Combined forced and natural convection

heat transfer in a deep lid-driven

cavity flow
Ajay K. Prasad
Department of M e c h a n i c a l Engineering, University of Delaware, Newark, DE

Jeffrey R. Koseff
Environmental Fluid Mechanics Laboratory. Stanford University, Stanford, CA

In this study, we describe the combined forced and natural convection (also known as
mixed-convection) heat transfer process within a recirculating flow in an insulated lid-driven
cavity of rectangular cross section (150 mm x 450 mm) and depth varying between 150
mm and 600 mm. The forced convection is induced by a moving lid, which shears the
surface layer of the fluid in the cavity, thereby setting up a recirculating flow, while the
natural convection flow is induced by heating the lower boundary and cooling the upper
one. By appropriately varying the lid speed, the vertical temperature differential, and the
depth, we obtained Gr/Re 2 ratios for these flows from 0.1 to 1000. Flow visualization using
liquid crystals and heat flux measurements at specific locations over the lower boundary
provided an insight into the nature of the heat transfer process under different flow and
temperature conditions. The mean heat flux values over the entire lower boundary were
analyzed to produce Nusselt number and Stanton number correlations which should be
useful for design applications.

Keywords: Lid-driven cavity flow; forced and natural convection; thermochromic liquid
crystals

Introduction tal plates. For example, Wang et al. (1983) conducted heat
transfer measurements in a water channel flow that was uni-
Heat transfer in flows in which the influence of forced convec- formly heated from below and compared their results with exist-
tion and natural convection are of comparable magnitude (com- ing correlations. Similarly, Imura et al. (1978) focused on the
monly referred to as "mixed-convection" flows) occur frequently transition from the laminar to the turbulent regime, by measur-
in engineering situations. Gebhart (1971) separates mixed-con- ing the heat flux on a horizontal heated plate. At about the same
vection processes into external flows and internal flows and time, a numerical investigation of a mixed convection boundary-
provides a review of some of the more common geometries. layer flow on a horizontal surface was performed by Chen et al.
Mixed-convection flows may be further subdivided into those (1977).
where the inertia force is parallel to the buoyancy force and All the studies mentioned in the preceding paragraph pertain
those where the inertia force is perpendicular to the buoyancy to external flows. However, internal, or confined, flows are also
force. Lloyd and Sparrow (1970) developed a similarity solution of great interest to designers of such alternate energy systems as
for a mixed-convection flow on a vertical surface for a range of solar ponds and solar storage devices. For example, Mufioz and
Prandtl numbers; in this study, the forced-convection flow aided Zangrando (1986) describe an experimental investigation of mix-
the free-convection component. Siebers (1983) carried out exper- ing in a double-diffusive fluid layer. They studied the effect of
iments with a heated vertical wall in the presence of a cross surface shear as well as bottom heating on entrainment and
stream, with a view towards predicting the behavior of a central developed correlations for the same. Cha and Jaluria (1984)
receiving station in a solar power plant. Kays and Crawford performed a numerical investigation of a recirculating mixed-
(1993) list a number of studies of mixed convection over horizon- convection flow in a stratified energy storage system, involving
withdrawal of hot fluid and discharge of cold fluid into the
storage chamber. For the range of governing parameters studied,
their results indicate that the recirculating flow is strongly af-
Address reprint requests to Prof. A. K. Prasad, Department of fected by buoyancy, as well as the geometry. In addition, the
Mechanical Engineering, University of Delaware, Newark, DE cooling of electronic components on a printed circuit board, and
19716, USA. the heat transfer from rectangular cut-outs on the surface of a
Received 7 January 1996; accepted 16 April 1996 heat exchanger in the presence of an oncoming stream could also

Int. J. Heat and Fluid Flow 17: 460-467, 1996

© 1996 by Elsevier Science Inc. 0142-727X/96/815.00
655 Avenue of the Americas, New York, NY 10010 PII S0142-727X(96)00054-6
 Convection heat transfer in rid-driven cavity. A. K. Prasad and J, R. Koseff

be modeled as internal or recirculating flows, and provide a The ratio G r / R e 2 traditionally has been used to represent
special incentive for our study. the relative magnitudes of forced and natural convection in a
The configuration described in the present investigation con- mixed-convection flow. In our configuration, we have produced
sists of an internal flow within a rectangular cavity driven by two the same value of the ratio, but by using different combinations
mechanisms. First, by shearing the surface layer by means of a lid of AT, D, and Us. By examining the heat flux from these
moving at a constant velocity, we induce a forced-convection flow different cases, we want to determine if the ratio G r / R e 2 is
within the cavity (see Figure 1). The lid speed can be increased sufficient to characterize this class of mixed-convection flows, or
continuously to produce Reynolds numbers, (Re, based on the lid if it needs to be modified by such factors as the depthwise aspect
speed UB and the cavity width B) up to 12,000. This recirculating ratio (DAR).
motion is extremely complex and three-dimensional (3-D) in
nature and has been described in detail by Koseff and Street
(1984) and Prasad et al. (1988). Second, by heating the lower Facility and instrumentation
boundary of the cavity and cooling the upper boundary, we can
create a natural convection flow, as described by Rhee et al. The i/d-driven cavity facility
(1985). The temperature differential between the lower and
upper boundaries and the depth of the cavity D can be varied to The lid-driven cavity facility, constructed from 12.5-mm thick
obtain Grashof numbers for these flows (Gr, based on the cavity Plexiglas, consists of two attached "shoe-boxes," as shown in
depth) between 107 and 5 x 109. The simultaneous use of both Figure 2. The lower of the two boxes is the main area of interest.
forcing mechanisms enables us to create flows in which the It is rectangular in cross section, with streamwise width (B) of
parameter G r / R e 2 varies between 0.1 and 1000. By varying the 150 mm, spanwise length (L) of 450 mm, and depth (D) varying
dimensions of the cavity, it is also possible to create flows of between 150 and 600 mm. The upper box houses the lid and its
widely differing geometric attributes, yet possessing the same driving system, which consists of a variable speed motor con-
G r / R e 2 ratio. nected by a chain drive to one of a pair of rollers. The "lid" is a
Flow visualization using thermochromic liquid crystals (TLC) 0.08-mm thick copper belt, which is mounted on and driven by
provides a simultaneous view of the temperature and velocity the two rollers. During operation, the belt-and-drive assembly in
fields, thereby identifying the relevant flow structures and their Figure 2 is lowered down into the upper box so that the belt just
influence on the heat transfer mechanisms in the cavity. This was
supplemented by quantitative data, in the form of heat flux
measurements at specific locations over the lower boundary, COOLED WATER
using microfoil heat flux sensors. The heat flux data in this paper BELT DRIVE (
are averaged over the entire lower boundary of the cavity SUPPORT STRUCTURE
(hereafter referred to as the cavity floor).

UPPER PLATE
0 p r

"(~1 ~/ ~" ~ ~ DRIVE-SYSTEM

t ' "° J j / ~ : . ~ : [ / SUPPORT BOX

M -- ~ LOWER PLATE

T R N
I ; II ~ I R-N = L: SPAN OF CAVITY
I
• , 225 rata -I
L • 450 mm II
! II ill
; 11 III K-N
'°°°°
= D: DEPTH OF CAVITY
, ~ HI = 150 rnm to 600 mm
{4
BOUNDARY DEFINITIONS
K-T =B: WIDTH OF CAVITY
END-WALL, POSR sIDE-WALL UPSTREAM, PKNR = 150 mm
, KTMN 0OWNSTRIEAM OTMS
, HEATED WATER
LID, POTK LOWER HORIZ,, SMNR Figure 2 The lid-driven cavity facility (from Koseff and Street
Figure 1 Definitions for the lid-driven cavity f l o w 1984)

Notation Re Reynolds number based on lid velocity and cavity width

St Stanton number based on lid velocity
B cavity width UB lid velocity
cp specific heat of water
D cavity depth Greek
g acceleration due to gravity
Gr Grashof number based on cavity depth thermal diffusivity of water
h heat transfer coefficient 13 thermal expansion coefficient of water
k thermal conductivity of water AT temperature difference between lower (hotter) and up-
L cavity length or span per (colder) cavity boundaries
Nu Nusselt number based on cavity depth v kinematic viscosity of water
q heat flux ( W / m 2) p density of water

Int. J. Heat and Fluid Flow, Vol. 17, No. 5, October 1996 461
Convection heat transfer in lid-driven cavity." A. K. Prasad and J. R. Koseff

touches the upper edges of the working area. The belt speed can Visualization photographs (color)were obtained for AT = 8°C,
be varied continuously to produce Reynolds numbers up to at Re of 3200 and 7500, for all four DARs. In this paper,
12,000. At a particular setting, the belt speed is constant to however, we have included only two representative prints for
within +_0.25%. D A R of 2 : 1.
The lower boundary of the cavity is a heat exchanger plate. In
addition, the copper belt is in continuous contact with a similar Heat flux measurements technique
plate. By circulating water of the appropriate temperature
through each plate, it is possible to generate a natural convection The heat flux meters used for the present study were microfoil
flow within the cavity. The maximum vertical temperature differ- heat flow sensors manufactured by the RdF Corporation. These
ential that can be attained is 10 _+ 0.05 °C. The four vertical walls heat flux meters are in the form of a thin, rectangular strip or
of the cavity are adiabatic. This is ensured in practice by adding foil and are mounted flush on the surface at which the heat flux
styrofoam insulation to the side-walls and by maintaining the is to be measured. A good thermal contact between the sensor
mean fluid temperature as close to the laboratory ambient tem- and the surface can be achieved with double-sided adhesive tape.
perature as possible. The heat flux meters used in this study measured 12 × 30 mm,
Finally, the cavity depth (D) can be varied from 150 to 600 with a foil thickness of 0.11 mm.
mm by lowering the bottom heat exchanger plate. In this study, The foil consists of an electrical insulator, in which a number
heat flux measurements were performed at four cavity depths: (typically 20) of copper-constantan thermocouple junctions are
150, 300, 450, and 600 mm. These correspond to D A R (DAR = embedded. The junctions are connected in a single series circuit
D/B) of ] : 1 , 2:1, 3:1, and 4:1, respectively. For each case, and are arranged so that alternate junctions are on opposite
three temperature differentials were studied: 1, 4, and 8 °C.
surfaces of the foil. Therefore, the voltage generated at the leads
Furthermore, for each combination of D A R and temperature
of the heat flux meter is directly proportional to the difference in
differential, nine belt-speeds were used providing Re numbers
the temperatures of the two surfaces, which in turn, depends on
ranging from 0 to 12,000. This resulted in a total of 108 different
cases. the heat flux through the foil. The manufacturer-supplied ther-
mal resistance of the sensors used in our experiments was
Flow visualization technique 5.3 x 10 4 K / ( W / m 2) with a response time of 50 Hz.
In our measurement configuration, the heat flux meters were
Thermochromic liquid crystals possess the ability to change color submerged in water for prolonged periods of time. Consequently,
with temperature, and therefore, can be used to visualize the the manufacturer-supplied calibration constants could not be
temperature field in mixed-convection flows. The liquid crystals relied upon for high accuracy. Therefore, we devised an in-situ
are encapsulated in microspheres which are advected by the flow, calibration technique, and recalibration was performed at fre-
and thereby enable visualization of the flow field as well. For the
quent intervals. On the basis of slight drift in the calibration
present study, we chose a chiral nematic type of TLC manufac-
constants, we estimate that the uncertainty in the heat flux
tured by Hallcrest. This particular sample is available in the
measurements is limited to _+5%.
microencapsulated form with 50-100 p~m sized capsules. The
A total of eight heat flux meters were attached to the lower
active range for this TLC starts at 23 °C and has a bandwidth of
boundary of the cavity, as shown in Figure 4. This particular
2 °C. A concentration of 0.2 g/1 is adequate to visualize the flow
satisfactorily. arrangement was chosen to distribute the measurement locations
Specific planes of the cavity were illuminated, as shown in uniformly over the lower boundary of the cavity, and thus to
Figure 3. The "end view" refers to the view of the symmetry obtain an accurate estimate of the average, overall heat transfer
plane through the end wall, while the "side view" offers a view of from the lower boundary. The voltage output of the meters
a plane parallel to the downstream side wall (DSW) through the (which typically, is about 6 Ixvolts for every 100 W / m 2 of heat
upstream side wall (USW). The illumination for both views was flux) was then amplified by a factor of 40,000 using an 8-channel
provided by an 800-W ILC Technology xenon lamp in conjunc- amplifier, before digitization. An HP2100 data acquisition system
tion with a slit-and-lens arrangement, which focused the light was used to sample the heat flux output at a rate of 100 Hz, for a
into a thin (3-mm) sheet. Kodacolor 400 ASA film and an period of 11.92 minutes. The heat flux for each meter was then
exposure time of 3 seconds with an f stop of 5.6, produced the analyzed on an HP1000 system, providing time traces and spec-
best results. tra, and the mean heat flux over the entire lower surface was
determined by taking the average of all eight heat flux meters.
D O ~ s t ream Side-Wall
(However, as mentioned previously, this paper focuses specifi-
cally on the averaged heat flux values.)

Downstream Side Wall

D E4

ISide-View
Upstrea~ Side Wall
End-View visualizes SyE~etry Plane [ ]
450 m~ m
Side-View VisualiEes Plane p a r a l l e ~
to Downstream Side Wall
Figure 3 Illuminated planes for flow visualization Figure 4 Location of heat flux meters over the cavity floor

462 Int. J. Heat and Fluid Flow, Vol. 17, No. 5, October 1996
 Convection heat transfer in lid-driven cavity. A. K. Prasad and J. R. Koseff

Observations Figure 5b is the corresponding side-view picture that was

exposed 4 seconds after Figure 5a, in a plane parallel to the
F l o w visualization results DSW, and 22 mm from it. It is possible to pick out counter-rotat-
Visualization of two perpendicular planes of the cavity provided ing pairs of longitudinal vortices along the floor, separated at
valuable insight into the nature of the 3-D flow structures in the regular intervals in the spanwise direction. These are the
lid-driven cavity flow. At the outset, the heat exchangers were set Taylor-G6rtler-like (TGL) vortices that originate over the con-
to produce the required temperature differential and allowed to cave separation surface of the downstream secondary eddy (DSE)
run for a period of 24 hours. The two light sources and cameras (see Figure 1). The mechanism that generates these vortices
were then positioned as shown in Figure 3. Next, the belt was involves the centrifugal instability resulting from the flow over a
started and set at the desired speed, and the flow eventually concave surface and have been extensively described in Koseff
reached a fully developed state. Subsequently, a measured quan- and Street (1984) and Prasad et al. (1988) for isothermal flows at
tity of TLCs was introduced slowly through a small port in the a D A R of 1 : 1. Rhee et al. (1985) discovered that TGL vortices
DSW and distributed uniformly by the mixed-convection flow. also form at a D A R of 1 : 1 in a mixed-convection flow (AT = 4
Figures 5a and 5b form an end-view/side-view pair for a °C, Re = 3200); however, the size of the TGL vortices are dimin-
D A R of 2:1, AT of 8 °C, and Re of 3200. Figure 5a is an ished when compared with that of the isothermal case.
end-view picture of the symmetry plane. The belt moves from left
to right at the top of the picture. The fluid adjacent to the belt is Heat flux results
cold, as it is cooled by the upper heat exchanger plate. This cold
fluid (which is red in color) is stripped off at the corner formed Heat flux measurements were performed in the following man-
by the lid and the DSW and plunges down along the DSW in the ner. The calibrated heat flux meters were attached on the cavity
form of a wall jet. The inertia imparted to this fluid by the lid is floor at the locations indicated in Figure 4, and the lower plate
aided by its larger density (the red fluid is colder, and therefore, was moved to the appropriate depth. As in the case of flow
denser), and the wall jet continues all the way to the floor of the visualization, the heat exchangers were run for a period of 24
cavity. At the floor, the fluid absorbs heat provided by the lower hours at the required temperatures before data acquisition com-
heat exchanger plate, and the fluid becomes warmer and changes menced. The first case studied during an experiment was the
in color to a dark blue (the hot end of the color spectrum). natural convection case (lid is stationary Re = 0). Subsequently,
Finally, the warmed-up fluid ascends the USW in the form of a the belt was started and the belt speed was set to produce the
buoyant plume and penetrates into the core of the recirculating next Re. Data were acquired only after allowing sufficient time
fluid. for the new flow conditions to reach a fully developed state

Direction of Lid

Ill
f~
$
I=
tl} 3
E O.
?
O.

Lower Boundary

(b) Side-View

Lower Boundary
(a) End-View
Figure 5 Flow visualization with liquid crystals: DAR 2:1, R e = 3 2 0 0 , A T = 8 °C; (a) symmetry plane (end view); (b) plane
parallel to DSW and 20-ram from it (side view)

Int. J. Heat and Fluid Flow, Vol. 17, No. 5, October 1996 463
Convection heat transfer in lid-driven cavity." A. K. Prasad and J. R. Koseff

(about 1 hour). This process was repeated every hour, until the ing range of G r / R e 2 spanned about four orders of magnitude,
highest Reynolds number had been studied. from 0.1 to 1000.
Figures 6a, b, c, and d show the mean heat flux over the cavity The ratio G r / R e 2 is used traditionally to indicate the relative
floor (averaged over all eight heat flux meters for a period of 12 strengths of the two modes of convection in a mixed-convection
minutes) for DARs 1:1, 2:1, 3:1, and 4:1, respectively. Each environment. Gebhart (1971) deduced from a simple manipula-
plot contains heat flux data for the three temperature differen- tion of the Navier-Stokes equations that the buoyancy effects
tials, as a function of the Reynolds number. The corresponding become noticeable when G r / R e 2 approaches unity. For G r / R e 2
uncertainties ( + 5 % ) are also indicated. These plots show that << 1, the forced-convection component controls the heat transfer
the heat flux is strongly influenced by the temperature differen- processes, while for G r / R e 2 >> 1, buoyancy effects predominate.
tial, and at a given AT, it usually increases with the Reynolds Based on this principle, Siebers (1983) defined the mixed convec-
number. This behavior is seen for all DARs. We see that the
tion limits for heat transfer from a vertical heated plate in a
D A R seems to have a rather weak influence on the heat flux.
horizontal stream as 0.7 _< G r / R e 2 _< 10.
However, for the DARs of 3 : 1 and 4 : 1, we see that at higher
Unfortunately, the ratio G r / R e 2 cannot be applied univer-
Reynolds numbers ( > 8000), the heat flux begins to level off: an
increase in Re does not cause an increase in the heat flux. In sally to every flow configuration. Researchers have often modi-
fact, for the AT = 1 °C case at these DARs, the heat flux actually fied this parameter in order to interpret their data. For instance,
decreases with Re, for Re > 5000. Wang et al. (1983) delineated the forced-, mixed-, and free-con-
vection regions with the parameters G r / R e 3/2. They determined
the mixed-convection regime to apply for 10 _< G r / R e 3/2 < 400.
Similarly, Imura et al. (1978) found that transition from laminar
Discussion
forced convection to turbulent convection occurred in the range
The experiments described above were made in a flow driven of 100_< Gr/Re3/2_< 300. Chen et al. (1977) report that for
jointly by shear and buoyancy. The strength of shear-driven G r / R e 5/2 > 0.05, buoyancy effects became significant.
motion is given by the Reynolds number Re, and the strength of One of our objectives was to determine if the ratio G r / R e 2
buoyancy is given by the Grashof numbers Gr. In these experi- could be used to classify the cavity flow driven jointly by shear
ments, 0 _< Re _< 12,000 and 107 < Gr _< 5 x 109. The correspond- and buoyancy adequately.

2000 'C', , , I , , r , I , , , , I , , , , I , , , I I , , , , 2000 . . . . I' ' ' ' 1 ' ' ' ' i ' ' ' ' 1 ' ' ' ' I' ' ' '

O AT= 1°C 0 AT= 1°C

o AT = 4oc o AT= 4°C
•,, A T = 8 o c ~, A T = 8oc
1600 A ~' ~" 1500

E E

x 1000, 1ooo
,'7

1:3 O o
T
O
O
o 0
500, 500

0 0 0 0 0 0 0 0
i0 ' 0 0 0 0 0 0 0
0 . . . . I . . . . I . . . . I,,a,l,~,l,,,,
0
0 2000 4000 6000 8000 10000 12000 0 2000 4000 6000 8000 10000 12000

Reynolds Number Reynolds Number

(a) D A R 1:1 (b) DAR 2:1

2000 2000
' ' ' ' 1 ' ' ' ' 1 ' ' ' ' I ' ' ' '1' ' ' ' 1 ' ' '' , r , , I . . . . i , , , , i , , , , i , , r , t , , , i

O &T= 1°C O AT= 1°C

[] & T = 4°C z~ I o A T = 4oc
A &T = 8oc ~, ~ &T=8°C ~. ,',
1500 1500

E A
LX
E z~

lOOO ~ 1000
g
u.

o o 0
T 0 o 0 o
0 "I- O
13 O

500~ 500

0 0 0 0 0 0 0 0 0 0 0 0 O 0 0 0
0 . . . . I , , , , I , , , , I , , , , I , , , ,I . . . . . . . . I , , , , I , , , , I , , , , I , ,, , I , , ~ ,
0 2000 4000 6000 ~00 10000 12000 2000 4000 6000 6000 10000 12000

Reynolds Number Reynolds Number

(c) DAR3:I (d) DAR 4:1

Figure 6 Variation of heat flux with Reynolds number and the temperature differential

464 Int. J. Heat and Fluid Flow, Vol. 1 7, No. 5, O c t o b e r 1 9 9 6

 Convection heat transfer in lid-driven cavity. A, K. Prasad and J. R. Koseff

Free-convection m e a s u r e m e n t s at R e = O i i ' ' ' ' ' ' I

DAR
First, we focus on our free-convection heat transfer measure- 10 3 - 1:1 2:1 3:1 4:1
ments made at Re = 0. This pertains to the case of : AT=I°C 0 c~ A o
Rayleigh-BEnard convection at high Rayleigh number, a prob- ?. t, T = 4°C
&T=8°C
• • • •
~ E~ ~ ®
lem that has been actively studied over the past few decades.
Previous experimental work has shown that the heat flux through
the fluid layer may be expressed as u~

t.J
Nu = clRa 1/3 (1)
¢u

where Nu is the Nusselt number: h = 23.6Re °22

-'r"

q D i i i i i 1 i I ]
Nu = - - - - (2) ~o3 10 4
AT k
Reynolds Number, Re
and Ra is the Rayleigh number: Figure 7 Variation of the heat transfer coefficient with
R e y n o l d s n u m b e r a n d t e m p e r a t u r e d i f f e r e n t i a l f o r all D A R s
gf3A TD 3
Ra
p~
Influence o f G r / R e 2 on Nu

The 1/3 dependence of Nu on Ra arises from the classical The heat flux data can also be presented in the form of the
argument that when the convecting layer is highly turbulent, the Nusselt number (Equation 2). The resulting Nusselt numbers for
heat flux q becomes independent of the layer depth D. The all four DARs were plotted on a single graph, against the ratio
implication is that, there exists a thin conduction boundary layer G r / R e 2 (Figure 8). The purpose of plotting Nu against G r / R e 2
which controls the heat flux adjacent to the lower and upper was to identify forced-, free-, or mixed-convection regimes in the
boundaries; whereas, the rest of the fluid layer is well mixed by G r / R e 2 spectrum spanning four orders of magnitude from 0.1 to
turbulence. The thickness of the boundary layer does not change 1000. From Figure 8, the Nusselt number distribution indicates a
appreciably, even when the total depth D is varied. Conse- large spread. However, for a given D A R (especially D A R = l : 1)
quently, q does not vary with D. Our results for Re = 0 can be the data appear to collapse quite well. Therefore, we decided to
fitted by Equation 1 with c~ = 0.05, which agrees reasonably with incorporate the D A R as an extra variable and calculated the
Goldstein and Tokuda (1980) who give c~ = 0.0556 for convection coefficients for the following correlation, which is valid for the
in water at 109 < Ra < 2 × 10 H. data from all four DARs.

The heat transfer coefficient l Gr l - ° . ° 2 r D ]11

From Figure 6, the independence of q on D continues, even (4)
when Re > 0. This implies that the addition of shear to the
buoyancy-driven flow does not alter the heat transfer mechanism
adjacent to the boundaries. However, as Re is increased, there is The corresponding least-squares fit is shown in Figure 9. The
a definite increase in the heat flux for a given AT. surprising aspect of this correlation is that the exponent in
The dependence of the heat flux on the temperature differen- G r / R e 2 is close to zero (in fact, owing to experimental errors in
tial can be removed by dividing the heat flux by AT. This yields the measurement of heat flux and temperature, the exponent
the heat transfer coefficient h: may be safely rounded off to zero; however, we have deliberately
retained the G r / R e 2 term in the correlation in Figure 9 to
q highlight this unexpected result). The result is unexpected given
h ~ - -
AT

10 3 t i t~rlll[ i iiiiii1[ i ill,lit I i iil~lf~] i r iill~l[ i IrJ~lr.

Figure 7 shows the variation of the heat transfer coefficient as a
function of Reynolds number. This plot contains the data for the DAR
three ATs and the four DARs. The good collapse of the data 1:1 2:1 3:1 4:1
AT= 1°C 0 o ~ 0
points suggests that the heat transfer processes occurring along AT=4°C
& T = 8°C
the cavity floor are fairly uniform for all the cases studied. A
least-squares fit of the data provided the following correlation:
:~ 102
h = 23.6 Re °'22 (3)

which is also indicated in Figure 7. In this correlation, the units

of h are W / m 2 °C. The uncertainty in the calculated value of the
heat coefficient is somewhat larger than that in the correspond-
ing heat flux. This additional uncertainty arises because of the
J fliH.I i i ,...,.I i JiHinl J i,,,,,d i j=,,..I j ill.
error in the measurement in the AT (+0.05 °C), and we expect 10
10-2 1 0 "1 1 10 102 10 3 10 4
the relative error to be larger for a smaller AT. Therefore, the
uncertainty in the heat transfer coefficient is approximately Gr/Re 2
+ 10% for the 1 °C case, + 7% for the 4 °C, and ___6% for the 8 Figure 8 V a r i a t i o n of t h e N u s s e l t n u m b e r w i t h G r / R e 2 f o r
°C case. These uncertainties are indicated in Figure 7. all D A R s

Int. J. Heat and Fluid Flow, Vol. 17, No. 5, October 1996 465
Convection heat transfer in ~M-driven cavity." A. K. Prasad and J. R. Koseff

1021 ' ' '''"'l . , ,ill. i v I ,i,..| ' ' ~'""1 ' ' *'""1 ' ' ''"'1
From Equation 4 we see that Nu increases with Re, indicating
DAR
that shear does influence heat transfer. However, we also see
1:1 2:1 3:1 4:1 that Nu varies almost linearly with D (i.e., it varies as Gr 1/3)
AT=I°C 0 o ~ o indicating the influence of natural convection. It is appropriate
A T - 4°C • • • •
& T = 8°0 • a • e to say that forced and natural convection effects actually comple-
ment each other: The shear imparted by the lid drags cold fluid
< towards the DSW. The shear-driven, descending wall jet is aided
13
oo
10 in its downward motion along the DSW because of its colder

Z
0)
¢r

/ Nu_-790.e0,.[ ] (denser) nature. At the lower boundary, the boundary-layer flow

is heated up and is driven upwards not only by the remnants of
shear but also by the additional buoyancy generated along the
lower boundary. Hence, in this configuration and for the range
studied, the flow shows the combined effects of forced and free
convection. Both effects operate seamlessly, as confirmed by the
1 . ,~ ..... I , .,,..,I , ,,,,,,,I , ~ ...... I ....... ,I , JiJ,,.. complete lack of sensitivity to G r / R e 2 over the entire range.
10 2 10"1 1 10 102 10 3 10 4

Gr/Re 2 The Stanton n u m b e r

Figure 9 V a r i a t i o n of t h e N u s s e l t n u m b e r as a f u n c t i o n of From the discussion in the preceding paragraph, it is apparent
R e y n o l d s n u m b e r , D A R , and G r / R e 2 that the heat flux q is independent of cavity depth D, weakly
dependent on Re, and independent of G r / R e 2. In fact, Equa-
tions 3 and 4 show very clear similarities. If we neglect the term
involving G r / R e 2 (on account of its very small exponent) and
the large range of variation of G r / R e 2 from 0.1 to 1000, and the approximate the exponent of D/B as unity, then Equation 4 can
traditional belief that G r / R e 2 ~ 1 separates the natural convec- be rewritten in a form very similar to Equation 3 (h = CRe°18),
tion regime from the forced-convection regime. where C is a constant (C = 8k W/m2K). This exercise reveals
Second, the coefficient for D A R is close to unity, which that the heat transfer coefficient h adequately represents the
indicates that the Nusselt number increases almost linearly with heat transfer within the lid-driven cavity.
D A R (i.e., q is independent of D). In other words, we see that However, Equation 3 relates a dimensional quantity h to a
the addition of shear to the buoyancy-driven flow does not alter nondimensional number Re. This drawback can be circumvented
the fact that the heat transfer is governed by effects that are by employing the Stanton number:
restricted to a small thickness adjacent to each boundary. The
rest of the fluid layer is highly turbulent, disorganized, and
well-mixed (Figure 5) very similar to the free-convection situa-
h
St=--
tion. Figure 5(a) shows that for Re > 0, the flow along the walls pCpUB
of the cavity is well organized and dominated by inertia. For
instance, the region adjacent to the DSW is always occupied by a
cold, plunging wall jet, while the fluid adjacent to the USW is a Figure 10 displays the Stanton number variation with Re for all
warm, buoyant plume. In addition, the lid and the floor have the DARs and ATs studied in this investigation. The least
boundary-layer flows attached to them. In contrast, the internal, squares fit shown in Figure 10 is given by:
core region of the recirculating fluid is dominated by the more
chaotic, natural convection flow structures.
The pattern of a well-organized boundary-layer flow with a St = 0.94 Re -°'78
chaotic internal flow, seen in Figure 5(a) for D A R = 2: 1, is
exactly duplicated at D A R s of 1:1, 3:1, and 4:1. Therefore, The uncertainty in St is essentially equal to the uncertainty in the
because the flow dynamics over the cavity floor are similar for all corresponding h on account of the low error in UB; this uncer-
DARs, we expect the magnitudes of the heat transfer coefficient tainty is also indicated in Figure 10.
to exhibit a similar behavior over a wide range of DARs and
ATs.
Third, the coefficient of Re in Equation 4 of 0.18, reflecting a
i I r ~ I r v
rise in heat transfer rates at the upper and lower boundaries as
DAR
the lid speed is increased. This result corroborates the correla-
10"2 1:1 2:1 3:1 4:1
tion presented in Equation 3. Re °~8 is a fairly weak dependence &T-I°C O. o ~. o
when compared to Re ~/2 seen, for example, in the Pohlhausen &T=4OC • • • •
&T=8°C ¢' a • e
relationship for laminar flow over an isothermal fiat plate. How-
ever, despite the apparent similarities between the flow over the
lower boundary of the lid-driven cavity and Blasius-type flow E
over a fiat plate, there exist significant differences. For instance,
#_
g
as seen in Figure 5(a), the flow over the lower boundary may be 10 .3
divided into three zones: (1) impingement of the wall jet as it
descends along the DSW with the lower boundary; (2) attached
flow over the lower boundary; and (3) separation or updraft
region near the USW. Moreover, the effect of buoyancy (cold
plumes dropping off the upper boundary layer and hot plumes I I I I I I I I

rising from the lower boundary layer) destabilizes the boundary- 10 Re 10 4

layer flow. Consequently, it is improbable that the Re 1/2 behav- Figure 10 Variation for the Stanton number with Reynolds
ior would be replicated by the lid-driven cavity flow. n u m b e r a n d t e m p e r a t u r e d i f f e r e n t i a l f o r all D A R s

466 Int. J. Heat and Fluid Flow, Vol. 17, No, 5, October 1996
 Convection heat transfer in lid-driven cavity. A. K. Prasad and J. R. Koseff

Conclusions References
Flow visualization and heat flux measurements have been con- Cha, C. K. and Jalurai, Y. 1984. Recirculating mixed convection flow
ducted in a bottom-heated, lid-driven cavity flow. The range of for energy extraction. Int. J. Heat Mass Transfer, 27, 1801-1812
Reynolds numbers studied were from 0 to 12,000. The vertical Chen, T. S., Sparrow, E. M. and Mocoglu, A. 1977. Mixed convection
in a boundary layer flow on a horizontal plate. J. Heat Transfer,
temperature differential, as well as the cavity depth, were varied
99, 66-71
to produce Grashof numbers from l0 T to 5 x 10 9. The corre- Gebhart, B. 1971. Heat Transfer, 2nd ed. McGraw-Hill, New York,
sponding G r / R e 2 varied between 0.1 to 1000. 388-397
Our results indicate the following. Goldstein, R. J. and Tokuda, S. 1980. Heat transfer by thermal
(l) The heat transfer coefficient (obtained by dividing the aver- convection at high Rayleigh numbers. Int. J. Heat Mass Transfer,
aged heat flux over the entire floor by the temperature 23,738-740
differential) is insensitive to G r / R e 2 for the range men- Imura, H., Gilpin, R. R. and Cheng, K. C. 1978. An experimental
tioned above. Flow visualization studies (for DARs 1 : 1, 2 : 1, investigation of heat transfer and buoyancy-induced transition
3:1, and 4: 1) indicate that this is because the nature of the from laminar forced convection to turbulent free convection over
a horizontal isothermally heated plate. J. Heat Transfer, 100,
flow adjacent to the cavity floor remains unchanged with
429-434
D A R and AT. Kays, W. M. and Crawford, M. E. 1993. Convective Heat and Mass
(2) The heat transfer within the cavity is independent of G r / R e 2 Transfer, 3rd ed. McGraw-Hill, New York, 413-414
over the range studied. However, the fact that q is indepen- Koseff, J. R. and Street, R. L. 1984. The lid-driven cavity flow: A
dent of D merely indicates that Nu ~ G r 1/3. This relation- synthesis of qualitative and quantitative observations. J. Fluids
ship is expected for pure free convection; our results indicate Eng., 106, 390-398
that the relationship holds even when external shear is im- Lloyd, J. R. and Sparrow, E. M. 1970. Combined forced and free
parted to the flow. The inference is that both free- and convection flow on vertical surfaces. Int. J. Heat Mass Transfer,
forced-convection effects are important and operate seam- 13,434-438
Mufioz, D. and Zangrando, F. 1986. Mixing in a double-diffusive,
lessly over the G r / R e 2 range studied.
partially stratified fluid. SERI/TR-252-2942, Solar Energy Re-
(3) The low exponent on the Reynolds n u m b e r Re °Is when search Institute, Golden, Colorado, USA
compared with the Blasius flow exponent of Re 1/2, confirms Prasad, A. K., Perng, C-Y and Koseff, J. R. 1988. Some observations
the flow visualization results that the flows along the upper on the influence of longitudinal vortices in a lid-driven cavity
and lower boundaries of the cavity are significantly different flows. In A Collection of Technical Papers, Part 1, A I A /
from Blasius-type flow over a flat plate. Also, buoyancy A S M E / S I A M / A P S ]st National Fluid Dynamics Congress,
destabilizes the boundary layer, adding to the deviation from Cincinnati, OH, 288-295
the Blasius Re ~/2 behavior. Rhee, H. S., Koseff, J. R. and Street, R. L. 1985. Visualization of
natural and mixed convection flows in a cavity, Proc. Int. Sympo-
sium on Refined Flow Modeling on Turbulence Measurements, Uni-
versity of Iowa, Iowa City, IA, 111, 1-9
Acknowledgments Siebers, D. L. 1983. Experimental mixed convection heat transfer
from a large vertical surface in a horizontal flow. Ph.D. thesis,
This work was supported by the Department of Energy under Stanford University, Stanford, CA, USA
G r a n t DE-FG03-84ER13240. The authors appreciate the valu- Wang, G. S., Incropera, F. P. and Viskanta, R. 1983. Mixed-convec-
able contributions of R. L. Street, and R. J. Moffat in interpret- tion heat transfer in a horizontal open-channel flow with uniform
ing the experimental results. bottom heat flux. J. Heat Transfer, 105, 817-822

Int. J. Heat and Fluid Flow, Vol. 17, No. 5, October 1996 467