4 Cylinders

International Journal of Heat and Mass Transfer xxx (2017) xxx–xxx
Contents lists available at ScienceDirect
International Journal of Heat and Mass Transfer

journal homepage: www.elsevier.com/locate/ijhmt
Two-dimensional flow instability induced by natural convection in a

square enclosure with four inner cylinders. Part I: Effect of horizontal
position of inner cylinders
Young Min Seo a, Yong Gap Park b, Minsung Kim a, Hyun Sik Yoon c, 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
Rolls-Royce and Pusan National University Technology Centre in Thermal Management, Jang Jeon 2-Dong, Geum Jeong Gu, Busan 609-735, Republic of Korea
c
Global Core Research Centre 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: Two-dimensional numerical simulations were conducted for the natural convection phenomena in a cold
Received 26 October 2016 square enclosure with four hot inner circular cylinders with the Rayleigh number in the range of
Received in revised form 4 February 2017 103 Ra 106 and a Prandtl number of Pr = 0.7. The immersed boundary method (IBM) was used to cap-
Accepted 4 February 2017
ture the virtual wall boundary of the four inner cylinders based on the finite volume method (FVM). The
Available online xxxx
effects of the horizontal distances of 0:1L 6 eh 6 0:5L between the inner cylinders were analyzed by a
visualization technique, and the heat transfer characteristics over time were investigated in regard to
Keywords:
the transition of the flow regime from the steady state to the unsteady state at a relatively high
Natural convection
Immersed boundary method
Rayleigh number of Ra = 106. In addition, the heat transfer correlation equation was developed as a func-
Flow instability tion of the horizontal distance between cylinders.
Horizontal distance of four inner cylinders Ó 2017 Elsevier Ltd. All rights reserved.
1. Introduction nates mainly from the interaction effect in the flow and thermal
fields around neighboring cylinders. This effect is induced by the
Natural convection in an enclosure has been the subject of buoyant plume that is generated by each cylinder and may
numerous investigations in the past several decades. The thermo- impinge on another cylinder. In addition, the heat transfer in an
fluid behavior of systems based on natural convection has various enclosure generally increases with increasing space between cylin-
applications in engineering, such as cooling of electronic equip- ders, and the heat transfer increases with the misalignment of the
ment, heat exchangers, solar collectors, nuclear safety systems, cylinders [19–27].
and chemical reactors. In addition, almost all technologies involv- Corcione [19,20] numerically studied the steady, laminar, free
ing passive heat transfer for thermal dissipation rely on natural convection from flat, vertical arrays of equally spaced cylinders
convection. Many investigations have dealt with natural convec- and a pair of vertical arrays of equally spaced horizontal isothermal
tion under various thermal boundary conditions within a square cylinders set in free air. Simulations were performed for arrays of
enclosure both experimentally and numerically. The influence of 2–6 circular cylinders with center-to-center separation distances
an immersed object on natural convection within an enclosure of 2 to more than 50 cylinder diameters [19]. The Rayleigh number
has been studied extensively by many researchers [1–5]. The based on the cylinder diameter ranged between 5 102 and
buoyancy-induced convection mechanisms in heated cylinders 5 105. The heat transfer rate at the bottom cylinder remained
become more complex when there is interaction between the flow the same as that for a single cylinder. In contrast, the downstream
and the enclosure walls. Interacting effects between inner bodies cylinders exhibited either enhanced or reduced Nusselt numbers,
alter the heat transfer from neighboring cylinders in an array with depending on their location in the array and the array geometry.
two or more cylinders compared to a single cylinder [6–18]. This Simulations were also performed for pairs of tube arrays con-
difference between a single cylinder and multiple cylinders origi- sisting of 1–4 circular cylinders for center-to-center horizontal of
1.4–24 cylinder diameters and vertical spacing of 2–12 cylinder-
diameters [20]. The Rayleigh number was between 102 and 104.
⇑ Corresponding author. The cylinders exhibit either enhanced or reduced Nusselt numbers
E-mail addresses: breakfree86@pusan.ac.kr (Y.M. Seo), pyg777@pusan.ac.kr (Y.G.
compared to a single tube array, depending on their location in the
Park), kms514v@pusan.ac.kr (M. Kim), lesmodel@pusan.ac.kr (H.S. Yoon), myha@-
pusan.ac.kr (M.Y. Ha).
http://dx.doi.org/10.1016/j.ijheatmasstransfer.2017.02.011
0017-9310/Ó 2017 Elsevier Ltd. All rights reserved.
Please cite this article in press as: Y.M. Seo et al., Two-dimensional flow instability induced by natural convection in a square enclosure with four inner
cylinders. Part I: Effect of horizontal position of inner cylinders, Int. J. Heat Mass Transfer (2017), http://dx.doi.org/10.1016/j.
ijheatmasstransfer.2017.02.011
2 Y.M. Seo et al. / International Journal of Heat and Mass Transfer xxx (2017) xxx–xxx
Nomenclature
fi momentum forcing xi Cartesian coordinates

x
g acceleration of gravity xi dimensionless Cartesian coordinates (¼ Li )
L length of square enclosure
n normal direction to the wall Greek symbols
Nu local Nusselt number a thermal diffusivity
hNui surface-averaged Nusselt number b thermal expansion coefficient
hNui time and surface-averaged Nusselt number e dimensionless distance between the four cylinders
P⁄ pressure di2 Kronecker delta
2
P dimensionless pressure (¼ PqaL2 ) q density
Pr Prandtl number (=m/a) m kinematic viscosity
r dimensionless radius of the cylinder (=R/L) h dimensionless temperature ¼ TTT c
h T c
R radius of circular cylinder
3
Ra Rayleigh number (¼ gbL ðTmah T c Þ) Superscripts/subscripts
t⁄ time ⁄
dimensional value

t dimensionless time (¼ tL2a) –
surface-averaged quantity
T dimensional temperature C cylinder
Th hot temperature EN enclosure
Tc cold temperature
h horizontal
ui velocity v vertical
u L
ui dimensionless velocity (¼ ai )
W surface area of walls
array, the array geometry, and the Rayleigh number. Also, heat The natural convection heat transfer from an array of heated
transfer dimensionless correlation equations were proposed. cylinders confined to a rectangular enclosure was investigated pre-
Yousefi and Ashjaee [21] experimentally investigated the heat viously. However, little work has been reported on natural convec-
transfer in natural convection about elliptical cylinders with a ver- tion heat transfer generated from heated cylinders with various
tical major axis. Experiments were carried out using a Mach- patterns inside the enclosure. Lipnicki et al. [24] experimentally
Zehnder interferometer with cylinder spacing from two to five and theoretically studied combined natural convection and solidi-
major axis lengths and at Rayleigh numbers between 103 and fication in a system of cylinders arranged vertically in a regular tri-
2.5 103. They found that the free convection heat transfer from angular pattern. The cylinders were cooled below the solidification
any individual cylinder in the array depends on the Rayleigh num- temperature of water. They investigated the influence of the con-
ber, cylinder separation distance, and cylinder position in the tact layer between the ice-layer and the cold surface of the cylin-
array. They also proposed a heat transfer correlation for a single ders and its effect on the solidification process.
elliptical cylinder with a vertical major axis. Garoosi et al. [25,26] numerically investigated the natural and
Yaghoubi et al. [22] experimentally studied steady-state two- mixed convection heat transfer of nanofluids around several pairs
dimensional free convection heat transfer from a horizontal, of hot and cold cylinders in an adiabatic enclosure. They reported
isothermal cylinder in a horizontal array of cylinders consisting the effect of parameters on the fluid flow and heat transfer charac-
of three isothermal cylinders located underneath a nearly adiabatic teristics such as the Rayleigh number, size and type of the
ceiling. The effects of the cylinder spacing relative to the diameter nanoparticles, enclosure shape, and the orientation and number
(S/D) and the cylinder distance from the ceiling relative to the of the hot and cold cylinders. They showed that the heat transfer
diameter (L/D) on the heat transfer from the center cylinder were rate was enhanced by altering the orientation of the hot and cold
investigated for Rayleigh numbers of 1500–5000. The recirculation cylinders in horizontal to vertical modes. In another second study,
flow strength increased and the boundary layers of neighboring they considered the effects of the position and number of the cold
cylinders combined and formed a developing flow between cylin- cylinders, location and rotation direction of the heat source/sink,
ders with decreasing ratio S/D. Also, the variation of the average Rayleigh number, Richardson number, volume fraction, size and
Nusselt number of the center cylinder depends highly on L/D, type of the nanoparticles on fluid flow, and heat transfer rate
while the trend with respect to S/D depends on the Rayleigh [26]. They found that when changing the location of the heat
number. source/sink from a bottom-top configuration to top-bottom config-
Harsini and Ashjaee [23] experimentally and numerically inves- uration, the heat transfer rate decreased significantly.
tigated the effect of a vertical adiabatic wall on the natural convec- Park et al. [27] carried out a numerical simulation to investigate
tion heat transfer from a vertical array of attached cylinders, which natural convection heat transfer induced by a temperature differ-
can be considered as wavy surface. They focused on the effect of ence between a cold outer square enclosure and four hot inner cir-
spacing between the wall and wavy surface and Rayleigh number cular cylinders placed in a rectangular array. They used the
variation of the local and average free convection heat transfer immersed boundary method based on the finite volume method
coefficients from each cylinder and the wavy surface. They pro- to handle four inner cylinders placed equal distances away from
posed a single correlation that gives the average Nusselt number each other within an enclosure moving along the diagonal center-
as a function of the ratio of the wall and wavy surface spacing to lines for different Rayleigh numbers of 103 Ra 106. This study
the cylinder diameter and the Rayleigh number. They showed an investigated the effects of the location of the cylinders in the enclo-
optimum distance between the wall and wavy surface in which sure on the heat transfer and fluid flow along the diagonal center-
the Nusselt number attain its maximum value, depending on the lines of the enclosure.
Rayleigh number.
Y.M. Seo et al. / International Journal of Heat and Mass Transfer xxx (2017) xxx–xxx 3
The literature shows that the flow and heat transfer character- The governing equations for mass, momentum, and energy con-
istics are heavily affected by the location of hot bodies within an servation are defined in non-dimensional forms as follows:
enclosure and the geometry of the array of hot bodies. However,
@ui
it is hard to find numerical research on natural convection from q¼0 ð1Þ
@xi
four isothermally heated circular cylinders with a rectangular array
in the enclosure. Thus, the main objective of the present work is to
numerically examine the effects of the locations of such a setup @ui @ui @P @ 2 ui
þ uj ¼ þ Pr þ Ra Pr h di2 þ f i ð2Þ
with different locations of the cylinders in the Rayleigh number @t @xj @xi @xj @xj
range of 103 Ra 106. The effect of the horizontal distance
between cylinders was analyzed using a visualization technique @h @h @2h
þ uj ¼ þh ð3Þ
for the phenomena inside the enclosure, and the heat transfer @t @xj @xj @xj
characteristics in the system were addressed at a relatively high
The dimensionless variables are defined as:
Rayleigh number of Ra = 106. In addition, a heat transfer correlation
equation was developed as a function of the horizontal distance t a xi u L P L2 T Tc
between cylinders. t¼ ; xi ¼ ; ui ¼ i ; P ¼ ; h¼ ð4Þ
L 2 L a qa2 Th Tc
2. Numerical methodology where q, T, and a represent the density, dimensional temperature,

and thermal diffusivity, respectively. The superscript ⁄ in Eq. (4)
Fig. 1 shows the configuration of the system considered. The represents the dimensional variables: the Cartesian coordinates xi,
system consists of a square enclosure with four inner cylinders. corresponding velocity components ui, time t, pressure p, and tem-
The length of each side wall of the enclosure is L, and the radius perature h. The fluid properties are assumed to be constant except
of the inner cylinders is 0.1L. The four inner cylinders were moved for the density in the buoyancy term, which follows the Boussinesq
along the horizontal directions in the ranges of 0:1L 6 eh 6 0:5L in approximation. According to Gray and Giorgini [30], the range of
increments of 0.1L, where eh represents the dimensionless horizon- the Rayleigh number considered in this study is appropriate for
tal distance between inner cylinders. The dimensional vertical dis- using the Boussinesq approximation.
tance between the inner cylinders ev was fixed to 0.3L. The Rayleigh The non-dimensionalization produces two important dimen-
number range was 103 Ra 106, and the Prandtl number was set sionless parameters: Pr = m/a and Ra = gbL3(Th Tc)/ma, where m,
to Pr = 0.7 for all numerical simulations. This study considers the g, and b are the kinematic viscosity, gravitational acceleration,
problem of a viscous incompressible fluid with thermal convection and volume expansion coefficient, respectively. The terms q, fi,
with an immersed boundary with no-slip and isothermal boundary and h in Eqs. (1)(3) are extra terms in the governing equations to
conditions imposed on both boundaries, as shown in Fig. 1. which the immersed boundary method is applied. The mass
The immersed boundary method was used to represent the source/sink q in Eq. (1) and momentum force fi in Eq. (2) were
boundary of an inner cylinder in the square enclosure. According applied to the surface or inside the body to satisfy the no-slip con-
to a method proposed by Kim et al. [28], in order to capture the vir- dition and mass conservation in the cell containing the immersed
tual boundary, there are two different approaches for simulating a boundary. In Eq. (3), the heat source/sink h was applied to satisfy
complex flow: an unstructured grid method and the immersed the isothermal boundary condition on the surface of the inner
boundary method. The immersed boundary method has a benefit body. A second-order linear or bilinear interpolation scheme was
of more easily handling an inner object with a complex geometry applied to satisfy the no-slip and isothermal conditions on the
in a Cartesian coordinate system. In addition, Further details on immersed boundary.
the immersed boundary method are described by Kim et al. [28] Spatial discretization was approximated using the central dif-
and Kim and Choi [29]. ference scheme with second-order accuracy based on the finite-
volume method. In the discretization process, the advection terms
were explicitly treated by using the second-order Adams–Bash-
forth scheme, while the diffusion terms were implicitly treated
by using the second-order accuracy Crank–Nicolson scheme.
Once the velocity and temperature fields were obtained, the
instantaneous local Nusselt number Nu and instantaneous
surface-averaged Nusselt numbers Nu were defined as:
Z W
@h 1
Nu ¼ ; Nu ¼ Nu dW ð5Þ
@n wall W 0
where n is the normal direction with respect to the walls, and W is

the surface area of the walls. The time-averaged local Nusselt num-
ber hNui and the time and surface-averaged Nusselt number hNui
are defined as:
Z t p Z t p
1 1
hNui ¼ Nu dt; hNui ¼ Nu dt ð6Þ
tp 0 t p 0
where t p is the period of time integration.

In Fig. 2, dense grids with uniform distribution in the x and y
directions are generated in the computational domain in order to
capture the virtual wall boundary of the inner circular cylinders
accurately using the immersed boundary method. A mesh size of
Fig. 1. Computational domain and coordinate system along with boundary L/500 along the horizontal (x) and vertical (y) directions was used
conditions of the physical model. in our computations according to the results of a grid dependency
Ra = 105, the numerical results show time-independent character-

istics, regardless of the variation in horizontal distance between
inner cylinders. In addition, the flow and thermal fields obtained
with the Rayleigh number ranging from 103 to 105 are symmetric
about the vertical centerline of the enclosure. However, when the
Rayleigh number increases to Ra = 106, the flow experiences the
transition from the steady state to the unsteady state. In addition,
the flow and thermal fields inside the enclosure have various
unsteady states with the variation in horizontal distance between
inner cylinders at a relatively high Rayleigh number of Ra = 106.
Therefore, we focus on the flow instability and the consequent heat
transfer in the unsteady state with variation in the horizontal dis-
tance between inner cylinders at Ra = 106. The correlations
between the Rayleigh number and the horizontal distance are
highlighted by the heat transfer correlation equation.
3.1.1. Flow and thermal fields in the enclosure

Fig. 4 shows the distribution of isotherms and streamlines
Fig. 2. Grids with uniform distribution in the x and y directions at eh = 0.3L and inside the enclosure according to the variation in the Rayleigh
ev = 0.3L. number of 103 Ra 106. The ranges of contour values in the iso-
therms are from 0 to 1 with 10 levels. At Ra = 103, the flow and
thermal fields show a symmetric pattern about the vertical and
test. The test was conducted for much denser grids with a mini-
horizontal centerlines in the enclosure. This means that the heat
mum mesh size of L/600. The results showed that the numerical
transfer in the enclosure is governed mainly by the conduction
solutions are almost independent of the increase in the number
mode. For this reason, the flow is circulated by two rotating sym-
of grid points. The discrepancy between the result in the current
grid system and that in a denser grid system is less than 1% in
terms of the surface-averaged Nusselt number on the cylinder sur-
face because the resolution of the grid for the computational
domain is sufficiently high.
The present numerical simulation code for natural convection
has already been used to validate the available results in various
studies [3,14–17,27]. This code was validated using a natural con-
vection problem in a cold square enclosure with a hot inner circu-
lar cylinder at the center was tested. The surface-averaged Nusselt
(a) Ra=103
numbers were compared with those from the previous works,
which showed good agreement.
3. Results and discussion
3.1. Bifurcation map for different values of Ra and eh
Fig. 3 shows the bifurcation map for the variation in both the
Rayleigh number and the horizontal distance between inner cylin- (b) Ra=104
ders. At relatively low Rayleigh numbers of Ra = 103, Ra = 104, and
(c) Ra=105
(d) Ra=106
Fig. 4. Distribution of isotherms and streamlines for eh = 0.3L at different Ra
Fig. 3. Bifurcation map for different values of Ra and eh. (contour values range from 0 to 1 with 10 levels).
metric eddies with inner vortices because the space between the
upper cylinders and the lower cylinders is large enough. At
Ra = 104, the system inside the enclosure is still maintained by
the dominant conduction heat transfer, and the effect of convec-
tion is weakly increased as the Rayleigh number is increased to
Ra = 104. As a result, the flow and thermal fields show an asymmet-
ric pattern about the horizontal centerlines due to the convective
flow generated by the increasing Rayleigh number. However, the (a) εh = 0.1L
overall flow direction at Ra = 104 is generally similar to that at
Ra = 103. For Ra = 105, the effect of convection on the flow and ther-
mal fields also increases compared to those at Ra = 103 and
Ra = 104. As shown in Fig. 4(c), the distribution of isotherm at
Ra = 105 is skewed to the positive y-direction in the enclosure,
especially in the region between lower cylinders, while the rotat-
ing motion of the fluid is similar to those at Ra = 103 and
Ra = 104. When the Rayleigh number increases to Ra = 106, the flow
transitions from the steady state to the unsteady state at eh = 0.3L. (b) εh = 0.2L
Fig. 4(d) shows the result obtained from the time-averaged flow
and thermal fields. As mentioned, the flow and thermal fields at
eh = 0.3L and Ra = 106 show an asymmetric pattern about the verti-
cal centerline due to the unsteady characteristics. The flow pat-
terns at relatively low Rayleigh numbers of Ra = 103, Ra = 104,
and Ra = 105 are replaced, and the main eddy at a low Rayleigh
number turns into many small vortices at Ra = 106 due to the flow
instability generated with increasing convection effect. In addition,
the thermal field becomes more complex with increasing number
(c) εh = 0.3L
of thermal plumes from the inner cylinders at Ra = 106. Therefore,
we focus on the flow instability in the unsteady state at Ra = 106,
which depends on the variation in horizontal distances between
inner cylinders.
Fig. 4 shows the distribution of isotherms and streamlines with
cylinder distances ranging from 0.1L to 0.5L at Ra = 106. Fig. 5
(a) and (b) shows the results obtained from the numerical solu-
tions in the steady state, and Fig. 5(c)–(e) are the results from
the time-averaged flow and thermal fields in the unsteady state.
For Ra = 106, the flow and thermal structures reach various states (d) εh = 0.4L
with the variation in the horizontal distance due to the convection
effect generated at a relatively high Rayleigh number.
Generally, the lighter heated fluid flow generated from the
inner cylinders moves up along the hot surface of the cylinders
and then reaches the cold top wall when Ra = 106. Then, the fluid
gradually becomes colder and heavier while it moves outward
and down along the cold top and side walls of the enclosure. Simul-
taneously, a descending plume from the top wall of the enclosure is
formed at the space between the inner cylinders in the upper part (e) εh = 0.5L
of the enclosure by the strong ascending plumes generated from
the inner cylinders. As shown in Fig. 5(a), the rotating motion of Fig. 5. Distribution of isotherms and streamlines for different eh values at Ra = 106
(contour values range from 0 to 1 with 10 levels).
the fluid in the enclosure is restricted at eh = 0.1L due to the narrow
space between the upper cylinders. Therefore, the fluid is more
likely to pass in the space between inner cylinders when the hor- cylinders increases to eh = 0.3L, the rotating flow in the space
izontal distance between inner cylinders is larger. between the upper cylinders become stronger compared to
The flow and thermal fields at eh = 0.1L reach the steady state eh = 0.2L. In addition, the influence of the ascending plumes from
with the asymmetric structure because the distance between the lower cylinders increases around the upper cylinders in the
cylinders is very small at eh = 0.1L. The flow and thermal fields at thermal field with increasing eh. As a result, flow instability is gen-
eh = 0.2L reach steady state with a symmetric pattern about the erated by the interaction between the large convection and the
vertical centerline in the enclosure. When the horizontal distance horizontal distance between inner cylinders. The flow transition
between cylinders increases from 0.1L to 0.2L, the convective flow generated by the flow instability results in an asymmetric pattern
at eh = 0.2L is stabilized in the space between cylinders compared in the time-averaged flow and thermal fields at eh = 0.3L. Therefore,
to eh = 0.1L. As a result of this convective flow, the strength of the the flow and thermal fields reach an unsteady state due to the flow
descending plume increases in the upper part of the enclosure, instability at eh = 0.3L compared to eh = 0.1L and eh = 0.2L. When the
and a plume is more likely to ascend in the space between cylin- horizontal distance between cylinders increases to eh = 0.4L and
ders in the lower part of the enclosure. Therefore, the flow and eh = 0.5L, the numerical solutions reach the unsteady state with a
thermal fields show a symmetric pattern at eh = 0.2L. symmetric structure due to the space limitation between the cylin-
The flow transitions from steady state to unsteady state when ders and side walls of the enclosure, as shown in Fig. 5(d) and (e).
the horizontal distance between cylinders increases from eh = 0.2L Therefore, the unsteady dynamic characteristics over time were
to eh = 0.3L at Ra = 106. When the horizontal distance between
analyzed in order to investigate the effect of the horizontal dis- of the enclosure except for that on surface of cylinders. Thus, hNuTi
tance between inner cylinders at Ra = 106. is considered in order to investigate the dynamic characteristics of
unsteady states.
As shown in Fig. 6(a), hNuTi shows unsteady characteristics over
3.1.2. Dynamic characteristics of unsteady state
time, and its power spectrum in Fig. 6(b) has a primary frequency
Fig. 6(a) shows the time history of the surface-averaged Nusselt
of approximately 38 with several harmonics in a wide frequency
number on the top wall of the square enclosure hNuTi at eh = 0.3L
band. As a result, there is clear periodicity in the hNuTi signal with
and Ra = 106. Fig. 6(b) presents the power spectrum as a function
two high peaks and two low peaks in a cycle of sP. The maximum
of the frequency, and Fig. 6(c)–(f) shows the time evolution in
amplitude in hNuTi over time is approximately 12.3% at t = A com-
the distribution of isotherms and streamlines at four instants
pared to the time-averaged value. Therefore, the flow and thermal
(t = A–D) specified in Fig. 6(a). The contour values in the isotherms
fields formed in the enclosure vary with the motion of plumes from
are equal to those in Figs. 4 and 5. The surface-averaged Nusselt
the cylinder over time, as shown in Fig. 6(c)–(f).
number on the top wall is almost the largest value among walls
(a) (b)
(c) t = A
(d) t = B
(e) t = C
(f) t = D
Fig. 6. eh = 0.3L at Ra = 106; (a) time history and (b) power spectrum of Nu on top wall of the enclosure, (c)–(f) time evolution of isotherms and streamlines at four instants
(t = A–D).
When the horizontal distance is increased from eh = 0.2L to eddies near the side walls decreases compared to the case of
eh = 0.3L at Ra = 106, the temperature decreases as the space eh = 0.4L.
between cylinders increases. As a result, the descending plume In contrast, the magnitude of vortices between cylinders is gen-
from the top wall easily penetrates into the space between cylin- erally increased as the space between cylinders increases. At
ders in the upper part of the enclosure. The flow instability inten- eh = 0.5L, a regular pattern is observed in the flow and thermal
sifies and the heat transfer on the top wall changes over time. As fields over time, as shown in Fig. 8(c)–(f). The flow and thermal
shown in Fig. 6(c)–(f), four eddies are generated near each side wall fields at t = A have almost mirror-symmetry with those at t = C
(the left and right walls) and they are almost fixed at each time about the vertical centerlines. This is also the case for t = B and
except for t = B. The large and small eddies consist of complex con- t = D. As a result, hNuTi at t = A is approximately 0.53% compared
vection structures and vary in the space between cylinders at each to that at t = C, and hNuTi at t = B is approximately 2.74% compared
time. For this reason, hNuTi is affected by the direction of ascending to that at t = D. As mentioned, the flow and thermal fields are
plumes generated from the upper cylinders. In particular, if the therefore repeated over time with a pattern showing two high
ascending plumes from the upper cylinders are generated in a peaks and two low peaks in hNuTi in every cycle at eh = 0.5L and
diagonal direction, as shown at t = B and t = D, the surface- Ra = 106.
averaged Nusselt number at the top wall has low peaks, as shown Fig. 9 shows the distribution of the local Nusselt number on the
in Fig. 6(a). top wall in the x-direction over time at eh = 0.3L, eh = 0.4L, and
Fig. 7(a) shows the time history of the surface-averaged Nusselt eh = 0.5L at Ra = 106. Due to the presence of eddies in the upper part
number on the top wall at eh = 0.4L and Ra = 106. Fig. 7(b) presents and the resulting thermal field, the distribution of the local Nusselt
its power spectrum as a function of the frequency. Fig. 7(c)–(f) number varies along the x-direction over time, as shown in Fig. 9.
shows the time evolution of the distribution of isotherms and In particular, the local Nusselt number is affected by the horizontal
streamlines at four instants (t = A–D) specified in Fig. 7(a). distance between inner cylinders as well as unsteady characteris-
In Fig. 7(a), the time history of hNuTi shows unsteady character- tics over time.
istics with larger amplitude compared to eh = 0.3L, and its power In Fig. 9(a), at t = A, NuT has the highest value of 19.76 at x =
spectrum in Fig. 7(b) has a primary frequency of approximately +0.31, where an ascending plume from the upper-right cylinder
12 with several harmonics in a wide frequency band. As the hori- impinges on the top wall. The lowest value is 3.92 at x = 0, where
zontal distance between cylinders is increased to eh = 0.4L, the a descending plume exists around the center of the top wall. In
number of the harmonics around the primary frequency increases addition, a single high peak of 18.5 exists at x = 0.33 due to an
compared to eh = 0.3L. As a result, periodicity is identified in the ascending plume from the upper-left cylinder. At t = B, NuT has
signal of hNuTi with three high peaks and two low peaks in a cycle the highest value of 18.84 at x = +0.34, while NuT has the lowest
of sP. The transition in the amplitude from the lowest peak to the value of 3.46 at x = 0.14 due to the existence of eddies on the
highest peak over time is more complicated compared to upper-right cylinder, and hence, distortion of the plumes is gener-
eh = 0.3L. The maximum amplitude of hNuTi over time is within ated in the thermal field. In addition, two high peaks of 14.3 and
approximately 23.5% at t = C compared to the time-averaged value. 15.67 at x = 0.39 and x = 0.14 and a single low peak of 6.61 at
At eh = 0.4L, because the inner cylinders are close to the side x = 0.28 occur at 0.5 x 0. At t = C, the high peak is 19.11 at
walls, there is strong interaction between the large convection x = 0.31, while the low peak is 4.14 at x = 0.02. In addition,
and the small gap distance between the four cylinders and the side two high peaks occur of 15.61 at x = 0.13 and 17.77 at x = 0.33
walls. Consequently, the number of small eddies in the space because the ascending plumes impinge on the top wall, and hence,
between the cylinders and the side walls is increased compared finer plumes are generated in the two diagonal directions from the
to eh = 0.3L. In addition, because the ascending plumes from the upper-right cylinder. At t = D, the local Nusselt number on the top
upper cylinders impinge on the top wall when the descending wall has almost mirror-symmetry about x = 0. NuT has high peak
plume is clearly observed in the upper part of the enclosure, the values of 19.72 and 20.54 at x = 0.33 and x = +0.34, while NuT
Nusselt number has a high peak like at t = A or t = C. In particular, has low peak values of 3.74 and 3.91 at x = 0.13 and x = +0.13.
hNuTi has a maximum value at t = C when the flow and thermal At x = 0, NuT has a high peak value of about 9 according to the
fields have an almost symmetrical pattern about the vertical cen- motion of fluid and consequent the distribution of temperature
terlines. In contrast, when the descending plumes are unable to in the upper part of enclosure.
penetrate the space between upper cylinders, the Nusselt number Generally, the peak NuT on the right side of 0 x 0.5 is higher
has a low peak like at t = B or t = D. In the cases of t = B or t = D, sec- than that on the left side of 0.5 x 0 at the top wall, as shown
ondary eddies generated in the upper part of the enclosure have a in Fig. 9(b). At t = A, NuT has the highest value of 19.32 at x = +0.22
bad influence on the heat transfer of the top wall. and the lowest value of 3.39 at x = +0.02, where an ascending
Fig. 8(a) shows the time history of the surface-averaged Nusselt plume from the upper-right cylinder impinges on the top wall,
number on the top wall of the square enclosure at eh = 0.5L and and there is a descending plume around the center of the top wall.
Ra = 106. Fig. 8(b) presents its power spectrum as a function of In addition, NuT has a high peak value of 16.24 at x = 0.2 due to
the frequency, and Fig. 8(c)–(f) presents the time evolution in the the ascending plume from the upper-left cylinder. At t = B, the high
distribution of isotherms and streamlines at four instants (t = A– peak of NuT has a low value compared to other times due to the
D) specified in Fig. 8(a). secondary eddy in the upper part, as shown in Fig. 7(b). NuT has
As shown in Fig. 8(a), hNuTi shows unsteady characteristics with the highest value of 14.31 at x = +0.29, the lowest value of 4.23
periodicity, and its power spectrum in Fig. 8(b) has a primary fre- at x = 0.07, and a high peak value of 8.29 at x = 0.28. At t = C,
quency of approximately 37 with several harmonics in a wide fre- the local Nusselt number on the top wall has almost mirror-
quency band. The periodicity of hNuTi is identified with two high symmetry about x = 0 in the flow and thermal fields because the
peaks and two low peaks in a cycle of sP. The maximum amplitude fluid motion is almost symmetrical about the vertical centerlines
in hNuTi over time is approximately 23% at t = A compared to the at the upper part, as shown in Fig. 7(c). As a result, NuT has high
time-averaged value. As the horizontal distance between cylinders peak values of 18.73 and 19.78 at x = 0.19 and x = +0.16, while
is increased to eh = 0.5L, the space between the cylinders and side NuT has a low peak value of 4.55 at x = 0.02. At t = D, NuT has high
walls is extremely small. As a result, the circulating motion of peak values of 16.88 and 18.91 at x = 0.21 and x = +0.34, where
the fluid in this space is suppressed, and the magnitude of the the highest peak is slanted toward the right side, while NuT has a
low peak value of 2.88 at x = +0.14 due to a secondary eddy over
(a) (b)
(c) t = A
(d) t = B
(e) t = C
(f) t = D
(t = A–D).
the upper-right cylinder and the descending plume generated by and distortion in the plumes, NuT has a low peak value of 4.88 at
the motion of fluid. x = 0.38 and high peak value of 5.80 at x = 0.43.
At eh = 0.5L, the flow and thermal fields at t = A have almost
mirror-symmetry with those at t = C about the vertical centerlines,
as do t = B and t = D. This feature is clearer in Fig. 9(c). The peak val- 3.2. Time and surface-averaged Nusselt number
ues of NuT at eh = 0.5L are visible at t = A and t = B. At t = A, NuT has
high peak values of 17.89 and 20.49 at x = 0.29 and x = +0.20, Fig. 10(a) and (b) shows the time and surface-averaged Nusselt
while NuT has a low peak value of 3.01 at x = 0.01. At t = B, NuT number on each surface of the four cylinders and each wall of the
has high peak values of 9.39 and 19.18 at x = 0.41 and x = +0.22, enclosure as a function of the horizontal distance between cylin-
while NuT has a low peak value of 2.73 at x = 0.27. In addition, ders at Ra = 106. At Ra = 106, hNui is smaller at the upper cylinders
due to two small eddies generated from the upper-right cylinder than at the lower cylinders, as shown in Fig. 10(a). This is due to
(a) (b)
(c) t = A
(d) t = B
(e) t = C
(f) t = D
(t = A–D).
the ascending plumes from the lower cylinders. These plumes gen- and side walls. In particular, when the horizontal distance between
erally reduce the heat transfer rate at the upper cylinders com- inner cylinders increases to eh = 0.5L, hNui at the lower cylinders
pared to the lower cylinders. hNui at the lower-left cylinder is increases because the inner cylinders are very close to the side
similar to that at the lower-right cylinder due to the symmetric walls of enclosure. hNui at the upper-left cylinder is different from
pattern of the flow and thermal fields in the lower part of the with that at the upper-right cylinder at the cases of eh = 0.1L and
enclosure at Ra = 106, regardless of the variation in the horizontal eh = 0.3L. The pattern of the flow and thermal fields is also asym-
distance between cylinders.hNui at the lower cylinders decreases metric, as shown in Fig. 5(a) and (c). hNui at the upper-left cylinder
with increasing eh from 0.1L to 0.4L because the fluid flow has more increases with increasing eh due to the narrow space between the
difficulty passing through the space between the cylinders and side upper cylinders and the left walls of the enclosure. In addition,
walls. This is followed by the increase of hNui with increasing eh hNui at the upper-right cylinder decreases with increasing eh from
from 0.4L to 0.5L due to the confined space between the cylinders 0.1L to 0.2L, and then it increases with increasing eh from 0.2L to
Fig. 10. Time and surface-averaged Nu as function of eh when Ra = 106; (a) each
surface of four inner cylinders and (b) each wall of the enclosure.
eh = 0.3L. As shown in Fig. 10(b), hNui at the top wall has a maxi-
mum peak at eh = 0.2L due to the presence of the strong ascending
plumes, which impinge on the top wall from the upper cylinders. It
has minimum value at eh = 0.5L due to the weaker ascending
plumes on the upper inner cylinders. hNui at the right wall is sim-
ilar to that at the left wall because the flow and thermal fields near
the side wall are symmetric about the vertical centerline. hNui at
the side walls decreases with increasing eh from 0.1L to 0.2L due
to the fluid motion generated by increasing the number of eddies
near the side walls, followed by an increase with increasing eh from
0.2L to 0.5L due to the effect of the distance between the cylinders
and side walls. As a result, hNui at the side walls has a minimum
peak at eh = 0.2L. The variation in hNui at the bottom wall according
Fig. 9. Local Nu distribution on top wall of the enclosure in the x-direction at (a)
eh = 0.3L, (b) eh = 0.4L and (c) eh = 0.5L when Ra = 106. to eh is very slight compared to that at the other wall because the
stagnant region between the lower inner cylinders and the bottom
wall increases at Ra = 106 compared to relatively low Rayleigh
0.5L. At eh = 0.1L, the descending plume from the top wall is slanted numbers of Ra = 103, Ra = 104, and Ra = 105.
toward the upper-right cylinder. Therefore, hNui at the upper-right Fig. 11(a) and (b) shows the time and surface-averaged Nusselt
cylinder is higher than that at the upper-left cylinder by about numbers on all surfaces of the cylinders hNuC i and walls hNuEN i as a
18.7% at eh = 0.1L. Similarly, hNui at the upper-right cylinder is function of the horizontal distance between cylinders in the Ray-
higher than that at the upper-left cylinder by about 4% at leigh number range of 103 Ra 106. In Fig. 11(a), hNuC i generally
increased with increasing Rayleigh number due to the convection
low Rayleigh numbers of Ra = 103, Ra = 104, and Ra = 105. At

Ra = 106, the hNuEN i decreases from eh = 0.1L to eh = 0.3L and then
increases from eh = 0.3L to eh = 0.5L due to the combined effect of
between the variation in hNui at the top wall and that at the side
walls with increasing eh compared to that at bottom wall, as shown
in Fig. 10(b). Due to the transition of the flow regime from the
steady state to unsteady state at Ra = 106, the tendency of the
surface-averaged Nusselt number at Ra = 106 according to eh is dif-
ferent from with that of a relatively lower Rayleigh numbers, and
this feature is clearly observed in Fig. 11(a) and (b).
Fig. 12 shows the main effect of each parameter considered.
hNuC i and hNuEN i are represented as functions of the Rayleigh num-
ber and the horizontal distance between inner cylinders, respec-
tively. hNuC i and hNuEN i tend to increase as the Rayleigh number
and the horizontal distance between cylinders increase, and the
positive slopes of the plots gradually increase with increasing Ra
and eh. hNuC i increased by about 70.6% when the Rayleigh number
increased from Ra = 103 to Ra = 106, and it increased by about 65.4%
when the horizontal distance between cylinders increased from
eh = 0.1L to eh = 0.5L. This means that the horizontal distance and
Rayleigh number are important variables for the natural convec-
tion. In addition, hNuEN i increased by about 70.7% when the Ray-
leigh number increased from Ra = 103 to Ra = 106, and it
increased by about 65.2% when the horizontal distance increased
Fig. 11. Time and surface-averaged Nu distribution against eh variation at Ra = 103,

Ra = 104, Ra = 105, and Ra = 106; (a) all surface of four inner cylinders and (b) walls
of the enclosure.
effect. However, at Ra = 103 and 104, hNuC i has almost the same
value, regardless of the variation in the horizontal distance
between inner cylinders. In addition, hNuC i generally increases
with increasing eh at relatively low Rayleigh numbers of Ra = 103,
Ra = 104, and Ra = 105 because the distance between the cylinders
and walls decreases. In particular, when the horizontal distance
between inner cylinders is increased from eh = 0.4L to eh = 0.5L,
the space between cylinders and walls is reduced and the heat
transfer characteristics monotonically increases. The difference of
hNuC i between Ra = 104 and Ra = 105 decreases with increasing eh
because the heat transfer of cylinders is governed by the effect of
the horizontal distance between cylinders. When the Rayleigh
number increased to Ra = 106, the values of hNuC i rapidly increased
with increasing convection effect compared to Ra = 103, Ra = 104,
and Ra = 105. In addition, hNuC i decreases from eh = 0.1L to
eh = 0.2L and then increases from eh = 0.2L to eh = 0.5L at Ra = 106
due to the considerable variation in hNui at the upper-right cylin-
ders with increasing eh compared to that at the lower cylinders,
as show in Fig. 10(a).
In Fig. 11(b), the variation in hNuEN i is similar to that of hNuC i.
The values of hNuEN i increase slightly from eh = 0.1L to eh = 0.4L
Fig. 12. Main effect of parameters for time and surface-averaged Nu; (a) four inner
and increase monotonically from eh = 0.4L to eh = 0.5L at relatively cylinders and (b) outer enclosure.
from eh = 0.1L to eh = 0.5L. As shown in Fig. 12, The parameters con-

sidered have a considerable influence on hNuC i and hNuEN i.
In order to analyze the correlation of each parameter, hNuC i and
hNuEN i are expressed as a combined function with cubed terms
using the given Rayleigh number and horizontal distance between
inner cylinders:
hNuC or EN i ¼ a0 þ a1 Ra þ a2 eh þ a3 Ra2 þ a4 Raeh þ a5 e2h

þ a6 Ra3 þ a7 Ra2 eh þ a8 Rae2h þ a9 e3h ð7Þ
Table 1 shows a the estimated regression coefficients for each
term of the correlation equation, as well as the mean error e and
the maximum error emax for different Rayleigh numbers in the
range of 103 Ra 106 and horizontal distance between inner
cylinders of 0:1L 6 eh 6 0:5L. The mean error and maximum error
of hNuC i are less than 0.65% and 1.55%, and those of hNuEN i are less
than 0.66% and 1.68% compared to the numerical data,
respectively.
Fig. 13 shows a comparison between the numerical data and the
correlation equation. The case numbers are selected by full facto-
rial design according to the variation in horizontal distance
between inner cylinders at different Rayleigh numbers. The details
about the case numbers are presented in Table 2. The maximum
error between the numerical data and the correlation equation in
Table 1 is in case 18, where the horizontal distance between inner
cylinders is 0.3 at Ra = 106, as shown in Fig. 13. However, in this fig-
ure, the surface-averaged Nusselt number hNuC i on the surfaces of
cylinders and hNuEN i on the walls of the correlation equations
match well with those of the numerical data. As a result, excellent
agreement with the numerical data was achieved, as shown in
Table 1 and Fig. 13.
4. Conclusion
Two-dimensional numerical analysis was conducted on a natu-

ral convection problem in a square enclosure with four inner circu-
lar cylinders. The key simulation parameters were the Rayleigh
number and horizontal distance between cylinders. At Ra = 103, Fig. 13. Comparison between the numerical data and the correlation equation; (a)
four inner cylinders and (b) outer enclosure.
Ra = 104, and Ra = 105, the numerical solutions show time-
independent characteristics, and the flow and thermal structures
show a mirror-symmetric pattern about the horizontal centerlines, increases to eh = 0.3L at Ra = 106, the flow experiences a transition
regardless of the variation in the horizontal distance between from steady state to unsteady state. The flow and thermal fields
cylinders. When the Rayleigh number is increased to Ra = 106, inside the enclosure have various unsteady states with the varia-
the numerical solutions still show the characteristics of steady tion in horizontal distance between cylinders at Ra = 106.
states over time at eh = 0.1L and eh = 0.2L. However, the flow and When the horizontal distance between cylinders is increased
thermal fields show an asymmetric pattern about the horizontal from eh = 0.2L to eh = 0.3L for Ra = 106, the temperature decreases
centerlines at eh = 0.1L. In addition, as the horizontal distance as the space between cylinders increases. As a result, the descend-
ing plume from the top wall easily penetrated the space between
Table 1 cylinders in the upper part. In this process, the flow instability is
The correlating equations for the total surface-averaged Nusselt number of a cylinder intensified, and the heat transfer on the top wall is affected over
hNuC i and enclosure hNuEN i at different Rayleigh numbers in the range of 103 - time. When the horizontal distance between cylinders increases
Ra 106 and horizontal distance between inner cylinders of 0:1L 6 eh 6 0:5L. further to eh = 0.4L, because the cylinders are close to the side walls,
Variable Surface-averaged Nusselt numbers there is stronger interaction between the large convection and the
small gap distance between the four cylinders and the side walls at
hNuC i hNuEN i
Ra = 106. Finally, when the distance between cylinders is increased
a0 3.85093 2.08672
to eh = 0.5L, the space between the cylinders and side walls is extre-
a1 1.28E05 7.14E06
a2 25.1699 14.336 mely small. As a result, the circulating motion of the fluid in the
a3 7.13E11 3.95E11 space between the cylinders and side walls is suppressed, and
a4 3.36E05 1.99E05 hence the magnitude of eddies near the side walls decreases com-
a5 8.94E+01 5.08E+01 pared to the case of eh = 0.4L.
a6 7.54E17 4.17E17
In order to analyze the effect of each parameter, the heat trans-
a7 2.16E11 1.18E11
a8 6.10E06 5.82E06 fer characteristics on the surfaces of cylinders and walls were con-
a9 1.43E+02 8.03E+01 sidered. hNuC i and hNuEN i tended to increase as the Rayleigh

e 0.65% 0.66% number and the horizontal distance between cylinders increased.
emax 1.55% 1.68%
Table 2
Detailed case numbers of the correlating equation.
Case numbers
Parameter eh
0.1 0.2 0.3 0.4 0.5
Ra 103 1 2 3 4 5
104 6 7 8 9 10
105 11 12 13 14 15
106 16 17 18 19 20
In addition, the correlation equations were obtained using full fac- [13] Y. Lu, Q. Yu, W. Du, Y. Wu, Natural convection heat transfer of molten salts
around a vertically aligned horizontal cylinder set, Int. Commun. Heat Mass
torial design according to the variation in horizontal distance
Transf. 76 (2016) 147–155.
between cylinders at different Rayleigh numbers. The results of [14] H.S. Yoon, Y.G. Park, J.H. Jung, Natural convection in a square enclosure with
the equations matched well with the numerical data. differentially heated two horizontal cylinders, Numer. Heat Transf. A: Appl. 65
(2014) 302–326.
[15] H.S. Yoon, J.H. Jung, Y.G. Park, Natural convection in a square enclosure with
Acknowledgements two horizontal cylinders, Numer. Heat Transf. A: Appl. 62 (2012) 701–721.
[16] Y.G. Park, M.Y. Ha, C. Choi, J. Park, Natural convection in a square enclosure
This subject is supported by Korea Ministry of Environment with two inner circular cylinders positioned at different vertical locations, Int.
J. Heat Mass Transf. 77 (2014) 501–518.
(MOE) as ‘‘the Chemical Accident Prevention Technology Develop- [17] Y.G. Park, M.Y. Ha, H.S. Yoon, Study on natural convection in a cold square
ment Project (No. 2015001950002).” enclosure with a pair of hot horizontal cylinders positioned at different
vertical locations, Int. J. Heat Mass Transf. 65 (2013) 696–712.
[18] J. Stafford, V. Egan, Configurations for single-scale cylinder pairs in natural
References convection, Int. J. Therm. Sci. 84 (2014) 62–74.
[19] M. Corcione, Correlating equations for free convection heat transfer from
[1] D. Angeli, P. Levoni, G.S. Barozzi, Numerical predictions for stable buoyant horizontal isothermal cylinders set in a vertical array, Int. J. Heat Mass Transf.
regimes within a square cavity containing a heated horizontal cylinder, Int. J. 48 (2005) 3660–3673.
Heat Mass Transf. 51 (2008) 553–565. [20] M. Corcione, Interactive free convection from a pair of vertical tube-arrays at
[2] C. Shu, Y.D. Zhu, Efficient computation of natural convection in a concentric moderate Rayleigh numbers, Int. J. Heat Mass Transf. 50 (2007) 1061–1074.
annulus between an outer square cylinder and an inner circular cylinder, Int. J. [21] T. Yousefi, M. Ashjaee, Experimental study of natural convection heat transfer
Numer. Meth. Fluids 38 (2002) 429–445. from vertical array of isothermal horizontal elliptic cylinders, Exp. Therm.
[3] B.S. Kim, D.S. Lee, M.Y. Ha, H.S. Yoon, A numerical study of natural convection Fluid Sci. 32 (2007) 240–248.
in a square enclosure with a circular cylinder at different vertical locations, Int. [22] M. Yaghoubi, M. Ashjaee, A.H. Eshtiaghi, T. Yousefi, Thermal study of an array
J. Heat Mass Transf. 51 (2008) 1888–1906. of inline horizontal cylinders below a nearly adiabatic ceiling, Exp. Therm.
[4] C.C. Liao, C.A. Lin, Transitions of natural convection flows in a square enclosure Fluid Sci. 33 (2009) 232–239.
with a heated circular cylinder, Appl. Therm. Eng. 72 (2014) 41–47. [23] I. Harsini, M. Ashjaee, Effect of adiabatic wall on the natural convection heat
[5] C.C. Liao, C.A. Lin, Influence of Prandtl number of the instability of natural transfer from a wavy surface created by attached horizontal cylinders, Exp.
convection flows within a square enclosure containing an embedded heated Therm. Fluid Sci. 34 (2010) 666–676.
cylinder at moderate Rayleigh number, Phys. Fluids 27 (2015) 013603. [24] Z. Lipnicki, D. Sasse, B. Weigand, Experimental and analytical investigation of
[6] R. Chouikh, A. Guizani, A.E. Cafsi, M. Maalej, A. Belghith, Experimental study of the solidification around cooled cylinders subjected to free convection, Int. J.
the natural convection flow around an array of heated horizontal cylinders, Heat Mass Transf. 78 (2014) 321–329.
Renewable Energy 21 (2000) 65–78. [25] F. Garoosi, F. Hoseininejad, M.M. Rashidi, Numerical study of natural
[7] C. Cianfrini, M. Corcione, E. Habib, Free convection heat transfer from a convection heat transfer in a heat exchanger filled with nanofluids, Energy
horizontal cylinder affected by a downstream parallel cylinder of different 109 (2016) 664–678.
diameter, Int. J. Therm. Sci. 45 (2006) 923–931. [26] F. Garoosi, F. Hoseininejad, Numerical study of natural and mixed convection
[8] O. Reymond, D.B. Murray, T.S. O’Donovan, Natural convection heat transfer heat transfer between differentially heated cylinders in an adiabatic enclosure
from two horizontal cylinders, Exp. Therm. Fluid Sci. 32 (2008) 1702–1709. filled with nanofluid, J. Mol. Liq. 215 (2016) 1–17.
[9] M.S. Chae, B.J. Chung, Effect of pitch-to-diameter ratio on the natural [27] Y.G. Park, M.Y. Ha, J. Park, Natural convection in a square enclosure with four
convection heat transfer of two vertically aligned horizontal cylinders, Chem. circular cylinders positioned at different rectangular locations, Int. J. Heat
Eng. Sci. 66 (2011) 5321–5329. Mass Transf. 81 (2015) 490–511.
[10] S. Grafsronningen, A. Jensen, Natural convection heat transfer from two [28] J. Kim, D. Kim, H. Choi, An immersed-boundary finite volume method for
horizontal cylinders at high Rayleigh numbers, Int. J. Heat Mass Transf. 55 simulations of flow in complex geometries, J. Comput. Phys. 171 (2001) 132–
(2012) 5552–5564. 150.
[11] J.H. Heo, M.S. Chae, B.J. Chung, Influences of vertical and horizontal pitches on [29] J. Kim, H. Choi, An immersed-boundary finite-volume method for simulation
the natural convection of two staggered cylinders, Int. J. Heat Mass Transf. 57 of heat transfer in complex geometries, KSME Int. J. 18 (2004) 1026–1035.
(2013) 1–8. [30] D.D. Gray, A. Giorgini, The validity of the Boussinesq approximation for liquids
[12] J.H. Heo, B.J. Chung, Natural convection of two staggered cylinders for various and gases, Int. J. Heat Mass Transf. 19 (1976) 545–551.
Prandtl numbers and vertical and horizontal pitches, Heat Mass Transf. 50
(2014) 769–777.

4 Cylinders

Uploaded by

Copyright:

Available Formats

4 Cylinders

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

4 Cylinders

Uploaded by

Copyright:

Available Formats

International Journal of Heat and Mass Transfer xxx (2017) xxx–xxx

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer

Two-dimensional flow instability induced by natural convection in a

fi momentum forcing xi Cartesian coordinates

2. Numerical methodology where q, T, and a represent the density, dimensional temperature,

where n is the normal direction with respect to the walls, and W is

where t p is the period of time integration.

Ra = 105, the numerical results show time-independent character-

3.1.1. Flow and thermal fields in the enclosure

3. Results and discussion

3.1. Bifurcation map for different values of Ra and eh

low Rayleigh numbers of Ra = 103, Ra = 104, and Ra = 105. At

Fig. 11. Time and surface-averaged Nu distribution against eh variation at Ra = 103,

from eh = 0.1L to eh = 0.5L. As shown in Fig. 12, The parameters con-

hNuC or EN i ¼ a0 þ a1 Ra þ a2 eh þ a3 Ra2 þ a4 Raeh þ a5 e2h

Two-dimensional numerical analysis was conducted on a natu-

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.