hf. J. Hear Mass Transfer. Vol. 35, No. 8, pp. 1881-1892, 1992 0017-9310/92 s5.oo+o.

oo
Printed in Great Britain 0 1992 Pergamon PressLtd

Prandtl number effects on laminar mixed

convection heat transfer in a lid-driven cavity
M. K. MOALLEMI and K. S. JANG
Department of Mechanical and Industrial Engineering, Polytechnic University,
333 Jay Street, Brooklyn, NY 11201, U.S.A.

(Received I March 1991 and injnalform 14 August 1991)

Abstract-This paper considers the flow and heat transfer in a square cavity where the flow is induced by
a shear force resulting from the motion of the upper lid combined with buoyancy force due to bottom
heating. The work is motivated by the application in the production of plane glass where the glass sheet
is pulled over a bath of molten metal while being cooled and solidified. The numerical simulations,
therefore, are performed for two-dimensional laminar flow (100 < Re < 2200), and effects of small to
moderate Prandtl numbers (i.e. 0.01 < Pr < 50) on the flow and the heat transfer in the cavity are
investigated for different values of Richardson number. The temperature and the flow fields in the cavity
are calculated and presented to illustrate the strong influence of Prandtl number. The local and average
Nusselt numbers are also reported for different values of Re, Ri, and Pr.

1. INTRODUCTION aspect ratios of 0.5, 1 and 2. These simulations were

performed for fixed values of Re = 100 and Pr = 1,
MIXED-CONVECTION flow and heat transfer occur fre- with the three cavity walls at a constant temperature
quently in engineering and natural situations. One different from the temperature of the top driving lid.
important configuration is a lid-driven (or shear- Their results indicated that the Richardson number
driven) flow in a differentially heated/cooled cavity, (Ri = Gr/Re*) was a governing parameter of the prob-
which has applications in crystal growth, flow and lem. They particularly concluded that the flow pattern
heat transfer in solar ponds [l], dynamics of lakes [2], in the cavity was influenced by the buoyancy if the
and thermal-hydraulics of nuclear reactors [3]. Flow absolute value of Ri was greater than unity. However,
in cavities driven by a combined shear and buoyancy the effects of buoyancy on the heat transfer rate in the
force also arises in industrial processes such as food cavity were not characterized quantitatively.
processing, and float glass production [4]. Mixed convection heat transfer in a lid-driven
The lid-driven cavity flow has served over and over cavity was recently investigated by Prasad and Koseff
again as a standard test case for the evaluation of [ 161. In a series of experiments which were performed
numerical solution procedures for the Navier-Stokes on a cavity filled with water, the heat flux was mea-
equations [5-lo]. In these works, however, the effects sured at different locations over the hot cavity floor
of buoyancy were not addressed since isothermal for a range of Re and Gr. Their results indicated that
cavity flows were studied [5-S], or the boundary con- the overall (i.e. area-averaged) heat transfer rate was a
ditions represented thermally stable (i.e. stratified) very weak function of Gr for the range of Re examined
situations [9-lo]. Flow in thermally stratified and iso- (2200 < Re < 12000). The data were correlated by
thermal lid-driven cavities has been the subject of Nusselt number vs Reynolds number, as well as Stan-
extensive investigations at the Environmental Fluid ton number vs Reynolds number relations. The point
Mechanics Laboratory at Stanford University (e.g. that heat transfer is independent of buoyancy is rather
Koseff et al. [ 111,Koseff and Street [ 12,131 and Freitas surprising considering the range of Gr examined, i.e.
and Street [ 141). Through flow visualization, velocity 10’ < Gr < 5 x 109. It is among the objectives of this
measurements, and numerical simulations, recir- work to provide an explanation for this point.
culation flow patterns were characterized over a wide In some mixed convection literature [ 15-171 Rich-
range of Re and Gr. These revealed that three-dimen- ardson number has been reported to be the indicator
sional features, such as corner eddies near the end of the relative importance of buoyancy induced flow.
walls, and Taylor-Gortler-like longitudinal vortices, This observation is based on analyses and experiments
have significant influences on the flow for Reynolds on fluids with Prandtl numbers of order of unity, and
numbers as low as 3200. These works emphasized the cannot be treated as a universal rule. In fact, through
fluid mechanics aspects of the problem and did not a scale analysis of mixed convection flow over a ver-
deal with the heat transfer issues vigorously [ 11, 121. tical wall, Bejan [18] showed that the criterion for the
Torrance et al. [15] numerically investigated both transition from forced convection dominant flow to
thermally stable and unstable shear-driven flows natural convection dominant flow was not the same
(- lo6 < Gr < + 106) in cavities with depthwise for Pr > 1 and Pr < 1 fluids. The transition criterion
1881
1882 M. K. M~ALLEMIand K. S. JANG

NOMENCLATURE

A aspect ratio, and constant for equation ul. lid velocity [m s ‘1

(12) (x, y) horizontal and vertical coordinates [m]
B constant for equation (12) (X, Y) dimensionless coordinates, (x, y)/L.
9 gravitational acceleration [m ss’]
Gr Grashof number, gp(TH - T,)L3/v2
h heat transfer coefficient [W mm-* Km ‘1 Greek symbols
k thermal conductivity [W mm ’ Km ‘1 thermal diffusivity [m’ s-- ‘1
L cavity depth and width [m] ; coefficient of thermal expansion
NU Nusselt number, dimensionless heat W’l
transfer coefficient, - a@/8 Y = hL/k 0 dimensionless temperature,
P pressure [N mm ‘1 (T- Tc)I(Tn - Tc)
P dimensionless pressure, p/pUf V kinematic viscosity [m’ s- ‘I.
Pr Prandtl number, v/a
Re Reynolds number, U,L/v
Ri Richardson number, Gr/Re2 Subscripts
T temperature [K] C cold
(u, v) horizontal and vertical velocity F forced convection
components [m s- ‘1 H hot
(U, V) dimensionless velocity components, L lid
(u, r)lU, N natural convection.

is not yet validated for Pr << 1 fluids, due to the non- vertical walls of the cavity are insulated, and the lid
existence of experimental data for this range of Pr. and bottom surfaces are maintained at constant tem-
Recently, however, Mohamad and Viskanta [19] peratures, Tc and TH, respectively, with Tc < TH. The
reported on the effects of Pr on the onset of instability flow is assumed to be laminar, and the fluid properties
in a shallow lid-driven cavity heated from below. are constant except for the density in the buoyancy
Through a linear stability analysis, they found that Pr term (the Boussinesq approximation).
influenced the conditions for the initiation of the The governing equations of the problem are
mixed convection regime. expressions of conservation of mass, momentum, and
This paper considers the flow and heat transfer in thermal energy. Using L, U,, and (T, - T,-) as length,
a square cavity where flow is induced by a shear force velocity, and temperature scales, the governing equa-
resulting from the motion of the upper lid combined tions are non-dimensionalized to yield
with the buoyancy force due to bottom heating. The
work is motivated by the application in the production
au av
Tgy+ar=o
of plane glass where a glass sheet is pulled over a bath
of molten metal while being cooled to solidify. The
numerical simulations, therefore, are performed for ug+vgL -g+L&+!g) (2)
two-dimensional laminar flows. The effects of small
to moderate Prandtl numbers (i.e. 0.01 < Pr < 50) on
the flow and heat transfer in the cavity are investigated
for different values of Reynolds and Richardson num- U = UL, V = 0, Tc

bers. The temperature and flow fields in the cavity are

presented to illustrate the strong influence of Prandtl
number. The effects of Pr on the variation of the heat
transfer distribution over the cavity floor, and the aT aT
-_=O
al: z=o
driving lid are discussed, and correlations for the aver-
(/=v=o ll=V=O
age Nusselt number are reported.

2. PROBLEM FORMULATION

Fluid flow and heat transfer are modeled in a two-

dimensional square cavity of width L in which the top U = V = 0, T,,
(lid) surface moves across the cavity from left to right FIG. 1. Schematic diagram of the cavity and boundary
at a constant speed UL, as shown in Fig. 1. The two conditions.
 Laminar mixed convection heat transfer 1883

av av ap where i and i refer to the computational nodes, k is

uz+vay= -z
the iteration loop counter, $ is u, v, or T, and Nun
and Nut are area-averaged Nusselt numbers on the
lid and bottom surfaces, respectively. When the above
+~(~+~)+~o (3)
criteria were satisfied, the residual source of mass was
less than lo-’ for all the cases examined. The con-
ao ao i vergence of the local heat transfer rate over the bottom
++g>. (4)
UZ2+VZ=Re~r ax2 surface, as judged by the position and magnitude of
its relative maximum, was within 0.5% when the cri-
The governing parameters of the problem are Reyn- teria of equation (6) were satisfied. Depending on
olds, Grashof, and Prandtl numbers which are defined the input parameters, 1000 to 1800 iterations were
in the Nomenclature. The coupling between the heat required for the solution to converge, with cases with
transfer and the fluid flow in the cavity is through the lower Ri or higher Pr requiring a fewer number of
buoyancy term in the y-momentum equation, equa- iterations.
tion (3), which has Gr/Re’ = Ri as a dimensionless The numerical procedure was validated by per-
coefficient. forming simulations of isothermal flow in a square
The velocity and temperature boundary conditions cavity with a driving lid for Re = 400 and 1000. The
are shown in Fig. 1 and have the following dimen- results of these simulations were compared with the
sionless forms : calculations of Schreiber and Keller [7], and Thomp-
son and Ferziger [8]. These comparisons revealed bet-
atX=OandX= 1 U=V=O and g=O
ter than 95% agreement in the strength and position
atY=O U=O, V=O and O=l of the primary [7, 81 and secondary [7] eddies. The
solution procedure was also validated by comparing
atY=l U=l, V=O and O=O. the predicted heat transfer rates with the experimental
(5) results of Prasad and Koseff [16]. The predicted area-
averaged heat transfer rates agreed to within 5% of
3. METHOD OF SOLUTION the experimental data for Pr = 6.0, Re = 2200, and
The dimensionless governing equations are dis- Gr = 10’ and 4 x 10’.
cretized using the control volume approach with the
power-law scheme [20] for the calculations of the 4. RESULTS AND DISCUSSIONS
fluxes at the faces of the control volumes. The
SIMPLER algorithm [20] is used with the inertia The main objective of this investigation was to
relaxation method of ref. [3] to accelerate conver- study the effects of Pr on the flow and heat transfer
gence. To resolve the large velocity and temperature in a square cavity driven by a combined shear and
gradients in the boundary layers near the cavity walls, buoyancy force. Forty-six numerical simulations were
a nonuniform grid system was used in both X- and Y- performed for a range of Pr, i.e. Pr = 0.01, 0.1, 6.0,
directions. The position of the control surfaces of the 7.1, and 50. With four values for Re (i.e. Re = 100,
grid system were determined with an exponentially 500, 1000, and 2200) and four values of Gr (i.e.
clustering scheme. A series of grid sensitivity runs Gr = 104, lo’, 106, and lo’), these simulations
were performed on nonuniform grids ranging from covered a range of Ri from 0.01 to 10.
22 x 22 to 52 x 52, which suggested that grid inde- The predicted flow and temperature fields for
pendent results could be achieved using a 42 x 42 grid. Re = 1000, and Gr = lo4 (i.e. constant value of
The results of two nonuniform grids of 42 x 42 ana Ri = 0.01) and for three different values of Pr are
52x 52 for Re = 1000, Gr = 106, and Pr = 0.1 shown in Fig. 2. The flow fields for all the three cases
revealed differences less than 1% in the streamlines are essentially established by the shear induced by the
and isotherms. The solution procedure is iterative and moving lid. This may be examined by recalling the
requires initial guesses for all the dependent variables. general features of the lid-driven cavity flow in the
To facilitate the convergence of the solution for a absence of buoyancy. The fluid flow in a two-dimen-
given set of parameters, the converged solution of a sional lid-driven cavity is characterized by (a) a pri-
case with smaller Ri or Re was used as the initial mary recirculating eddy of the size of the cavity gen-
guess. Sensitivity tests have indicated that the final erated by the lid dragging the adjacent fluid ; (b) a
results were independent of the initial guess. secondary eddy formed in the apex of the vertical wall
For any set of input parameters, the solution was and the bottom surface (referred to as the downstream
considered converged if secondary eddy, DSE) as a result of frictional losses
and stagnation pressure ; (c) another secondary eddy
IV+‘G7A--V(i,j)l < 1o_3 (referred to as the upstream secondary eddy, USE)
Max Itf‘+‘(i, j)( formed in the upstream lower corner due to the nega-
tive pressure gradient generated by the primary cir-
and culating fluid as it deflects upward over the upstream
vertical wall ; and (d) a third secondary eddy which
1884 M. K. MOALLEMIand K. S. JANE

F&l 000, Ri-0.01, Pk7.1 Re=lOOO, Ri=O.Ol , Pr=7.1

71

Level STR Level TEMP

A 0.001 S 1.000
8 0.000 A 0.000
6 -0.006 Q 0.600
7 -0.016 6 0.700
6 -0.029 7 0.600
5 -0.039 6 0.600
4 -0.049 5 0.400
3 -0.069 4 0.300
2 -0.069 3 0.200
1 -0.080 2 0.100
1 0.006

R&l 000, f&0.01, Pr-1 Re=lOOO, RM.01, Pr-1

Lwd STR .wel TEMP

B 0.0007 B 1.000
A 0.0001 A 0.000
9 0.0000 9 0.600
6 -0.0063 6 0.700
7 -0.0166 7 0.660
6 -0.0207 6 0.600
6 -0.0269 5 0.460
4 -0.0400 4 0.306
3 -0.0692 3 0.200

2 -0.0604 2 0.100
1 -O.WOO 1 0.000

Red 000, RkO.01, PkO.1 Re=lOOO, RM.01, PkO.1

Lwd STR Level TEMP

B 0.0007 B 1.000
A 0.0001 A 0.000
9 0.0000 9 0.800
0 -0.0063 6 0.700
7 -0.0166 7 0.800
6 -0.0287 6 0.600
5 -0.0269 6 0.400
4 -0.0490 4 0.300
3 -0.0692 3 0.200
2 -0.0684 2 0.100
1 -0.oaOO 1 0.000

FIG. 2. Effect of Pr on the flow temperature fields in the cavity, for Re = 1000 and Gr = 104.

may develop on the vertical wall near the lid, for of the value of Pr. Moreover, the size, position, and
Re > 1000, as the lid draws fluid from the comer into strength of the secondary eddies are almost identical
the lid boundary layer [l l-131. All these features are for the three cases presented in this figure. These con-
present in the flow fields shown in Fig. 2, regardless firm the established notion that for Ri << 1, the flow
 Laminar mixed convection heat transfer 1885

in the cavity is defined by the shear action of the heat transfer is mostly via convection in the boundary
moving lid [ 151, and thus, is independent of Pr which layers, the core of the cavity is isothermal, and the
defines the temperature and buoyancy variations in buoyancy effects are only significant near the cavity
the cavity. wall. And these together result in the enlargement of
The temperature distributions in the cavity, shown the DSE, and the elimination of the USE. For
on the right side of Fig. 2, exhibit strong dependence Pr << 1.O, the heat transfer in the cavity is mostly due
on Pr. These indicate that the thermal boundary layers to conduction, resulting in a rather gradual variation
over the bottom and the lid surfaces thicken as Pr is of temperature in the cavity, and therefore, a small
reduced. The large isothermal region which extends buoyancy field. In the bottom flow field in Fig. 4, the
over most of the cavity for Pr = 7.1, shrinks with presence of USE and upper secondary eddies in the
decreasing Pr, and disappears for Pr < 0.1. The cavity is the result of weak buoyancy or, in other
strong dependence of the temperature distribution on words, strong shear-induced recirculation.
Pr does not reflect on the flow pattern in the cavity The flow and temperature fields for Ri = 4
which again demonstrates the insignificant con- (Re = 500 and Gr = 106) illustrate similar effects and
tribution of buoyancy in the flow development for interactions, as shown in Fig. 5. Comparing the flow
Ri cc 1. fields in Figs. 4 and 5 shows the influence of a change
Figure 3 presents the flow and temperature fields in shear at constant Gr for different values of Pr. In
calculated for Re = 500, Gr = 10’ (i.e. constant the top rows (Pr = 7.1), a decrease in Re results in
value of Ri = 0.4), and for Pr = 7.1, 1.0, and 0.1. The significant change in the flow structure, i.e. the
flow fields of this figure indicate to a more significant enlargement of the DSE. Whereas, for Pr = 0.1, a
contribution of buoyancy in defining the flow pattern decrease in Re causes a change in the strength of the
in the cavity, as compared with the flow fields of Fig. primary eddy, with little effect on the secondary eddies
2. The upstream secondary eddies have vanished, and or the temperature field. The influence of a change in
the downstream secondary eddies have changed size. buoyancy on flow and heat transfer in the cavity at
On the upstream wall, buoyancy assists the core flow constant lid shear may be revealed in a comparison
to make a sharp turn, and thus, the secondary eddy at of Figs. 3 and 5. For large values of the Prandtl
the upstream corner disappears. On the downstream number, i.e. Pr 2 1.0, an increase in Gr results in
wall, however, buoyancy is opposing the core flow significant change in the secondary eddies, as well as
and causes the downcoming boundary layer to detach the primary eddy. This is due to the fact that for this
from the wall at a larger Y, resulting in a larger eddy. range of Pr, the flow in the cavity is boundary layer
The effects of Pr on the temperature distribution in type. For Pr < 1.0, however, the cavity flow is more
the cavity in this figure are similar to those of Fig. 2, of a core flow type, and an increase in buoyancy results
i.e. large isothermal core region for large Pr, and very in a stronger primary eddy and this manifests in the
thick thermal boundary layers for small Pr. For the formation of an upper secondary eddy on the vertical
cases of Fig. 3, however, a change in Pr results in a wall near the lid, and the resurgence of the USE.
change in the flow pattern in the cavity, confirming The variations of local Nusselt number over the lid
the important role of buoyancy in defining the flow and bottom surfaces for different values of Re, Gr,
field in the cavity. For large values of Pr, since the and Pr are presented in Figs. 6-10. Over the cold lid
core is isothermal, the influence of buoyancy is limited surface, the heat transfer rate generally drops in the
to the boundary layers, and appears in the form of direction of the lid motion due to the formation and
reduction in the size or position of the secondary growth of the boundary layer over this surface. Over
eddies. For small values of Pr, the effects of buoyancy the bottom hot surface, however, the variation of the
manifest in the boundary layers, and also in the core local heat transfer rates is more complex, and exhibits
of the cavity. Therefore, a reduction in Pr will result local maxima and minima which are defined by the
in a stronger primary eddy while affecting the sizes action of the primary and secondary eddies on this
and positions of the secondary eddies, as well. surface. The local NuH has a relative maximum at a
The flow and temperature fields in the cavity for point that cold fluid impinges on the cavity floor. The
Re = 1000 and Gr = lo6 are presented in Fig. 4. This drop in the heat transfer subsequent to an impinge-
figure shows that for Ri = 1.0 the fluid flow in the ment point is due to the formation and growth of
cavity is established by a relatively balanced inter- boundary layers over the cavity floor. The points of
action of the two driving mechanisms involved, with relative minimum NuH correspond to the points of
a strong dependence on Pr. For Pr equal to unity, the detachment of these boundary layers from the cavity
balance of the shear and buoyancy effects is mani- floor. These may be examined by referring to the cor-
fested in the formation of two eddies of almost equal responding flow fields presented in Figs. 2-5.
size. The upper eddy is driven by the moving lid, and In Figs. 6 and 7, for which Ri < 1, the local and
encapsulates an isothermal core. This is in fact the area-averaged heat transfer rates over the lid as well
primary eddy which is deformed due to the opposing as the floor surfaces increase with an increase in Pr. This
action of the buoyancy. The eddy in the lower half of trend is consistent over both surfaces since neither the
the cavity is driven by buoyancy, and is the enlarged dashed lines nor the solid lines cross each other. For
form of the DSE. For Pr greater than unity, the these cases, it may be concluded that the heat transfer
1886 K. MOALLEMI and K. S. JANG

R6=500, Ri-0.4, Pe7.1 Re=500, Rb0.4, Pb7.1

LWel TEMP
B 1BOO
LWsl STR
A o.QOO
A o.OOo7
9 0.800
9 o.moo
6 0.700
6 -0.0083
7 0.600
7 -0.0186
6 OSQO
6 -0.0267
6 0.400
5 -0.0389
4 0.300
4 -0.0490
3 0.200
3 -0.06Q2
2 O.?OQ
2 -0.06Q4
1 -o.OwJ 1 0.000

AedjOO, Rk0.4, Pr=l Re=500, Rir0.4, Pr=l

t.emI TEMP
Level STR B 1 .OOo
ii 0.0030 A 0.900
9 O.OQfO e 0.800
8 0.0000 a 0.700

7 -0.0039 7 0.600
6 -0.0127 6 0.500
5 -0.0304 5 0.400
4 -0.0461 4 0.300
3 -0.0659 3 0.200
2 -0.0636 2 0.100
1 -0.1013 1 0.000

Re&OO, fGO.4, PrPO.1 Re=500, RidI.4, PraO.1

Lwel TEMP
S 1.QOO
Level STR
A WOO
A O.Oof5
0 0.0000 9 o.wO
8 -0.0020 B 0.700

7 -0.0196 7 0.600
6 -0.0371 6 0.600

5 -0.0547 5 0.400

4 -0.0722 4 0.300
3 0.200
3 -0.0898
2 0.100
2 -0. f 073
1 0.000
1 -0.1249

FIG. 3. Effect of Pr on the flow and temperature fields in the cavity, for Re = 500and Gr = IO’.

in the cavity is defined by the shear flow that dictates as the positions of maximum and minimum heat
the primary recirculation in the cavity, since the form transfer rates do not noticeably change with a change
of variation of Nut along the lid does not change with in Pr.The small change in the local variation of Nut.,
Pr.The effect of buoyancy on the local heat transfer with Pr is due to the weak influence of buoyancy on
rate from the cavity goor is recognized to be minimal, the secondary eddies.
 Laminar mixed convection heat transfer 1887

F&l 000, Ri=l , Pe7.1 Red 000, Rid, Pr-7.1

Loud STR Level TEMP

B 0.006 B 1.000
A 0.006 A 0.606
9 0.000 9 0.660
6 -0.010 6 0.700
7 -0.020 7 0.600
6 -0.031 6 0.500
5 -0.041 5 0.400
4 -0.052 4 0300
3 -0.062 3 0.200
2 -0.073 2 0.100
1 -0.063 1 0.000

Re=lOOO, Rid , Pr=l Re=lOOO, Ri=l , Pr=l

Level STR Level TEMP

0 0.0435 B 1.060
A 0.0312 A 0.600
9 0.0160 6 0.600
6 0.0067 6 0.700
7 -0.0056 7 0.606
6 -0.0176 6 0.W
5 -0.0302 5 0.400
4 -0.0425 4 0.306
3 -0.W 3 0.206
2 -0.0671 2 0.106
1 -0.0764 1 0.006

Re=lOOO, RI=1 , PkO.1 R-1 000. Ri-1, PbO.1

Level sm
B 0.001 B 1.060
A 0.000 A 0.600
9 -0.011 9 0.606
6 -0.032 6 0.700
7 -0.063 7 0.606
6 -0.073 6 0.600
5 -0.064 5 0.406
4 -0.115 4 0.300
3 -0.136 3 0.200
2 -0.156 2 0.100
1 -0.177 1 0.006

FIG. 4. Effect of Pr on the flow and temperature fields in the cavity, for Re = 1000 and Gr = 106.

In Figs. 8 and 9 (Ri > l), the variations of Nuu and correspond to the points of cold flow impingement on
Nut illustrate stronger dependence on Pr, indicating the floor, and points of boundary layer detachment
a more significant contribution of buoyancy to from the floor, respectively. In these figures, the effects
heat transfer and fluid flow in the cavity. For Nuu of Pr on Nu, and Nut may be noted to be both
variations, the points of maxima and minima again quantitative (i.e. mostly higher heat transfer rates for
1888 M. K. MOALLEMI and K. S. JANC

Re=!iOO, Rid, PM.1 F&500, Ri54, Pk7.1

lmd STR Loud TEW

A 0.0477 0 1.000
S 0.0304 A 0900
8 0.0291 S 0.000
0.0199 8 0.700
7
6 0.0106 7 0.600
5 0.0013 6 0.500
4 -0.0080 5 0.400
3 -0.0266 4 0.300

2 -O.O4!ii 3 0.200

1 -0.0637 2 0.100
1 0.000

RedOO, Rk4, Pr-1 Re=500, Ri=4, Pr=l

l.md STR Level TEMP

B 0.099 B 1.000
A 0.057 A 0900
0 0.037 0 0.800
8 0.016 B 0.700
7 0.001 7 0.800
6 0.000 6 0.600
5 -0.004 5 0.400
4 -0.025 4 0.300
3 -0.046 3 0.200
2 -0.066 2 0.100
1 -0.087 1 0.000

Re500, Ri-4, PhO.1 Re=500, Ri=4, PrmO.1

TEMP
B 1.000
Lwal STR A OS00
S 0.0100 0 0.800
8 0.0004 8 0.700
7 0.0000 7 0.600
6 -0.0064 6 0.5SO
5 -0.t we 5 0.400
4 -0.1725 4 0.300
3 -0.2261 3 0.200
2 -0.2796 2 0.100
1 -0.3332 1 O.ooO

FIG. 5. Effect of Pr on the flow and temperature fields in the cavity, for Re = 500 and Gr = 10’.

larger values of Pr) and qualitative (i.e. changes in the effects are felt by both the primary and the secondary
positions of maximum and minimum Nuu with a eddies, particularly for cases with Pr 2 1.O.
change in Pr). In either sense, however, the influence The inftuence of Pr on the variation of the area-
is more pronounced for larger Pr. This confirms the averaged heat transfer rate with Ri, for different values
earlier observations that for Ri 3 1.0 the buoyancy of Re, is presented in Fig. 10. The experimentat results
 Laminar mixed convection heat transfer 1889

0.00 0.25 0.50 0.75 1.00

X
FIG. 6. Variations of the local Nu over the cavity floor (solid FIG. 9. Variations of the local Nu over the cavity floor
lines), and the lid (dash lines), for Re = 1000 and Gr = 10’ (solid lines), and the lid (dash lines), Re = 500 and Gr = lo6
(Ri = 0.01). (Ri = 4.0).

0.00 0.25 0.50 0.75 1.oo

X 1
0.01 0.1
Ri 1 10
FIG. 7. Variations of the local Nu over the cavity floor
(solid lines), and the lid (dash lines), Re = 500 and Gr = lo5 FIG. 10. Effect of Pr on the variation of averaged Nu with
(Ri = 0.4). Ri.

80
Re = 500, only 3% above the conduction limit of
unity. The results also illustrate that up to a minimum
60
level of buoyancy, the heat transfer rate in the cavity
is essentially via forced convection and independent
of Ri. The departure from the forced convection heat
Nu
40 transfer- mechanism to a mixed convection one, for
which Nu = f (Re, Rl), is noted to be a function of Pr.
Prasad and Koseff [ 161have correlated their exper-
20 imental results as
-
Nu A’.’
cc ~eO.‘8 Ri-0.02
(7)
0 where A is the depthwise aspect-ratio of the cavity.
0.00 0.25 0.50 0.75 1.oo
X This correlation should be quite useful for design
FIG. 8. Variations of the local Nu over the cavity floor (solid applications, as it has a simple form and spans four
lines), and the lid (dash lines), Re = loo0 and Gr = lo6 decades of Ri (i.e. 0.1 < Ri < 1000). It does not,
(Ri = 1.0). however, provide any clues to the interaction of the
two heat transfer mechanisms involved. Moreover,
of Prasad and Koseff [16], for water (Pr = 6.0) at the experimental conditions of ref. [16] correspond to
Re = 2200 are also shown in the figure, which com- 2200 < Re < 12000 and 10’ < Gr < 5 x lo’, which
pare well with the results of simulations performed include laminar and turbulent regimes for both the
for the same set of parameters. The figure indicates a forced and natural convection mechanisms. The cast-
rapid drop of convective heat transfer in the cavity ing of the data for the two flow regimes into one
with a decrease in Pr. This
- drop which is also a func- correlation had resulted in a very weak dependence of
tion of Re brings the Nu to 1.03 for Pr = 0.01 and the heat transfer rate on buoyancy, which may be
1890 M. K. MOALLEMIand K. S. JANG

justified for turbulent flows but is an anomaly for These are in agreement with the results for heat trans-
laminar flows. fer over a flat plate which are proportional to Pr”‘l
It has now become a common practice in mixed and Pr’/3 for Pr < 1 and Pr > 1, respectively [22].
convection literature [Zl] to correlate heat transfer To find the correct form of variation of NU with Ri
results by an expression of the form and Pr for the mixed and natural convection regimes,
it is first noted that Ri is not the correct scaling par-
;\;i;” = Nu”+Nu”
F- N (8)
ameter for both Pr > 1 and Pr < 1 conditions, as the
where subscripts F and N refer to pure forced and curves for Pr < 1 in Fig. 11 have knees at around Ri
natural convection, respectively. The value of the equal to unity, while other curves in the figure seem
exponent n and the sign between the two terms on the to have departures from pure forced convection mech-
right-hand side of the equation depend on the flow anism at values of Ri greater than unity. This is in
configuration and geometry. ____-The correlations for the agreement with the scaling analysis of Bejan [18] for
average Nusselt numbers Nup and NuN are determined mixed convection heat transfer over a vertical wall,
with clues from the existing correlations for similar which suggests use of different scaling criteria for tran-
geometries. Equation (8) yields the correct form of sition from forced convection to natural convection
dependence on the governing parameters of the prob- regimes. These criteria are
lem in the two limiting cases (i.e. forced convection as
for Pr < 1 fluids
Gr -+ 0, and natural convection as Re -+ 0), a charac-
teristic that equation (7) does not have. < 0( 1) forced convection
$3114
To identify a correct form for the heat transfer > O( I) natural convection (10)

correlation, an expression similar to equation (8) is

sought. The Nusselt number is first scaled by Re’.‘, and for Pr > 1 fluids
as suggested by correlations for laminar forced con-
vective flow over flat plates [22]. Figure I1 indicates Ri ~4 < 0( 1) forced convection
that this scaling is appropriate as curves for different Pr “3
(6) natural convection ’ (11)
i > 0( 1)
values of Re collapse on one another, for the entire -
range of Ri examined. The heat transfer results for Figure 12 presents the variation of NU scaled with
Pr = 0.01 and Re = 500 do not follow a trend similar Re”,’ x Pr” against the scaling parameters of equations
to the other curves in the figure, and will not be con- (IO) and (1 l), depending on the range of Pr. The
sidered for finding the correlation. This is justified by exponent M also depends on the range of Pr as given
the fact that the heat transfer for this set of conditions by equation (9). This figure indicates the proper
is mostly via conduction (i.e. Nu = 1.03). In finding scaling of both the abscissa and ordinate, and shows
the form of dependence of Nt+ on Pr, two ranges are the correct forced convection and natural convection
considered, namely, Pr < 1 and Pr 2 I. Least-squares limits for both ranges of Pr. The Nu variations in the
fits of the data at Ri = 0.01, which may be regarded natural convection limit seem to be proportional to
as pure forced convection, provide the following the abscissa parameter raised to the 0.20 power, rather
relations : than the 114th power suggested by criteria of equa-
tions (10) and (11). From this figure it may be con-
cluded that the overall effect of buoyancy in the cavity
(9)
is to assist the heat transfer, and no opposing effects

“.“b.*..1”. Ri 1
FIG. 11. Variation of averaged i”iu scaled by Re”-*with Ri FIG. 12. Variation of heat transfer in the cavity as a function
for different values of Pr. of Re, Ri, and Pr (symbols are the same as those of Fig. 11).
 Laminar mixed convection heat transfer 1891

appears to be present for the range of parameters efficient numerical techniques, J. Comput. Phys. 49,3 I&
333 (1983).
examined.
8. M. C. Thompson and J. H. Ferxiger, An adaptive multi-
grid technique for the incompressible Navier-Stokes
5. CONCLUSIONS equations, J. Comput. Phys. 82,94-121 (1989).
D. L. Young, J. A. Liggett and R. H. Gallagher,
The flow and heat transfer is investigated in a bot- 9’ Unsteady stratified circulation in a cavity, ASCE J. The
tom heated lid-driven square cavity flow. The effects Engineering Mechanics Division 102(EM6), 1009-1023
of Prandtl number on the flow structure and heat (1976).
transfer in the cavity are studied for laminar ranges of 10. M. Morzynski and Cz. 0. Popiel, Laminar heat transfer
Re and Gr. The inibrence of buoyancy on the flow
in a two-dimensional cavity covered by a moving wall,
Namer. Heat Transfer 13,26>273 (1988).
and heat transfer in the cavity is found to be more l1 J. R. Kose.ff, R. L. Street, P. M. Gresho, C. D. Upson,
pronounced for higher values of Pr, if Re and Gr ’ J. A. C. Humphrey and W. M. To, A three-dimensional
are kept constant. The natural convection effects are lid-driven cavity flow : experiment and simulation, Proc.
always assisting the forced convection heat transfer, 3rd Int. Co@ on Numerical Method in Laminar and
Turbulent Flow, Univ. Washington, 8-11 August, pp.
and the extent of the contribution is a function of Pr
564-581 (1983).
and Ri. The average heat transfer in the cavity is 12. J. R. Koseff and R. L. Street, The lid-driven cavity flow:
correlated as a synthesis of qualitative and quantitative observations,
ASME .I. Fluicls Engng 106,390-398 (1984).
- J. R. Koseff and R. L. Street, Circulation structure in a
Nu = A Re’,’ Pr” stratified cavity flow, AXE J. ~y~~lic Engng 111,
334-3.54 (1985).
where A and B are constants, the exponents m and n t4. C. J. Freitas and R. L. Street, Non-linear transient
phenomena in a complex recirculating flow : a numerical
depend on Pr, as given by equations (9) to (1 l), investigation, Int. J. Numer. Methodr FIuia!s 8, 769-802
and q is a constant of the order of 0.20 to 0.25. (1988).
15. K. Torrance, R. Davis, K. Eike, P. Gill, D. Gutman,
A. Hsui, S. Lyons and H. Zien, Cavity flows driven by
REFERENCES buoyancy and shear, J. Fluid Mech. 51(2), 221-231
1. 6. K. Cha and Y. Jaluria, Recirculating mixed con- (1972).
vection flow for energy extraction, Int. j. Heat Mass 16. A. K. Prasad and J. R. Koseff, Combined forced and
Transfer 27.1801-1810 11984). natural convection heat transfer in a deep lid-driven
2. J. Imberger’and P. F. Hamblin, Dynamics of lakes, reser- cavity flow, in Heat Transfer in Convective Flows, HTD-
voirs, and cooling ponds, A. Rev. FIuid Mech. 14, 153- 10’7,pp. 155-162. ASME: New York (1989).
187 (1982). 17. J. R. Llovd and E. M. Soarrow. Combined forced and
3. F. J. K. Ideriah, Prediction of turbulent cavity flow free convection flow on-vertical surfaces, Znt. J. Heat
driven by buoyancy and shear, J. Mech. Engng Sci. 22, Muss Transfer 13,434438 (1970).
287-295 (1980). 18. A. Bejan, Convective Heat Transfer, pp. 141-146. Wiley,
4. L. A. B. Pilkington, Review lecture: The float glass New York (1984).
process, Proc. R. Sot. Lond., IA 314, 1-25 (1969). 19. A. A. Mohamad and R. Viskanta, Effects of the upper
5. R. K. Agarwal, A third-order-accurate upwind scheme lid shear on the stability of flow in a cavity heated from
for Navier-Stokes solutions at high Reynolds numbers, below, ht. J. Heat Mass Transfer 32,2155-2166 (1989).
AIAA-81-0112, Proe. 19th AIAA Aerospace Sciences 20. S. V. Patankar, Numerical Heat Transfer and Fluid Flow.
Meeting, St. Louis, MI, January (1981). McGraw-Hill, New York (1980).
6. U. Ghia, K. N. Ghia and C. T. Shin, High-Be solutions 21. F. P. Incropera and D. P. Dewitt, F~~mentals of Heat
for incompr~sible flow using Navier-Stokes equations and Mass Transfer, 3rd Edn. Wiley, New York (1990).
and a multigrid method, J. Comput. Phys. 48, 387411 22. W. M. Kays and M. E. Crawford, Convective Heat and
(1982). Mass Transfer, 2nd Edn. McGraw-Hill, New York
7. R. Schreiber and H. B. Keller, Driven cavity flows by (1980).

EFFET DU NOMBRE DE PRANDTL SUR LA CONVECTION THERMIQUE MIXTE

LAMINAIRE CONDUITE DANS UNE CAVITE PAR SON COUVERCLE

R&arm&On considere l’ecoulement et le transfert de chaleur dans une cavite car&e pour laquelle le
mouvement du fluide est induit par le frottement resultant du mouvement de translation du couvercle et
par la force de flottemeut due au chauffage du plancher. Cette etude est motivee par l’application a la
production du verre plan dans laquelle celui-ci est pousse sur un bain de metal liquide, puis refroidi et
solidifii. Les simulations numeriques sont faites pour un kcoulement laminaire bidimensionnel
(100 Q Re < 2200) et, pour differentes valeurs du nombre de Richardson, on Btudie les effets du nombre
de Prandtl depuis les valeurs faibles (0,Ol < Pr < 50) sur l’ecoulement et le transfert de chaleur dans la
cavite. La temperature et le champ d’kcoulement sont calculbs et prisentes pour illustrer la forte influence
du nombre de Prandtl. Les nombres de Nusselt locaux et globaux sont aussi rapportcs pour diffkntes
valeurs de Re, Ri et Pr.
1892 M. K. MOALLEMI and K. S. JANG

EINFLUSS DER PRANDTL-ZAHL AUF DEN WARMEUBERGANG BE1 LAMINARER

MISCHKONVEKTION IN EINEM HOHLRAUM MIT BEWEGTER DECKFLACHE

Zusammenfassung-Die vorliegende Arbeit beschaftigt sich mit der Stromung und dem Wirmeiibergang
in einem quadratischen Hohlraum, wobei die Stromung durch die Reibungskraft als Folge der Bewegung
der oberen Deck&he, in Verbindung mit der Auftriebskraft als Folge der Bodenheizung verursacht wird.
Die Arbeit wurde durch ihre Bedeutung fur die Produktion von Flachglas angeregt. Dort werden die
Glasscheiben iiber ein Bad mit fliissigem Metal1 gezogen, wahrend sie gekiihlt werden und erstarren. Aus
diesem Grund werden die numerischen Simulationsrechnungen fur zweidimensionale laminare Stromung
(100 < Re < 2200) im Bereich kleiner bis mittlerer Prandtl-Zahlen durchgefiihrt (0,Ol < Pr < 50). Der
Warmetibergang im Hohlraum wird fur unterschiedliche Werte der Richardson-Zahl untersucht. Urn den
starken EinfluB der Prandtl-Zahl zu verdeutlichen, werden die Temperatur- und Striimungsfelder im
Hohlraum dargestellt. Dariiberhinaus werden lokale und mittlere Nusselt-Zahlen fiir verschiedene Werte
von Re, Ri und Pr prlsentiert.

BJIMRHME 4kiCJIA I-IPAHATJDI HA TEITJIOI-IEPEHOC I-IPM JIAMMHAPHOR

CMEBIAHHOB KOHBEKHHW B I-IOJIOCTM C flBMmYILIE$iCX KPbIIIIKOR

AlrsOTaIQSI--~CCneAyIOoTcnTeqeHHe a TennonepeHoc B ksanparnoii no.noc~H, B KOTO~O~~ Tevewie

Bb13BaHO CABHrOBOfi CE!nOir,o6ycnoBJIeHHoi-i
ABHXCCHHCM BepXHeiiKPblUlKW,H IlOA~MHOZi CHnOii 3a CWT
HarpeBa OCHOBaHHR.&CneAOBaHHeBblnOAHIleTCIlC UenbEO HCIIOnb30BWiHRerO pe3ynbTaTOB llpH ITpOW
3onoACTBe nnocKor0 crercna, B npouecce ~o~oporo oxnawaeh4blR a 3aTsepAeBahwvG creKnnHHb&i
Juicr BbITRI'HBaeTCI83 BaHHbI C paCMaBJIeHHbIM MeTaJlJlOM.B CBI13HC 3THM lIpOBOAHTC5lYHCneHHOe
Monennpoeatine ana neyh%epHoro naMHHapHoro TeYeHHn (100 Q Re Q 2200) H HccneAyeTcn BnHRHHe
9ncna Hpasnrna OT ~anblx A0 yMepenHbtx 3HaqeHiifi
(T.e. 0,Ol c Pr < 50) na Tenernie ti rennonepenoc B
non0cTa npH pa3nsiqHblx3HaqeHHnx 9ticna PziqapAcoHa.CanbHoe nnHnwe wicna IlpaHnTnr munocr-
pepyercr wcneHHblMu pe3ynbTaTardH Ann nonefi rehnreparyp H renetnin B nonocrn. Hpencrasnenbr
TaKxe AOKiUlbHOeHC~Ali~'lHCnaHyCCenbTaIIpHpa3nFi~HbIx3HaYeHHffX Re,RiH Pr.