You are on page 1of 13

International Journal of Engineering Science 43 (2005) 668–680

www.elsevier.com/locate/ijengsci

Finite element analysis of natural convection flows in a


square cavity with non-uniformly heated wall(s)
a,* b
S. Roy , Tanmay Basak
a
Department of Mathematics, Indian Institute of Technology Madras, Chennai 600 036, India
b
Department of Chemical Engineering, Indian Institute of Technology Madras, Chennai 600 036, India

Received 27 August 2004; accepted 20 January 2005

(Communicated by ERINGEN)

Abstract

A penalty finite element analysis with bi-quadratic rectangular elements is performed to investigate the
influence of uniform and non-uniform heating of wall(s) on natural convection flows in a square cavity. In
the present investigation, one vertical wall and the bottom wall are uniformly and non-uniformly heated
while the other vertical wall is maintained at constant cold temperature and the top wall is well insulated.
Parametric study for a wide range of Rayleigh number (Ra), 103 6 Ra 6 106 and Prandtl number (Pr),
0.2 6 Pr 6 100 shows consistent performance of the present numerical approach to obtain the solutions
as stream functions and temperature profiles. Heat transfer rates at the heated walls are presented in terms
of local Nusselt number.
 2005 Elsevier Ltd. All rights reserved.

Keywords: Penalty finite element method; Natural convection; Square cavity; Non-uniform heating

*
Corresponding author. Tel.: +91 44 22578492; fax: +91 44 22578470.
E-mail addresses: sjroy@iitm.ac.in (S. Roy), tanmay@iitm.ac.in (T. Basak).

0020-7225/$ - see front matter  2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijengsci.2005.01.002
S. Roy, T. Basak / International Journal of Engineering Science 43 (2005) 668–680 669

1. Introduction

In recent past, an ever-increasing awareness in thermally driven flows reflect that fluid motions
and transport processes generated or altered by buoyancy are of interest and significance in many
fields of science and technology. As a result, this subject is currently studied in diverse areas of
meteorology, geophysics, nuclear reactor systems, energy storage and conservation, fire control,
and chemical, food, and metallurgical industries, as well as in the more conventional fields of
the fluid and thermal sciences. Buoyancy driven flows are complex because of essential coupling
between the transport properties of flow and thermal fields. These problems can be classified as
either external ones known as free convection or internal ones called as natural convection. A lit-
erature survey shows that the comprehensive review of these problems was made by Ostrach [1–3],
Gebhart [4] and Hoogendoorn [5] in which each emphasizes essentially different aspects of the
subject.
It may be pointed out that internal problems are considerably more complex than external
ones. This is because at large Rayleigh number (product of Prandtl and Grashof numbers) clas-
sical boundary layer theory yields the same simplifications for external problems, namely, the re-
gion exterior to the boundary layer is unaffected by the boundary layer. For confined natural
convection, in contrast, boundary layers form near the walls but the region exterior to them is
enclosed by the boundary layers and forms a core region. Since the core is partially or fully encir-
cled by the boundary layers, the core flow is not readily determined from the boundary conditions
but depend on the boundary layer, which, in turn, is influenced by the core. The interactions be-
tween the boundary layer and core constitute a major complexity in the problem and is inherent to
all confined convection configurations, namely, that the flow pattern cannot be predicted a priori
from the given boundary conditions and geometry. In fact, the situation is even more intricate
because it often appears that more than one global core flow is possible and flow subregions, such
as cells and layers, may be embedded in the core. This physical complexity in confined convection
is not only a topic for analysis but has equal significance for numerical and experimental investi-
gations. The extensive research studies using various numerical simulations reported by Patterson
and Imberger [6], Nicolette et al. [7], Hall et al. [8], Hyun and Lee [9], Fusegi et al. [10], Lage and
Bejan [11,12], and Xia and Murthy [13] ensure that several attempts have been made to acquire a
basic understanding of natural convection flows and heat transfer characteristics in an enclosure.
However, in most studies, one vertical wall of the enclosure is cooled and another one heated
while the remaining top and bottom walls are well insulated.
The aim of the present investigation is to study the circulations, temperature distributions with-
in the cavity and heat transfer rate at the heated walls in terms of local Nusselt number when the
bottom wall and one vertical wall are heated (uniformly and non-uniformly) and the top wall is
well insulated while the other vertical wall is cooled by means of a constant temperature bath (see
Fig. 1). In case of uniformly heated walls, the finite discontinuity in temperature distribution ap-
pears at the right edge of the bottom wall. The discontinuity can be avoided by choosing a non-
uniform temperature distribution along the walls (i.e., non-uniformly heated walls) as discussed
by Minkowycz et al. [14] where an investigation is made for a mixed convection flow on a vertical
plate, which is non-uniformly heated /cooled. In the current study, Galerkin finite element method
with penalty parameter has been used to solve the nonlinear coupled partial differential equations
for flow and temperature fields with both uniform and non-uniform temperature distributions
670 S. Roy, T. Basak / International Journal of Engineering Science 43 (2005) 668–680

y, v

Adiabatic Wall

Tc
or
Th

x,u
Th
or

Fig. 1. Schematic diagram of the physical system.

prescribed at the bottom wall and at one vertical wall. It may be noted that Galerkin finite element
method has been used to solve fluid flow and heat transfer problems in recent investigations by
Nishimura et al. [15], Basak and Ayappa [16] and Asaithambi [17].

2. Governing equations

Thermophysical properties of the fluid in the flow model are assumed to be constant except the
density variations causing a body force term in the momentum equation. The Boussinesq approx-
imation is invoked for the fluid properties to relate density changes to temperature changes, and
to couple in this way the temperature field to the flow field. The governing equations for steady
natural convection flow using conservation of mass, momentum and energy can be written as:
ou ov
þ ¼0 ð1Þ
ox oy
 2 
ou ou 1 op o u o2 u
u þv ¼ þm þ ; ð2Þ
ox oy q ox ox2 oy 2
 2 
ov ov 1 op o v o2 v
u þv ¼ þm þ þ gbðT  T c Þ; ð3Þ
ox oy q oy ox2 oy 2
 2 
oT oT o T o2 T
u þv ¼a þ ; ð4Þ
ox oy ox2 oy 2
S. Roy, T. Basak / International Journal of Engineering Science 43 (2005) 668–680 671

with boundary conditions

uðx; 0Þ ¼ uðx; LÞ ¼ uð0; yÞ ¼ uðL; yÞ ¼ 0;


vðx; 0Þ ¼ vðx; LÞ ¼ vð0; yÞ ¼ vðL; yÞ ¼ 0;
px oT ð5Þ
T ðx; 0Þ ¼ T h ; or T ðx; 0Þ ¼ ðT h  T c Þ sin þ T c; ðx; LÞ ¼ 0;
L oy
py 
T ð0; yÞ ¼ T h ; or T ð0; yÞ ¼ ðT h  T c Þ sin þ T c; T ðL; yÞ ¼ T c
L
where x and y are the distances measured along the horizontal and vertical directions, respec-
tively; u and v are the velocity components in the x- and y-directions, respectively; T denotes
the temperature; m and a are kinematic viscosity and thermal diffusivity, respectively; p is the pres-
sure and q is the density; Th and Tc are the temperatures at hot and cold walls, respectively; L is
the side of the square cavity.
Using the following change of variables,
x y uL vL
X ¼ ; Y ¼ ; U¼ ; V ¼ ;
L L a a
pL2 T  Tc
P ¼ 2; h ¼ ; ð6Þ
qa Th  Tc
the governing equations (1)–(4) reduce to non-dimensional form:
oU oV
þ ¼0 ð7Þ
oX oY
 2 
oU oU oP o U o2 U
U þV ¼ þ Pr þ ; ð8Þ
oX oY oX oX 2 oY 2
 2 
oV oV oP o V o2 V
U þV ¼ þ Pr þ þ Ra Pr h; ð9Þ
oX oY oY oX 2 oY 2
oh oh o2 h o2 h
U þV ¼ þ ; ð10Þ
oX oY oX 2 oY 2
with the boundary conditions
U ðX ; 0Þ ¼ U ðX ; 1Þ ¼ U ð0; Y Þ ¼ U ð1; Y Þ ¼ 0;
V ðX ; 0Þ ¼ V ðX ; 1Þ ¼ V ð0; Y Þ ¼ V ð1; Y Þ ¼ 0;
oh
hðX ; 0Þ ¼ 1; or hðX ; 0Þ ¼ sinðpX Þ; ðX ; 1Þ ¼ 0;
oY
hð0; Y Þ ¼ 1; or hð0; Y Þ ¼ sinðpY Þ; hð1; Y Þ ¼ 0; ð11Þ
Here X and Y are dimensionless coordinates varying along horizontal and vertical directions,
respectively; U and V are, dimensionless velocity components in the X- and Y-directions, respec-
tively; h is the dimensionless temperature; P is the dimensionless pressure; Ra and Pr are Rayleigh
and Prandtl numbers, respectively.
672 S. Roy, T. Basak / International Journal of Engineering Science 43 (2005) 668–680

The heat transfer coefficient in terms of the local Nusselt number is defined by
oh
Nu ¼  ; ð12Þ
on
where n denotes the normal direction on a plane.

3. Method of solution

The momentum and energy balance equations (8)–(10) are solved using the Galerkin finite ele-
ment method. The continuity equation (7) will be used as a constraint due to mass conservation
and this constraint may be used to obtain the pressure distribution [16,18]. In order to solve Eqs.
(8)–(10), we use the penalty finite element method where the pressure P is eliminated by a penalty
parameter c and the incompressibility criteria given by Eq. (7) (see Reddy [18]) which results in
 
oU oV
P ¼ c þ . ð13Þ
oX oY
The continuity equation (7) is automatically satisfied for large values of c. Typical values of c
that yield consistent solutions are 107 [16,18].
Using Eq. (13), the momentum balance equations (8) and (9) reduce to
   2 
oU oU o oU oV o U o2 U
U þV ¼c þ þ Pr þ ; ð14Þ
oX oY oX oX oY oX 2 oY 2
and
   2 
oV oV o oU oV o V o2 V
U þV ¼c þ þ Pr þ þ Ra Pr h. ð15Þ
oX oY oY oX oY oX 2 oY 2
Expanding the velocity components (U, V) and temperature (h) using basis set fUk gNk¼1 as,
XN X
N X
N
U U k Uk ðX ; Y Þ; V  V k Uk ðX ; Y Þ; and h  hk Uk ðX ; Y Þ; ð16Þ
k¼1 k¼1 k¼1

for
0 6 X ; Y 6 1;
the Galerkin finite element method yields the following nonlinear residual equations for Eqs. (14),
(15) and (10), respectively, at nodes of internal domain X:
X N Z " X N
!
X N
! #
ð1Þ oU k oU k
Ri ¼ Uk U k Uk þ V k Uk Ui dX dY
k¼1 X k¼1
oX k¼1
oY
" Z Z #
X N
oUi oUk XN
oUi oUk
þc Uk dX dY þ Vk dX dY
k¼1 X oX oX k¼1 X oX oY

XN Z
oUi oUk oUi oUk
þ Pr Uk þ dX dY ð17Þ
k¼1 X oX oX oY oY
S. Roy, T. Basak / International Journal of Engineering Science 43 (2005) 668–680 673

X
N Z " X
N
!
X
N
! #
ð2Þ oUk oUk
Ri ¼ Vk U k Uk þ V k Uk Ui dX dY
k¼1 X k¼1
oX k¼1
oY
" Z Z #
X
N
oUi oUk XN
oUi oUk
þc Uk dX dY þ Vk dX dY
k¼1 X oY oX k¼1 X oY oY
Z Z !
X
N
oUi oUk oUi oUk X
N
þ Pr Vk þ dX dY  Ra Pr hk Uk Ui dX dY ð18Þ
k¼1 X oX oX oY oY X k¼1

and
X
N Z " X
N
!
XN
! #
ð3Þ oUk oUk
Ri ¼ hk U k Uk þ V k Uk Ui dX dY
k¼1 X k¼1
oX k¼1
oY

X
N Z
oUi oUk oUi oUk
þ hk þ dX dY . ð19Þ
k¼1 X oX oX oY oY
Bi-quadratic basis functions with three point Gaussian quadrature is used to evaluate the inte-
grals in the residual equations. In Eqs. (17) and (18), the second term containing the penalty
parameter (c) are evaluated with two point Gaussian quadrature (reduced integration penalty for-
mulation, [18]). The motivation for reduced integration is given below. The matrix vector notation
for the penalty finite element equations of the residuals i.e., Eqs. (17)–(19) may be expressed in
matrix vector notation as
ðK1 þ cK2 Þa ¼ F; ð20Þ
where a denotes the unknown vector, K1, K2 are the matrices obtained from the Jacobian of the
residuals, As c tends to a large value ( 107), the constraint equation (i.e., continuity equation) is
satisfied better, which in turn causes the magnitude of K1 is negligible when compared with cK2
resulting in
F
K2 a ¼ . ð21Þ
c
This implies that as c tends to infinity, governing equations are left with only the constraint con-
dition, i.e, the continuity equation. Hence, the contributions from the momentum and energy con-
servations are completely lost. In addition, as K2 is nonsingular for large c the resulting solution
obtained from Eq. (21) is trivial. To obtain the non-trivial solutions for large c ( 107) the matrix
K2 is needed to be a singular matrix. This is obtained by using two point Gaussian quadrature for
K2 and three point Gaussian quadrature for K1. In the absence of the above reduced integration
method velocities are underestimated ([18]).
The non-linear residual equations (17)–(19) are solved using a Newton–Raphson procedure to
determine the coefficients of the expansions in Eq. (16). At each iteration, the linear (3N · 3N)
system;

Jðan Þ an  anþ1 ¼ Rðan Þ; ð22Þ
674 S. Roy, T. Basak / International Journal of Engineering Science 43 (2005) 668–680

is solved where n is the iterative index. The elements of the Jacobian matrix, J(an) contains the
derivatives of the residual equations with respect to velocity components (Uj)Õs, (Vj)Õs and the tem-
perature (hjÕs) and R(an) is the vector of residuals. The linear system for each iteration is based on
efficient node numbering of the elements such that the jacobianhP forms i0.5 a banded matrix. The iter-
ðjÞ 2
ative process is terminated with the convergence criterion ðRi Þ 6 105 using two-norm of
residual vectors.
We have used nine node bi-quadratic elements with each element mapped using iso-parametric
mapping ([18]) from X–Y to a unit square n–g domain. Subsequently, the domain integrals in the
residual equations are evaluated using nine node bi-quadratic basis functions in n–g domain as:
X
9 X
9
X ¼ X i Ui ðn; gÞ and Y ¼ Y i Ui ðn; gÞ; ð23Þ
i¼1 i¼1

where Ui(n, g) are the local bi-quadratic basis functions on the n–g domain. The integrals in
Eqs. (17)–(19) can be evaluated in n–g domain using following relationships:
" oUi # " oY 2 3
oY # oUi
1 og
 on
oX
¼ 4 on 5
oUi jJ j  oX oX oUi
oY og on og

and
dX dY ¼ jJ jdn dg ð24Þ
 
oðX ;Y Þ
where J ¼  oðn;gÞ . The local Nusselt number on a surface (Nu) contains a normal derivative as
shown in Eq. (12) and the normal derivative is evaluated using the bi-quadratic basis set in n–g
domain using Eqs. (23) and (24).

4. Evaluation of stream function

The fluid motion is displayed using the stream function w obtained from velocity components U
and V. The relationships between stream function, w ([19]) and velocity components for two
dimensional flows are
ow ow
U¼ ; V ¼ ; ð25Þ
oY oX
which yield a single equation
o2 w o2 w oU oV
þ ¼  . ð26Þ
oX 2 oY 2 oY oX
Using the above definition of the stream function, the positive sign of w denotes anticlockwise
circulation and the clockwise circulation is represented by the negative sign of w. Expanding the
P
stream function (w) using the basis set fUk gNk¼1 as w ¼ Nk¼1 wk Uk ðX ; Y Þ and the relation for U, V
from Eq. (16), the Galerkin finite element method yield the following linear residual equations for
Eq. (26).
S. Roy, T. Basak / International Journal of Engineering Science 43 (2005) 668–680 675

X
N Z
oUi oUk oUi oUk
Rsi ¼ wk þ dX dY
k¼1 X oX oX oY oY
X n Z Xn Z
oUk oUk
þ U k Ui dX dY  V k Ui dX dY ð27Þ
k¼1 X oY k¼1 X oX
The no-slip condition is valid at all boundaries as there is no cross flow, hence w = 0 is used as
residual equations at the nodes for the boundaries. The bi-quadratic basis function is used to eval-
uate the integrals in Eq. (27) and wÕs are obtained by solving the N linear residual Eqs. (27).

5. Results and discussion

Eqs. (10) and (14), (15) with boundary conditions (11) have been solved numerically using a
penalty finite element method. The computational domain consists of 20 · 20 bi-quadratic ele-
ments which correspond to 41 · 41 grid points. The bi-quadratic elements smoothly capture the
non-linear variations of the field variables which are in contrast with finite difference/finite volume
solutions available in the literature [10–12]. Numerical solutions are obtained for various values of
Ra = 103–105 and Pr = 0.2–10 with uniform and non-uniform heating of a vertical wall and the
bottom wall where the other vertical wall is cooled and the top wall is well insulated. The jump
discontinuity in Dirichlet type of wall boundary conditions at the corner point (see Fig. 1) corre-
spond to computational singularity. The Gaussian quadrature based finite element method
provides the smooth solutions at the interior domain including the corner regions.
To ensure the convergence of the numerical solution to the exact solution, the grid sizes have
been optimized and the results presented here are independent of grid sizes. Further, in order to
verify the accuracy of our numerical procedure, we have tested our algorithm based on the grid
size (41 · 41) for a square enclosure with a side wall heated and the results are in well agreement
with Mallinson and Vahl Davis [20] for Ra = 103–106. Comparisons are not shown here for the
brevity of the manuscript. The numerical results are displayed as stream functions and isotherm
contours in Figs. 2–7. Also, the heat transfer rate at the heated walls are presented in terms of
local Nusselt number in Figs. 8 and 9.
Stream function and isotherm contours for various Ra = 103–105 and Pr = 0.710 with uni-
form heating of a vertical wall and the bottom wall are displayed in Figs. 2–4. As expected due
to a hot vertical wall, fluids rise up along the side of the hot vertical wall and flow down along
the cold vertical wall forming a roll with clockwise rotation inside the cavity. As Ra increases from
103 to 105, the values of the stream function increase i.e., the flow rate increases. Fig. 2 shows that
the isotherm lines change its value smoothly form hot vertical wall to cold vertical wall for
Ra = 103. But for Ra = 105 in Fig. 3 due to enhanced convection from the hot vertical wall side
to the cold vertical wall side, the isotherm lines with greater values (h > 0.5) covers almost the en-
tire cavity. Comparative studies on Figs. 3 and 4 show that as Pr increases from 0.2 to 10, the
values of stream function and isotherms in the core cavity increase slightly and the effect of Pr
on velocity and temperature fields is very little as compared to the effect of Ra on those fields.
Figs. 5–7 present the stream function and isotherm contours for various Ra = 103105 and
Pr = 0.210 when the bottom wall and one vertical wall are non-uniformly heated. As seen in
676 S. Roy, T. Basak / International Journal of Engineering Science 43 (2005) 668–680

STREAM FUNCTION, ψ TEMPERATURE, θ


1 1

0.8 0.8

0.9

0.8
0.6 0.6

0.7

0.1
0.3
0.6

0.2
0.5
0.4
0.4 0.4

0.2 0.2

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1

3
Fig. 2. Contour plots for Pr = 0.2 and Ra = 10 (uniform heating case).

STREAM FUNCTION, ψ TEMPERATURE, θ


1 1

0.9
0.8 0.8
0.8

0.6 0.6
0.7

0.4 0.4

0.1
0.2
0.2 0.2 0.6 4
0.5 0.

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1

5
Fig. 3. Contour plots for Pr = 0.2 and Ra = 10 (uniform heating case).

STREAM FUNCTION, ψ TEMPERATURE, θ


1 1

0.9
0.8 0.8
0.8

0.7
0.6 0.6

0.4 0.4
0.6

0.2 0.2 0.5


0.1
3

0.4
0.

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1

Fig. 4. Contour plots for Pr = 10 and Ra = 105 (uniform heating case).

Figs. 2–4, uniform heating of the bottom wall and one vertical wall causes a finite discontinuity in
Dirichlet type of boundary conditions for the temperature distribution at one edge of the bottom
S. Roy, T. Basak / International Journal of Engineering Science 43 (2005) 668–680 677

STREAM FUNCTION, ψ TEMPERATURE, θ


1 1
0.3 0.4
0.5
0.8 0.8

0.3
0.

0.2
0.4
6
0.6 0.6

0.1
0.8

0.7
0.4 0.4

0.2 0.2 0.6


9 0.7
0.4
0.9
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1

3
Fig. 5. Contour plots for Pr = 0.2 and Ra = 10 (non-uniform heating case).

STREAM FUNCTION, ψ TEMPERATURE, θ


1 1 0.3
5 0.6
0.
0
0.8 0.8
0.7

0.6
0.6 0.6
0.8

0.4 0.4
0.5
4
0.6 0.
0.2 0.2
0.23
0 0.

0.1
0.
5 0.7
0.
3 0.8
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1

Fig. 6. Contour plots for Pr = 0.2 and Ra = 105 (non-uniform heating case).

STREAM FUNCTION, ψ TEMPERATURE, θ


1 1
0 5
0. 0.6
0.8 0.8
0.6
0.7
0.8

0.6 0.6

0.5
0.4 0.4
0.1

0.4
0.2 0.2
0.3
2

0.
0.

5 0.6
0.
3 0.8
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1

Fig. 7. Contour plots for Pr = 10 and Ra = 105 (non-uniform heating case).

wall. In contrast, the non-uniform heating removes the singularity at the edge of the bottom wall
and provides a smooth temperature distribution in the entire cavity. Results indicate that the
678 S. Roy, T. Basak / International Journal of Engineering Science 43 (2005) 668–680

15

5
Local Nusselt No., Nu 10 Pr = 0.2, Ra = 10 Pr = 10, Ra = 10
5

5
Pr = 0.2, Ra = 103

-5
0.2 0.4 0.6 0.8
Distance, X

Fig. 8. Variation of local Nusselt number with distance at the bottom wall for uniform heating (—) and non-uniform
heating (- - -).

9
Pr = 10, Ra = 105

Pr = 0.2, Ra = 105
6
Local Nusselt No., Nu

3
Pr = 0.2, Ra = 10

-3
0.2 0.4 0.6 0.8
Distance, Y

Fig. 9. Variation of local Nusselt number with distance at the left vertical wall for uniform heating (—) and non-
uniform heating (- - -).

values of stream function are almost similar in both uniform and non-uniform heating cases ex-
cept the formations of small stagnant fluid region attached to the corners of the heated vertical
wall for the non-uniform heating case. Further for Ra = 105, due to the present non-uniform heat-
ing, the isotherm lines with greater values (h > 0.5) covers approximately 70% of the cavity which
is less as compared to the uniform heating case.
Figs. 8 and 9 display the effects of Ra and Pr on the local Nusselt number at the heated bottom
and vertical walls. In case of uniform heating, the heat transfer rate is very high at right edge of
S. Roy, T. Basak / International Journal of Engineering Science 43 (2005) 668–680 679

the bottom wall due to the discontinuity present in the temperature boundary conditions at this
edge and heat transfer rate is almost uniform at the hot vertical wall. On the contrary, for
Ra = 103 with non-uniformly heated bottom wall and one vertical wall, the heat transfer rate in-
creases from edges towards the center with its maximum value at the center. Further for Ra = 105,
the present non-uniform heating produces a sinusoidal type of local heat transfer rate with its
minimum values at the edges. The physical reason of this type of behaviour is due to the higher
values of stream function (i.e., high flow rate) for Ra = 105 in the core of the cavity. Due to the
variations of Pr from 0.2 to 10, the local Nusselt number at the heated walls increases slightly as
seen in Figs. 8 and 9.

6. Conclusions

The main objective of the current investigations is to study the influence of uniform and non-uni-
form heating of the bottom wall and one vertical wall on flow and heat transfer characteristics due
to natural convection within a square enclosure. The penalty finite element method helps to obtain
smooth solutions in terms of stream function and isotherm contours for uniform and non-uniform
heating cases with wide ranges of Pr and Ra. In case of uniform heating, the heat transfer rate is
very high at the right edge of the bottom wall and almost uniform at the rest part of the bottom wall
as well as at the hot vertical wall. In contrast, for the case of non-uniform heating the heat transfer
rate is minimum at the edges of the heated walls and the heat transfer rate reaches its maximum
value at the center of both the heated walls. It is interesting to note that a sinusoidal type of vari-
ations in heat transfer rate is observed for non-uniform heating with Ra = 105.

References

[1] S. Ostrach, Natural convection in enclosuresAdvances in Heat Transfer, Vol. VIII, Academic press, New York,
1972, pp. 161–227.
[2] S. Ostrach, Low-gravity fluid flows, Ann. Rev. Fluid Mech. 14 (1982) 313–345.
[3] S. Ostrach, Natural convection in enclosures, ASME Trans. J. Heat Transfer 110 (1988) 1175–1190.
[4] B. Gebhart, Buoyancy induced fluid motions characteristics of applications in technology: The 1978 Freeman
Scholar Lecture, ASME Trans. J. Fluids Eng. 101 (1979) 5–28.
[5] C.J. Hoogendoorn, Natural convection in enclosuresProc. Eighth Int. Heat Transfer Conf., Vol. I, Hemisphere
Publishing Corp., San Francisco, 1986, pp. 111–120.
[6] J. Patterson, J. Imberger, Unsteady natural convection in a rectangular cavity, J. Fluid Mech. 100 (1980) 65–86.
[7] V.F. Nicolette, K.T. Yang, J.R. Lloyd, Transient cooling by natural convection in a two-dimensional square
enclosure, Int. J. Heat Mass Transfer 28 (1985) 1721–1732.
[8] J.D. Hall, A. Bejan, J.B. Chaddock, Transient natural convection in a rectangular enclosure with one heated side
wall, Int. J. Heat Fluid Flow 9 (1988) 396–404.
[9] J.M. Hyun, J.W. Lee, Numerical solutions of transient natural convection in a square cavity with different sidewall
temperature, Int. J. Heat Fluid Flow 10 (1989) 146–151.
[10] T. Fusegi, J.M. Hyun, K. Kuwahara, Natural convection in a differentially heated square cavity with internal heat
generation, Numer. Heat Transfer Part A 21 (1992) 215–229.
[11] J.L. Lage, A. Bejan, The Ra–Pr domain of laminar natural convection in an enclosure heated from the sider,
Numer. Heat Transfer Part A 19 (1991) 21–41.
680 S. Roy, T. Basak / International Journal of Engineering Science 43 (2005) 668–680

[12] J.L. Lage, A. Bejan, The resonance of natural convection in an enclosure heated periodically from the side, Int. J.
Heat Mass Transfer 36 (1993) 2027–2038.
[13] C. Xia, J.Y. Murthy, Buoyancy-driven flow transitions in deep cavities heated from below, ASME Trans. J. Heat
Transfer 124 (2002) 650–659.
[14] W.J. Minkowycz, E.M. Sparrow, G.E. Schneider, R.H Fletcher, Handbook of Numerical Heat Transfer, Wiley,
New York, 1988.
[15] T. Nishimura, A. Wake, E. Fukumori, Natural convection of water near the density extremum for a wide range of
Rayleigh numbers, Numer. Heat Transfer Part A 27 (1995) 433–449.
[16] T. Basak, K.G. Ayappa, Influence of internal convection during microwave thawing of cylinders, AIChE J. 47
(2001) 835–850.
[17] A. Asaithambi, Numerical solution of the Falkner–Skan equation using piecewise linear functions, Appl. Math.
Comput. 81 (2003) 607–614.
[18] J.N. Reddy, An Introduction to the Finite Element Method, McGraw-Hill, New York, 1993.
[19] G.K. Batchelor, An Introduction to Fluid Dynamics, Cambridge University Press, 1993.
[20] G.D. Mallinson, G.D. Vahl Davis, Three-dimensional natural convection in a box: a numerical study, J. Fluid
Mech. 83 (1977) 1–31.

You might also like