between
the top wall of the enclosure and the hot cylinder. As d increases
7924 Y.G. Park et al. / International Journal of Heat and Mass Transfer 55 (2012) 7911–7925

Fig. 11. Surface-averaged Nusselt numbers as function of d for four different Rayleigh numbers at each wall: (a) top wall ðNuT Þ, (b) bottom wall ðNuB Þ, (c) right wall ðNuR Þ, and
(d) left wall ðNuL Þ.

from 0.2 to 0.25 at Ra = 106, the value of NuT increases again be- of the enclosure blocks the movement of isotherms through this
cause of the strong effect of the gap distance between the cylinders gap.
and top wall of the enclosure. Unlike to the distributions of NuT , NuB , and NuR , the distribution
NuB generally decreases as a result of the increasing stagnant re- of NuL does not depend very much on the Rayleigh number but
gion between the cylinders and bottom wall of the enclosure with strongly depends on the distance between the hot cylinder and left
increasing d. NuB decreases rapidly as d increases from 0.25 to 0.1 wall of the enclosure, resulting in small differences in the values of
because of the strong effect of the distance between the cylinders NuL for different Rayleigh numbers. When Ra = 103, 104, or 105, the
and bottom wall of the enclosure, followed by a gradual decrease
as d increases from 0.1 to 0.25. The value of NuB for Ra = 104 is al-
most the same as that for Ra = 103 because of the dominant heat
transfer mode of conduction. As the Rayleigh number increases
to 105 and 106, the values of NuB become lower than those for
Ra = 103 and 104 because of the increasing stagnant region caused
by the increasing convection with increasing Rayleigh number. In
particular, the values of NuB for Ra = 106 are almost zero in the re-
gion 0:1 6 d 6 0:25 .
When Ra = 103 and 104, the values of NuR are almost zero be-
cause the cold cylinder is located close to the right wall of the cold
enclosure under the dominant conduction heat transfer mode. As
the Rayleigh number increases to 105, the values of NuR are slightly
higher than those for Ra = 103 and 104 but are still very low. If the
Rayleigh number increases further to 106, the values of NuR are
much higher than those for Ra = 103, 104, and 105 because an
ascending plume from the hot cylinder moves to the upper right
part of the enclosure as a result of the strong effect of convection,
as shown in Fig. 9, and transfers heat to the right wall of the enclo-
sure. As d increases at Ra = 106, the value of NuR decreases because Fig. 12. Surface-averaged Nusselt numbers of enclosure ðNuen Þ as function of d for
the decreasing gap distance between the cold cylinder and top wall different Rayleigh numbers.
 Y.G. Park et al. / International Journal of Heat and Mass Transfer 55 (2012) 7911–7925
value of NuL has a maximum at d = 0 and decreases slightly with
increasing or decreasing d. When Ra = 106, the value of NuL at
d < 0 is slightly higher than that at d > 0 because of the greater
amount of secure space between the cylinders and top wall of
the enclosure for the movement of the ascending plume from the
hot cylinder. As a result, the values of NuL at d < 0 for Ra = 106
are slightly higher than those for Ra 6 105 , whereas those at d > 0
for Ra = 106 are slightly lower than those for Ra 6 105 .
Fig. 12 shows the surface-averaged Nusselt numbers of the
enclosure (Nuen ) as a function of d for different Rayleigh numbers.
Nuen represents the surface-averaged sum of NuT , NuB , NuR , and
NuL , as shown in Fig. 11(a)–(d). When Ra = 103 or 104, Nuen has a
parabolic profile with a minimum value at d = 0. The value of
Nuen at Ra = 104 is slightly higher than that at Ra = 103. However,
when Ra = 105 or 106, the symmetry of Nuen is broken because of
the increasing effect of convection with the increasing Rayleigh
number. Nuen increases with the increasing Rayleigh number as a
result of the increasing effect of convection.
4. Conclusions
The present study investigated the effects the presence of left
hot and right cold cylinders located at different positions along
the vertical centerline of an enclosure has on the fluid flow and
heat transfer for different Rayleigh numbers in the range
103 6 Ra 6 106 . The immersed boundary method was imple-
mented in a finite volume method to simulate the flow and heat
transfer over two inner circular cylinders in a cold square enclo-
sure using a Cartesian coordinate system.
For all Rayleigh numbers considered in the present study, the
flow and thermal fields eventually reached steady state. The distri-
butions of the flow and thermal fields depended on the positions of
the cylinders in the enclosure and the buoyancy-induced convec-
tion by the ascending thermal plume from the hot cylinder. If the
Rayleigh number was as low as Ra = 103 or 104, the fluid flow
and heat transfer were mainly governed by conduction and by
the positions of the cylinders in the cold enclosure. Thus, when
the two cylinders were moved closer to the top wall of the enclo-
sure, denser and coarser isotherms were formed in the narrow re-
gion facing the cylinder movement and in the opposite wide
region, respectively. Consequently, the local peaks of the Nusselt
numbers on the surfaces of the two cylinders and the enclosure ap-
peared at the locations of the shortest distances between the cylin-
der surfaces and enclosure wall. However, if the Rayleigh number
was increased to 105 or 106, there was an increase in the effect
of the buoyancy-induced convection, in addition to the effect of
the cylinder positions, on the fluid flow and heat transfer in the
enclosure. Thus, when Ra = 105 or 106, the ascending plume from
the hot cylinder slanted toward the upper right part of the enclo-
sure, depending on the vertical positions of the cylinders and the
Rayleigh number, and interacted with the cold cylinder and cold
enclosure, giving different distributions of local Nusselt numbers
at the cylinders and enclosure from those for Ra = 103 and 104.
As both d and the Rayleigh number increased, NuT generally in-
creased because of the combined effects of the increased convec-
tion and decreasing distance between the cylinders and top wall
of the enclosure, whereas NuB decreased because of the combined
effects of the increasing stagnant region at the lower part of the
enclosure with increasing convection and because of the increasing
distance between the cylinders and top wall of the enclosure.
When Ra 6 105 , the values of NuR were close to zero because the
cold cylinder was exposed to the right wall of the cold enclosure.
However, when Ra = 106, the values of NuR were relatively higher
that those when Ra 6 105 because the ascending plume from the
hot cylinder moved to the right upper part through the gap
7925
between the cold cylinder and top wall of the enclosure. The distri-
bution of NuL strongly depended on the small distance between the
hot cylinder and left wall of the enclosure, resulting in a small var-
iation according to the Rayleigh number and cylinder position d.
When Ra = 103 or 104, the profiles of Nuen as a function of d showed
almost symmetric shapes with respect to d = 0. However, when
Ra = 105 or 106, this symmetry of Nuen was broken because of the
increasing effect of convection with the increasing Rayleigh
number.
Acknowledgement
This research was supported by the Leading Foreign Research
Institute Recruitment Program through the National Research
Foundation of Korea (NRF) funded by the Ministry of Education,
Science and Technology (MEST) (No. K20703001798-11E0100-
00310) and this work was supported by the Korean Research Foun-
dation (KRF) grant funded by the Korea government (MEST) (Nos.
20110027445 and 20100025618).
References
[1] M.Y. Ha, I.K. Kim, H.S. Yoon, S.S. Lee, Unsteady fluid flow and temperature
fields in a horizontal enclosure with an adiabatic body, Phys. Fluids 14 (2002)
3189–3202.
[2] M.Y. Ha, I.K. Kim, H.S. Yoon, K.S. Yoon, J.R. Lee, S. Balachandar, H.H. Chun, Two-
dimensional and unsteady natural convection in a horizontal enclosure with a
square body, Numer. Heat Transfer Part A 41 (2002) 183–210.
[3] J.R. Lee, M.Y. Ha, A numerical study of natural convection in a horizontal
enclosure with a conducting body, Int. J. Heat Mass Transfer 48 (2005) 3308–
3318.
[4] J.R. Lee, M.Y. Ha, Numerical simulation of natural convection in horizontal
enclosure with heat-generating conducting body, KSME J. 29 (2005) 441–452.
[5] J.R. Lee, M.Y. Ha, S. Balachandar, H.S. Yoon, S.S. Lee, Natural convection in a
horizontal layer of fluid with a periodic array of square cylinders in the
interior, Phys. Fluids 16 (2004) 1273–1286.
[6] J.M. Hyun, J.W. Lee, Numerical solutions for transient natural convection in a
square cavity with different sidewall temperatures, Int. J. Heat Fluid Flow 10
(1989) 146–151.
[7] D. Misra, A. Sarkar, Finite element analysis of conjugate natural convection in a
square enclosure with a conducting vertical wall, Comput. Methods Appl.
Mech. Eng. 141 (1997) 205–219.
[8] J.L. Wright, H. Jin, K.G.T. Hollands, D. Naylor, Flow visualization of natural
convection in a tall, air-filled vertical cavity, Int. J. Heat Mass Transfer 49
(2006) 889–904.
[9] G.D. McBain, Natural convection with unsaturated humid air in vertical
cavities, Int. J. Heat Mass Transfer 40 (1997) 3005–3012.
[10] M. Jami, A. Mezrhab, M. Bouzidi, P. Lallemand, Lattice Boltzmann method
applied to the laminar natural convection in an enclosure with a heat-
generating cylinder conducting body, Int. J. Thermal Sci. 46 (2007) 38–47.
[11] M.Y. Ha, M.J. Jung, A numerical study on three-dimensional conjugate heat
transfer of natural convection and conduction in a differentially heated cubic
enclosure with a heat-generating cubic conducting body, Int. J. Heat Mass
Transfer 43 (2000) 4229–4248.
[12] 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.
[13] 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.
[14] 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.
[15] O. Reymond, D.B. Murray, T.S. O’Donovan, Natural convection heat transfer
from two horizontal cylinders, Exp. Thermal Fluid Sci. 32 (2008) 1702–1709.
[16] 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.
[17] J. Kim, P. Moin, Application of a fractional step method to incompressible
Navier–Stokes equations, J. Comput. Phys. 59 (1985) 308–323.
[18] 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.
[19] 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.
[20] 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.