h1. J. Hca, Mom Trmfir. Vol. 36. No. 6. pp. IM)I-I608. 1993 OOl7-9310/93$6.00+0.

00
Primed in Great Britain i 1993 Pergamon Pras Ltd

Mixed convection in a driven cavity with

a stable vertical temperature gradient
REIMA IWATSU
Institute of Computational Fluid Dynamics. l-22-3 Haramachi. Meguro. Tokyo 152. Japan

JAE MIN HYUN

Korea Advanced Institute of Science and Technology, P.O. Box 150, Cheongryang, Seoul. Korea

and

KUNIO KUWAHARA
The Institute of Space and Astronautical Science, Yoshinodai. Sagamihara, Kanagawa, Japan

Abstract-Extensive parametric studies are made of Row and heat transfer of a viscous fluid contained in
a square cavity. Flow is generated by the top horizontal boundary wall, which slides in its own plane at
constant speed. A stabilizing externally-applied vertical temperature gradient is imposed on the system
boundaries: the top wall is maintained at a higher temperature than the botom wall. Numerical solutions
to the Navier-Stokes equations are secured over broad ranges of the parameters, 0 < Ru i IO”.
0 < Re $ 3000. Pr c O(I). aspect ratio -O(I). Systematical evaluations of the numerical results are
carried out to ascertain the relative importance of natural and forced convections. Representative plots
illustrating the velocity and thermal fields are presented. These clearly identify the major dynamic elements
in various regimes of the parameter spaces. When Gr/Re’ < I. the flow features are similar to those of a
conventional driven-cavity of a non-stratified fluid. In the bulk of the interior, fluids are well mixed and
temperature variations are small. When Gr/Re’ >> I, much of the middle and bottom portions of the cavity
interior is stagnant. In these regions. isotherms are nearly horizontal. and vertically-linear temperature
distributions are seen. By inspecting separate plots of i;T,&s and - PrRe.vT, contributions to total heat
transfer by conduction and by convection are assessed in a quantitative manner. The Nusselt number at
the top wall is calculated, and these results indicate the intensification of heat transfer as Gr/Re’ K I.

1. INTRODUCTION however, reveals that studies of mixed convection

within a closed cavity are relatively scarce.
FLOW OF a viscous fluid contained in a rectangular Recently, Mohamad and Viskanta [IO] reported the
cavity, with one of its boundary walls sliding at results of numerical investigations of mixed con-
constant speed U,, constitutes an appealing bench- vective flow in a shallow cavity (A << I). It should be
mark problem in numerical exercises. The geometry noted that, in their analysis, the temperature at the
is straightforward and the boundary conditions are top sliding surface, TT, was lower than that at the
regular. The flow is characterized by two principal stationary bottom surface, r,. The two vertical side
non-dimensional parameters, i.e. the Reynolds num- walls were thermally insulated. This created a gravi-
ber, RE = C/,h/v, and the cavity aspect ratio, A = tationally unstable flow configuration. The major
h/W, in which h and Ware height and width of the elements of velocity and thermal characteristics
cavity, and v is the kinematic viscosity of the fluid. were exhibited in the ranges of 200 < RE < 1000
Results of extensive numerical computational studies and 500 < 1Ral < IO“. Here, the buoyancy effect is
are available in the literature [ l-91, and the consistency measured by the Rayleigh number JRal, and the value
between these numerical predictions and laboratory of Ra may be interpreted to be negative as
experiments [3-51 has been explored [8,9]. AT = T, - T, < 0. In particular, the computed results
The introduction of a temperature gradient into the of this study illustrated the changes in the basic
cavity boundaries produces an additional significant character of a gravitationally unstable configuration
dynamical ingredient. The thermal non-homogeneity in a shallow cavity as Re and Ra encompassed the
gives rise to buoyancy and this, in turn, impacts upon above-stated ranges.
the coupled fields of velocity and temperature in As is evident in the non-dimensionalized governing
the cavity. An understanding of this mixed forced- equations, the fluid flow and accompanying heat
natural convective flow is of value from the standpoint transfer are characterized principally by four dimen-
of basic fluid dynamics as well as in practical engin- sionless parameters, i.e. Re, Gr, A and Pr. In the
eering applications. A survey of the relevant literature, ensuing discussions, the aspect ratio is fixed to be

1601
1602 R. IWhTSU (‘1 (I/.

n = I .O (a square cavity). and the Prandtl number is

Y
set at PI = 0.71 in most cases. This will allow cvalu-
ations of the explicit influences of forced convection
t U=l T=l
(represented by Re) and of natural convection (typi-
1
fied by G,). The ratio Gr/Re’ emerges as a measure
of the relative imporlance of natural convection to
forced convection. 2L-J
As pointed out earlier. the numerical work of an
Mohamad and Viskanta [IO] was concerned with the
cases when the externally-applied temperature diffcr-
encc was negative (AT < 0). i.c. the bottom wall was
heated. It is now proposed in the present endeavdr i
to cxaminc the flow and heat transfer properties in 0 T-O 1 X

the cavity when the externally-imposed tempera- FIG. I. Flow configuration and coordinate system.
ture diffcrcncc leads to a gravitationally-stable con-
figuration. Specifically. the top surface wall. which
slides at constant velocity I/,,, is maintained at tem-
perature T,-, which is higher than the temperature at
the stationary bottom wall, 7,. i.e. AT = T-,-- TR > 0. pressible Navier-Stokes equations. The customary
In contrast to the gravitationally unstable con- Boussinesq approximation is employed. These equa-
figuration of Mohamad and Viskanta [IO]. the system tions, expressed in properly non-dimensionalized
under present study will sustain a state of rest if the form, are
top surface wall is stationary. In this situation, heat
div V = 0 (1)
transfer is entirely based on the conductive mode.
Howcvcr. the sliding motion of the top surface wall SV
- +(V*grad)V = -gradp+Rr-’ AV
brings about fluid flows inside the cavity. Of primary at
interest is the effect of the stable stratification on the
global flow and thermal fields. In practical terms, the +Gr*Re-’ Te (2)
augmentation of heat transfer rate, above that given tlT
by conduction alone, which is brought forth by forced Z +(V*grad)T= (Pr*Re)- ‘AT, (3)
convection, is of great significance. Stated alterna-
tively, under a given imposed stable vertical tem- where V = (u, u) is the velocity vector, p the pressure,
perature difference across the system, AT = T-r-T,,, T the temperature and e = (0, I) a unit vector in the
the present investigation aims to delineate the heat direction of the buoyancy force. In the above, the
transfer enhancement and the accompanying velocity relevant dimensionless parameters are identified as:
field, which are achieved by forcing the top wall to Re = U,,LIv
slide in its own plane. In general, it is expected that,
for a given value of Re. the stable stratification tends RN = bAT/r’/rti
to inhibit vertical motions. Consequently, effective Pr = v/ti
convective heat transfer is anticipated to decrease as
the stable system-wide stratification intensifies. Gr = Ra/Pr.
In the present account, comprehensive numerical In accordance with the problem description, the
solutions were acquired to the complete Navier- boundary conditions are stated as :
Stokes equations using very fine grid meshes. Both
the ranges of ] RN[ and Re were extended to cover the V=(l,O) at )‘=I, O<x<l,
broader spectra of flow characteristics. Details of flow V=O at s=O,l, O<~j<l
and thermal field structures will be depicted, and
quantitative assessments will be made to portray the and y=O, O,<s< I,
salient features in various parameter settings. The T= 0 at ~3= 0, 0 < I < I,
overriding motivation of this study is, as ascertained
earlier. to demonstrate the substantial increase in T&l at )‘=I, O<x,<l,
global heat transfer above that given by the con- i3T
ductive mode. This is accomplished by forced con- %=O at s=O and s= I, O<y,< I.
vection, which is caused by the sliding top surface
wall. The above system of equations is solved by resort-
ing to a well-established finite difference numerical
procedure. This numerical model represents an
2. PROBLEM FORMULATION
amended version of the MAC method [I I-141. Exten-
The definition sketch of the problem is displayed in sive model verifications have been documented as to
Fig. 1. The governing equations are the incom- the reliability and accuracy of this numerical
 Mixed convection in a driven cavity 1603

technique. For complete details, the reader is referred driven cavity flow of a non-stratified fluid. At high
to the prior publications [15-l 71, Reynolds numbers. the emergence of boundary laycr-
In order to test the spatial grid convergence, cal- like flow features is apparent. These characteristics
culations were repeated using several different spatial have been scrutinized in detail in preceding pub-
resolutions, i.e. As = 8~3 = l/64. l/l28 and l/256 for lications [8,9].
Re = 100 and Gr = IO”; and A.u = A?‘= I1128 and As anticipated, the prornincncc of buoyancy effects
l/256 for Rc = 3000 and Gr = IO”. For these tests, is discernible throughout the cavity when Gr/
Ar = 0.005. 0.002 and 0.001 for As = A?, = l/64, Re’ > O(l). Figure 3. at Gr = IO”. cxcmplitics this
l/l28 and l/256. respectively. The outcome of these trend. In Fig. 3. curves (a) and (b) correspond to the
cxerciscs indicated that the discrepancies in the com- casts Gr/Rc >> I. Due to the dominant influcncc 01
puted velocity and temperature fields wcrc less than stabilizing buoyancy effects. the flows are almost stag-
I% nant in the bulk of the cavity interior. excluding the
In summary, the above-stated sensitivity tests portions close to the sliding top wall. This typifcs the
clearly demonstrated that the solutions are, to a fair flow structure which is stabilized by buoyancy-driven
degree of accuracy, independent of the grids and time natural convection. It is strcsscd again that the prcscnt
intervals adopted in the present study. Consequently. result is in contrast to the previous work of Mohamad
in the numerical computations, As and A,r wcrc and Viskanta [IO], which dealt with a gravitationally
chosen tither as I / I28 or I /256. unstable flow configuration. Curves (c) and (d) are
At the initial state, it is assumed that the fluid is characterized by Gr/Re’ < O(l): the qualitative
motionless and a linear temperature stratification prc- velocity profiles arc similar to those shown in Fig. 2.
vails throughout the whole domain. Calculations are for which Gr/ RcZ cc O( I ).
continued until steady motion is attained. The tran- The numerical results arc rearranged to demon-
sitory approach toward the steady state is not of stratc the change in basic Row character as G/
primary concern here and it will not be treated in the incrcascs at a fixed Re, as displayed in Fig. 4. The
prcscnt treatise. changeover of the principal flow pattern is evident,
as the criterion Gr/Re’ < O(I) is strongly satisfied
(curves (a) and (b)) or not (curve (c)).
3. RESULTS
The global flow and temperature fields arc depicted
As is immediately clear from the non-dimcn- in Figs. 5 and 6. When the buoyancy effects arc
sionalized governing equations, the ratio GI./RL" pro- relatively small. Gr/Re’ << I, the gross flow fcaturcs
vides a measure of the importance of buoyancy-driven of Fig. 5 are similar to those of a conventional mcch-
natural convection relative to lid-driven forced con- anically-driven cavity flow of a non-stratified fluid at
vection. If Gr/Re' cc I. the buoyancy effect is over- comparable values of Re. The main circulation fills the
whelmed by the mechanical effect of the sliding lid. entire cavity ; minor cells arc visible near the bottom
Therefore, as displayed in Fig. 2. the qualitative corners. The isotherms are clustered close to the bot-
character of flow is similar to the conventional lid- tom lid, which points to the cxistcnce of steep tem-

***tJMIE,tJd\X,VMIN,VMAX
***l -0.20370626 1.0000000 -0.24480307 0.16998208
***'2 -0.31979227 1.0000000 -0.44589245 0.29554427
***3 -0.37819362 1.0000000 -0.51784348 0.36577463
*x*4 -0.42407644 1.0000000 -0.55746049 0.41593158

FE. 2. Velocity profiles. (a) Profles of horizontal velocity, U, along the vertical center line. .v = 0.5. (b)
Profiles of vertical velocity, (1,along the horizontal midheight, J’ = 0.5. Gr = IO’. The Reynolds numbers
are:(a) Re = 100; (b) Re = 400;(c)Re = 1000; (d) Re = 3000.
1604

***UMIN,UMAX,VMIN,VMAX
***I -0.25125504 1.0000000 0.00000000E+00 3.59019265E-03
**x2 -0.26322627 1.0000000 -8.97862948E-04 7.23083690E-03
***3 -0.17038333 1.0000000 -8.77381563E-0'2 9.6930921lE-02
**x4 -0.40658975 1.0000000 -0.54972339 0.40160173

FIG. 3. Same as in Fig. 2 except for Gr = IO”

perature gradients in the vertical direction in this middle and bottom interior regions. The vertical tem-
region. In the bulk of the cavity except this localized perature stratification is substantially linear in the
area, however, the temperature gradients are weak. stagnant bulk of the interior regions. This reflects the
This implies that, due to the vigorous actions of mech- fact that heat transfer is mostly conductive in the
anically-driven circulations, fluids are well mixed ; middle and bottom parts of the cavity. Only in a
consequently, the temperature differences in much of relatively small region in the top portion of the cavity
this interior region are very small. On the other hand, are fluids comparatively well mixed; temperature is
when the buoyancy effects outweigh the effects of the fairly uniform in this region, where the mechanically-
sliding wall, i.e. Gr/Re’ >> I, the flow and temperature induced convective activities are appreciable.
fields are typified in Fig. 6. The interior circulation is A systematic representation of the vertical tem-
restricted to a small zone close to the sliding top lid. perature structures is portrayed in Fig. 7. Evidently,
The impact of the mechanically-driven top lid pene- all the temperature profiles should satisfy the bound-
trates only small distances toward the interior region. aryconditions:T=Oaty=OandT=laty=l.In
Obviously, the stable stratification inhibits vertical the limit when the top lid is stationary, Re + 0 or
motions; much of the fluid remains stagnant in the Gr/Re’ x I at a finite Re, the interior fluid is motion-

***UMIN,UMAX,VMIN,VMAX
***I -0.31979585 1.0000000 -0.44589722 0.29555106
***2 -0.30951858 1.0000000 -0.43180406 0.28371251
***3 -0.26322138 1.0000000 -8.9775212lE-04 7.24285096E-03

1.0

V
a.

0.0

FIG. 4. Same as in Fig. 2. Re = 400. The Grashof numbers are: (a) Gr = IO’; (b) 10J; (c) 10”
 Mixed convection in a driven cavity 1605

TR:MIN,RMAX,LMAX -0.11521769 1.58714270E-03 2.56162835E-04 X,Y:MIN,RMAX,LMAX

053307390 0.56809336 0.86770427 0.11673146 8.94941092E-02 8.56030583E-02 T,M
d,MAX 0.00000000E+00 0.99785382

FIG. 5. Plots of the fields of velocity (a) and temperature (b). Re = 10’. Gr = IO’.

less and the corresponding temperature distribution which the mechanically-induced activities are appreci-
approaches the linear profile obtained by the con- able.
duction solution. In the opposite case, when the buoy- The total vertical heat transport at a given height
ancy effect is minor, Gr/Re’ << 1, much of the tem- y = y may be expressed as
perature variations is achieved in narrow strips
adjacent to the top and bottom lids. In the middle (4)
portions of the cavity, the temperature changes are
very small ; as asserted earlier, these regions of near-
uniform temperatures correspond to the areas in In the above. the first and second terms denote the

STREAU LINE TEHPERATURE

FIG. 6. Same as in Fig. 5, except for Re = 1000, Gr = IO’.

I606 R. IWATSU et d.

FIG;. 7. Vertical tcmpcraturc prolilcsalongs = 0.5. Rc = 400. FIG. 9. Vertical temperature profiles along s = 0.5.
The values of Gr are: (a) Gr = IO’; (b) Gr = IO’; (cl Rc = 1000. Gr = IOh. The values of Pr arc: (a) Pr = 0.01 :
Gr = IO”. (b) Pr = 0.71 : (c) Pr = 10.0.

contributions attributable to conductive mode and file in the interior shows a zone of quite uniform
convective mode, respectively. By inspecting separate tcmperaturc in the top portion of the cavity. However.
plots of these two terms. the relative magnitude of in the bottom part of the cavity, the influence of the
each contribution may be appraised clearly. Figure 8 moving top boundary wall is relatively weak. The
exemplifies such plots. Curve (cf. for which Gr/Rc’ >> 1. temperature distribution in this area illustrates a
points to an ovcrwhclming dominance of conduction, vertically-linear distribution. which reflects the preva-
i.e. ?Tj?j* + I .O and - Pr Rc - rT - 0 over nearly the lence of the conductive mode in this region.
entire region of the cavity. On the other hand. when Utilizing the wealth of computed results. the total
Gr/Re’ < O(I). the substantial contribution of con- heat transfer attainable at the top surface wall is cal-
vectivc heat transport, shown by the magnitudes of culated. This is represented by the Nusselt number
-PrRe * I:T in the middle and upper portions of NLI, which is defined as
the cavity is manifest. This scrutiny of the separate
profiles of (4) reaffirms the qualitative trends discussed ds. (5)
in the above. I’= I
Additional runs were made to inquire as to the The exemplary results of these calculations are con-
effect of the Prandtl number Pr. Figure 9 exemplifies tained in Table I. Clearly, as Gr/Rr’ >> I, the domi-
the qualitative aspects of the global temperature fields
when Gr/Re’ - O(1). Clearly. when Pr << I, the tem- Table I. The Nussclt number Nu at the lop wall
perature distribution in the interior tends to the ver-
tically-linear profile. This is demonstrative of the pre- RC Gr = IO-’ Gr = IO’ Gr = IO’
vailing conductive heat transfer. When Pr is large. the
100 I .94 1.34 I .O’
upper part of the cavity interior is occupied by a 400 3.84 3.62 I .22
strong circulatory cell. which renders a region of well- 1000 6.33 6.29 1.77
mixed fluids. Therefore, the vertical temperature pro-

O.&j---J 0. 10.
-P,R,VT
FIG. 8. Vertical profiles of ar/ii, and - Pr RevTalong s = 0.5. Re = 100. The values of Gr are : (a) Gr = 10’; (b) Gr = IO’;
(c) Gr = 10”.
 Mixed convection in a driven cavity 1607

Nu

1
X

FE. 10. Local Nusselt number profles at the top and bottom walls. Profiles are along: (a) top wall.
J’ = 1.0: (b)bottom wall.!’ = 0.0. (a) Re = 400. Gr = IO’. Pr = 0.71. (b) Re = 400, Gr = 100. Pr = 0.71.

nance by conduction is obvious. and Ntr + 1.0. teristics arc similar to those of a conventional lid-
accordingly. The intensification of heat transfer, driven cavity of a non-stratified fluid. The isotherms
induced by mechanically-driven forced convection. is arc clustcrcd only in narrow strips adjacent to the top
apparent as Gr/Re’ K I. and bottom walls. In the bulk of the central regions
The variations of local Nussclt number at the top of the cavity. fluids are well mixed and temperature
and bottom walls are of intcrcst in the practical design variations arc very small.
of heat transfer equipment. Figure IO(a) shows the Prominent buoyancy-driven convective features arc
results (Re = 400. Gr = IO”. Pr = 0.71); the Ruid discernible when Gr/Re’ >> I. The direct effect of the
motions are generally subdued and heat transport in sliding top wall penetrates only small distances into
the interior is dominated by conduction. Conse- the interior region. Much of the middle and bottom
quently. Nu at the bottom wall is close to unity. At portions of fluids is stagnant. In these near-stagnant
the sliding top wall. owing to the presence of the regions. the isotherms are fairly horizontal and ver-
circulation near the top left corner, convcctivc heat tically-linear tempcraturc distributions prevail. Only
transfer is appreciable. However. in the vicinity of in small regions near the top sliding wall arc fluids
the top right corner, the fluid temperature is fairly well mixed, and zones of frtirly uniform temperature
uniform in the vertical direction. Figure IO(b) por- are noticeable.
trays the local Nusselt number variations (Re = 400, The influence of Pr over the flow characteristics is
G, = 100, Pr = 0.71) when convective activities are studied for casts at Gr/Re’ - O(l). When Pr is low,
dominant. The behavior of the Nusselt number pro- the temperature field in the interior tends to a
files at both boundary walls clearly points to the exis- vertically-linear profile throughout the entire cavity.
tence of substantial fluid motions, leading to enhance- However, when Pr is high, the upper part of the cavity
ment of heat transport throughout the cavity. represents a region of well mixed fluids. The tem-
This baseline information is also useful to designers perature field is fairly uniform in the upper portion
of practical heat exchangers. In addition, these nu- of the cavity. In the bottom part of the cavity. the
merical results will serve as reference materials against temperature distribution shows a vertically-linear
which further numerical or observational data may profile.
be compared. Further inspections of the numerical results are cor-
roborative of the qualitative features described in the
above.
4. CONCLUDING REMARKS

Numerical examinations are performed of mixed

convective flow and heat transfer in a driven-cavity of
REFERENCES
aspect ratio O(l). The effect of the mechanical drive
of the top wall is represented by Re. A stabilizing U. Ghia. K. N. Ghia and C. T. Shin, High-Re solutions
externally-imposed vertical temperature differential. for incompressible flow using the NavierrStokes equa-
tions and a multigrid method. J. Co~pitr. P/r.rs.48, 387-
AT ( = Top - Ttmm ) > 0, is enforced on the system 41 I (1982).
boundaries. The flow configuration is in contrast to R. Schreiber and H. B. Keller. Driven cavity flows by
that adopted in the previous study of Mohamad and efficient numerical technique. J. Corn/~. P/Iw. 49, 310-
333 (1983).
Viskanta, who considered shallow, bottom-heated
J. R. KosetT and R. L. Street, Visualization studies of
cavities. a shear driven three-dimensional recirculation flow, J.
When Gr/Re’ << I, the buoyancy effect is out- Fhih Engr!y 106, 21-29 (1984).
weighed by forced convection, and the flow charac- J. R. KoselT and R. L. Street, On end wall effects in a
1608 R. IWATsU et al.

lid-driven cavity flow, J. Fluids Engng 106, 385-389 I I. F. H. Harlow and J. E. Welch, Numerical calculation of
( 1984). time-dependent viscous incompressible flow of fluid with
5. J. R. KoselTand R. L. Street, The lid-driven cavity flow: free surface, PI7ys. Nuids 8.2 182-2 I87 (1965).
a synthesis of qualitative and quantitative observations, 12. K. Kawamura and K. Kuwahara, Computation of high
J. Fhids Engng 106, 390-398 (I 984). Reynolds number flow around a circular cylinder with
6. C. J. Freitas, R. L. Street, A. N. Findikakis and J. KoselT, surface roughness, AIAA paper 84-0340 (1984).
Numerical simulation of three-dimensional flow in a cav- 13. H. A. Van der Vorst. Bi-CGSTAB : a fast and smoothlv
ity, In!. J. Nun7er. Merl7. Fluids 5, 561-575 (1985). converging variant of Bi-CC for the solution of non-
7. C. Y. Perng and R. L. Street, Three-dimensional symmetric linear systems, SIAM J. Sci. Sfurisr. Compu7.
unsteady flow simulations: alternative strategies for a 13(2), 63 l-644 (1992).
volume-averaged calculation, hf. J. Numer. Merh. Fluids 14. S. Fujino, Visualization of convergence behavior of
9, 341-362 (1989). Bi-CGSTAB method, Proc. hr. Syn7p. on Compuler
8. R. Iwatsu, K. Ishii, T. Kawamura, K. Kuwahara and 1. Aritl7metric and Scientific Con7pururion (1991).
M. Hyun, Numerical simulation of three-dimensional 15. R. Iwatsu, J. M. Hyun and K. Kuwahara, Numerical
flow structure in a driven-cavity, Nuid Qn. Rex. 5, l73- simulation of torsionally-oscillating lid in a square
189 (1989). cavity, J. Nuid~ Engng 114, 143-151 (1992).
9. R. Iwatsu, J. M. Hyun and K. Kuwahara, Analyses of 16. R. Iwatsu, J. M. Hyun and K. Kuwahara, Convection
three-dimensional flow calculations in a driven cavity. in a differentially-heated square cavity with a torsionally-
Fluid Dyn. Res. 6,9 I - 102 ( 1990). oscillating lid, Inr. J. Hear Muss Tramfir 35, 1069-1076
IO. A. A. Mohamad and R. Viskanta, Combined surface (1992).
shear and buoyancy-driven convection in a shallow 17. R. Iwatsu. J. M. Hyun and K. Kuwahara, Flows driven
cavity, Proc. AIAA/ASME Tllern7opl7ysics and Hear by a torsionally-oscillating lid in a square cavity with
Tronsjer Co@ ( 1990). stable stratification, AIAA paper 91-0163 (1991).