Int..I. Heat Mass Tran,~lbr. Vol, 40, No. 4. pp.

915 9129, 1997

'~ Pergamon Copyright !~ 1996 Elsevier Science Ltd

Printed in Great Britain. All rights reserved
0017 9310/97 $1 ZOO + 0.t)0

P I I : S0017-9310(96)00120-2

Effects of wall conduction, internal heat sources

and an internal baffle on natural convection heat
transfer in a rectangular enclosure
Y. S. S U N
Advanced Package Development. Advanced Semiconductor Engineering Inc., Kaohsiung, Taiwan,
Republic of China

and

A. F. E M E R Y

Department of Mechanical Engineering, University of Washington, Seattle, WA 98105, U.S.A.

(Received 25 October 1995 and in.final.[brrn 29 March 1996)

Abstrae~Conjugate natural convection heat transfer in a two-dimensional, air-filled enclosure containing

discrete internal heat sources and an internal baffle is examined. The enclosure formed of finite conductive
walls is designed to simulate the behavior of an experimental window calorimeter in order to correct for
losses from the calorimeter. The equations are solved using a finite-volume method for a wide range of
Rayleigh numbers, internalmxternal heat source ratios, solid fluid conductivity ratios and baffle heights.
The influences on the heat transfer and the flow characteristics resulting from the internal heat sources as
well as the conductive baffle are discussed. Measurements of temperature distributions in the window
calorimeter are also reported. The comparison between numerical predictions and experimental measure-
ments shows that it is inappropriate to specify simple boundary conditions on the window surface of the
calorimeter and to neglect the conduction through the baffle, window, and walls. A modified procedure
for calculating the temperature distributions on the window surface improves the predicted results and
yields good agreement with the measured data. Copyright ~ 1996 Elsevier Science Ltd.

INTRODUCTION frames with a thermal break. When the results were

analyzed [1], the computed U values were generally as
The present investigation is a detailed study of the expected, but at times some of the better insulated
coupled conduction and natural convection process windows assemblies performed poorer than the appar-
in a window calorimeter. The calorimeter is equipped ently less insulated assemblies. Our initial thought was
with a heater, a temperature sensor and a baffle. The that the baffle caused the variation, but an analysis by
baffle prevents nocturnal cooling of the sensor and Sun and Emery [2] indicated that it could not account
straightens the flow over the glazing. The window for all of the discrepancy. Some preliminary analysis
calorimeter is represented as a two-dimensional, air- suggested that the variation was due to heat lost
filled rectangular cavity with finite conductive walls, through the insulated walls, and to a non-uniform
several discrete internal heat sources and an internal thermal condition on the outer surface of the window.
baffle (Fig. 1). In operation, a set of resistance wires, This study reveals that it is primarily the non-uniform
the discrete sources, dissipate sufficient heat to keep window boundary condition which causes the unex-
the temperature measured by the sensors equal to the pected variation.
temperature in the room, Tw Ideally, if the air in the A benchmark study of a square cavity with differ-
calorimeter were isothermal and at T., no heat would entially heated vertical walls and insulated horizontal
be conducted through the calorimeter walls and all walls was given by de Vahl Davis [3]. Chenoweth
of the heat dissipated by the heater would transfer and Paolucci [4] conducted a seminal study of the
through the window, thus permitting the calculation Boussinesq assumption. Henks and Hoogendoorn [5]
of the window U-value, U = Q / A / ( T H - T). In reality, examined the stability for air in this benchmark cavity.
the calorimeter air is not isothermal and corrections Recently, increased interest in the study of natural
have to be made for the heat lost to the room and for convection in cavities lies in the thermal control of
the effect of the baffle. The calorimeter analyzed in electronic equipment. A number of studies on con-
this study was used to compute the U value of special vection cooling applicable to thermal control of elec-
different window assemblies: single, double, triple tronic equipment were reviewed by Incropera [6]. Sev-
glazing; wooden sash, aluminum frames, aluminum eral investigations of convection from discrete heat

915
916 Y. S, SUN and A. F. EMERY

NOMENCLATURE
Cp constant pressure specific heat [~' thermal expansion coefficient
g gravitational acceleration [ms -z] {K q
h dimensionless height Y diffusion coefficient
H dimensionless height of cavity K conductivity coefficient
k thermal conductivity [W mC -~] r/ heat transfer fraction ( = heat transfer
k* solid-fluid thermal conductivity ratio through hot wall/total heat
( = k,kr) transfer)
Nu Nusselt number v kinematic viscosity [m: s ~]
P dimensionless pressure 0 dimensionless temperature
Pr Prandtl number ( = v/~) r dimensionless time
RuE external Rayleigh number p density [kg m-~l
( :gfl A TL~o/a,') c~ void fraction (=hw/HW).
RaL internal Rayleigh number
(:gflOL~/~vk)
Q uniform volumetric heat generation Subscripts
[W m-] b baffle
q dimensionless heat transfer rate C cold side
U, V dimensionless velocities f fluid
X, Y space coordinates H hot side
W dimensionless width of cavity. h heater
s solid
Greek symbols S sensor
thermal diffusivity [m: s- ~] oo external.

adiabatic--~

- W
wal

T
! i
glazing

H
1 < --T
hb C
h
sensor H

Yh ,.,,

F
Xh eN e I
Xb

adiabatic--
X
Fig. 1. Model of the experimental system Window Calorimeter.
 Effects on natural convection heat transfer 917

sources in air have been reported. Acharya and ~V ~V ~V

Goldstein [7] computationally examined the heat
transfer characteristics in an externally heated vertical
or inclined square cavity with uniformly distributed
internal energy sources. Conjugate heat transfer for
C(r V) 0(r V) R. (3)
developing flow over finite heat sources on a parallel
plate was reported by Davalath and Bayazitoglu [8]. (?0 d0 60
Affid and Zebib [9] considered heat conduction CT
+ g~x + VUy
coupled with convection in air with heat sources pro-
~ [ K ~0~ c~ / K DO~ Ra 1
truding from adiabatic vertical surfaces. Discrete heat
gX~rr~)+ ~-~r~)+ ~avDr O. (4)
sources flush mounted on a substrate in a liquid-filled
square enclosure were investigated by Joshi et al. [10].
However. most of the studies have neglected the inter- The Ra~/RaE term in equation (4) is equal to
action between convection in a fluid-filled cavity and QL2/kAT, a dimensionless quantity, that is an inde-
conduction in the walls surrounding the cavity. Kim pendent parameter representing the strength of the
and Viskanta [11] reported the effect of wall heat internal heating. The same set of equations can be
conduction on natural convection heat transfer in a applied to the solid portions by using a harmonic-
square cavity without internal heat sources. mean formulation in the diffusivities (V and ~) for the
solid fluid interfaces. F is a general diffusion
coefficient which is equal to 1 in the fluid region and
10t5 in the solid region. By using these values, the
velocities automatically approach zero in the solid
ANALYSIS regions. K is a general conductivity coefficient which
Phvsical model and assumptions is equal to 1 in the fluid region, and equal to k* in
The physical model and coordinate system of the the solid region, where k* is the ratio of solid-fluid
problem are illustrated schematically in Fig. 1. The conductivity. This conjugate approach is very con-
two-dimensional, rectangular cavity is formed by venient, but it is inefficient because it requires solution
finite conductive walls. A conductive baffle of the same of the momentum equations in the area defined by
material as the wall and a discrete heat source are solid materials.
mounted inside the cavity. The horizontal walls tbr-
ming the enclosure are assumed to be insulated exter- Numerical procedure
nally. Different temperatures are imposed on the out- Equations 0)-(4) are solved by a control-volume-
side of the two vertical walls and the properties of the based finite difference procedure. The equations are
walls and the fluid are taken at the mean temperature. expressed in an integral form over each control vol-
The internal flow is assumed to be steady, laminar, ume and discretizations are made to represent the
Newtonian and incompressible, the Boussinesq interface values and the derivatives in terms of grid
approximation is also valid. A problem with an point values. Hybrid [12] and QU1CKE [13] schemes
internal heat source is that the volumetric heat source are employed to obtain the solutions. The results from
induces heat transfer and flow characteristics which these two schemes agree reasonably well for small
differ from those of an externally heated cavity, there- Rayleigh numbers, but differ at higher Rayleigh num-
fore, one extra nondimensional parameter, Ra~, has to ber flows [14]. Therefore, the second order QUICKE
be defined. This new parameter is called the 'internal scheme was employed to discretize the convective
Rayleigh number'. The Rayleigh number based on the terms. The velocity-pressure interlinkage was solved
external temperature difference is referred to as the using the SIMPLE algorithm [12] and the discretized
external Rayleigh number, RaE, equations were solved iteratively using a line-by-line
TDMA (tri-diagonal matrix algorithm) procedure,
gflQ L So gfi AOL 3 The false-time-step method, in which the solution of
Rat - and RaE= the steady-state results are found by solving the
~vk av
unsteady equations, was used to stabilize the relax-
The natural convection in this complex con- ation procedure.
fipuration can be described by the following set of A non-uniform mesh with a large concentration of
dimensionless equations: nodes in regions of steep gradients, such as close to
the solid portions, is employed. The computations
dU (?V were generally performed on a 50 x 50 mesh with an
acceptable grid independence [2]. Convergence was
assumed when the L2 residual norm of major variables
was no more than 0.001% per step. The conservation
of energy was satisfied; in the worst case the error
was no more than 1%. The computational time on a
(2) VMS/VAX-6530 computer ranged from 25 1o 4(1CPU
918 Y.S. SUN andA. F. EMERY

man depending upon the internal and external Ray- 2O

leigh numbers. k*=l : ,,
15
}"
i .¢
COMPUTATIONAL RESULTS Nu 10 ............... r
.f
i :
_
I
The dimensionless parameters that must be speci- 5
fied for the system are : Ra~, Rai, Pr, k*, the location
of the heat source, Yh, the location of the baffle, ~'b, 0

the height of the baffle, hu a n d the void fraction, ~, 0.3 0.5 0.7 0.9 1.1
which is defined as the ratio of the fluid area to the area void f r a c t i o n
of the entire cavity. Since so m a n y basic dimensionless
parameters are required to characterize the system, a
comprehensive analysis o f all c o m b i n a t i o n s of these
parameters is not practical. The numerical results will
20
be aimed at a small fraction of the possible situations,
and will explain the effect of several critical parameters
15 •
t7. - n"=:l'~/
,i.-,.< ........... ] k.:,o ................ .; 7" "
. ....
in the experimental calorimeter by simplifying the con-
ll,,-l.l~7 / . " "
figuration. Nu 10 ................. + ..;...= ~
. . -:" { ,,, - ''
E[lbct q[ conducth:e walls 5 - L , ": .................

i
In order to investigate the effect of the conductive
0 i i
walls (void fraction a n d conductivity ratio), the
internal baffle and heat source of Fig. I are neglected. 0.3 0.5 0.7 0.9 1.1
Figure 2 shows the variation of the average Nusselt void f r a c t i o n
n u m b e r for different void fractions and conductivity
ratios. As the external Rayleigh n u m b e r increases,
the average Nusselt n u m b e r increases. F o r the larger
thermal conductivity ratios k*, the heat transfer rate
25
is higher t h r o u g h the wall t h a n the fluid, due to a larger
temperature gradient at the solid fluid interface. The 20 .........
fluid in the cavity behaves as a n insulator for a wall 15 ......... i
of such large thermal conductivity, but, for a small Nu k*=
thermal conductivity enclosure, the heat transfer 10 .............................. ~-

t h r o u g h the solid is lower t h a n in the fluid. Therefore, 5 ......... ->

for a lower conductivity ratio (k* = 1), an increase in i i
o i
void fraction causes an increase in the average Nusselt
0.3 0.5 0.7 0.9 1.1
number. F o r a higher conductivity ratio (k* = 100),
the average Nusselt n u m b e r decreases with an increase void fraction
in the void fraction. F o r a mild conductivity ratio Fig. 2. Effect of void fraction and wall conduction o11 average
(k* = 10), the effect o f the void fraction is not sig- Nusselt number.
nificant for a m o d e r a t e Rayleigh n u m b e r (Ra = 10('),
since the effects from the wall c o n d u c t i o n and the the presence of internal heating (RaT > 0), the trends
fluid convection almost c o m p e n s a t e each other. The in the flow and heat transfer of the s t a n d a r d bench-
average Nusselt n u m b e r continuously increases for a mark solutions (without internal heating [3]) persist
higher Rayleigh n u m b e r (107) but decreases for a if Ral/Ra~: < 1 (Fig. 3a).
lower Rayleigh n u m b e r (10 s) when the void fraction F o r Ra~/Ralz >> 1, the internal heater causes the fluid
increases. Note t h a t the decrease continues even when to rise along the hot wall a n d then to turn into the
the void fraction falls below 0.5. These results show central region of the enclosure. The flow then moves
that increasing the thickness of a low conductivity d o w n b o t h the hot and cold surfaces (Fig. 3c). The
wall can reduce the heat loss t h r o u g h the walls, but flow adjacent to the upper part of the hot wall is
that the effect becomes marginal when the thickness directed d o w n w a r d and cooled. The overall heat
reaches a certain small value which depends on the transfer characteristics of the enclosure will be
conductivity. F o r highly conductive walls, the thinner described by the average Nusselt numbers, defined as
the wall is, the less heat transfers t h r o u g h it, with a the ratio of the heat flux at the wall if one-dimensional
consequent very strong reduction in Nu. conduction is the only mode of heat transfer.
(~q$/P Y) r <,
E[fects of internal heat sources NUll = -- [1 -- (1 -- XIORaI,/RaF]c
To consider only the effect of a heat source located
at the lower left corner of the cavity, the void faction (~'~i~ Y) ~ i
is set to one a n d the baffle in Fig. 1 is neglected. In [1 + Xl~Ral/RaE] "
 Effects on n a t u r a l c o n v e c t i o n heat transfer t) 19

; , ~ z E _ ! . . ! :.... ! ! .: - : z-z.,,,,

g/i, . . . . . . " " "'qlil

i,,. "''7 2". 7 " -_ "_ " . ' _ ' . - ' ' - " q L
!tll::: ............. :'3t
'~ . . . . . iiii 'i1 {-
'lhl,
. . . . . tft
i lill
1 =-5.00E-01
2 = -4.00E-01
'i~,.] !tll~l 3 = -3.00E-01
i~,.. !111~1 4 = -2.00E-01
h ...... ' lllll

il!t
1~,.. ' [llil 5 = 1.00E-01
. . . . . . . .. 11 6 = 0.00E+00
lll~ . . . . . . . . . . . . . . . . . . li 7 = 1.00E-01
8 = 2.00E-01
~'~'-Z''2"'" -" - 2 : ." 2 7. 2 " ";: I 9 = 3.00E-01
s = 4.00E-01
# = 5.00E-01

(a)

.............. - ~ ' - Z Z 2 -7. 2 2 2 - 2--'~'",'i

,, e l # , B I l l / " .~ . - ' . . . . . . . •

., ,~11m"17/7 : . . . . . " '

.,llllH~/t: . . . . . . "" .. "''~till"

1 =-3.00E-01
t,,llitt~,, " . 7].. "'~tlili"
I lllID ,. " ' ' ~l~P"
2 = - 1.00E-01
. . . . . . l t lt.ll,, 3 = 0.00E+00
,lilili]lll I I ,
,dl]]~lil , , . ' llllil" 4 = 1.00E-01
5 = 2.00E-01
6 = 3.00E-0t
l,lillI]ll~ . . . . . . . . . . . . . ' , I I ll'
l,Illillltt~ . . . . . . . . . . . . , I Iv 7 = 4.00E-01
,I . . . . . . . . . . 2 . _ " I"l 8 = 5.00E-01
. . . . . . . . . . . . . . . Z .... ' ~l ~,',:1 9 = 1.00E+00
s = 2.00E+00
# = 3.00E+00
li,~iil i i ~. l ~ i i iiiili :,

(b)
II!'-'!'!US~ ! . ! :: .

L'.:Z'-Z: • L ; " : L 2 22 27.-]~I"Y

. . . . . . ~ ~ i il l' 1 = 0.00E+00
_.,#/Jill" " ": : . . . 2= 5.00E-01
..n#lllL" ....... - . . . . . . . t , ~ kll~' 3= 1.00E+00
.n/gill' .... .... i , , i~ i" 4 = 1.50E+00
ltllT#~, ... ] . ....... "' 5= 2.00E+O0
6 = ZSOE+00
7 = 3.00E+00
~,----~ . . . . . . . . _-.5 . . . . . . . . . 8= 5.00E+00
9 = 1.00E+01
s= 2.00E+01
# = 3.00E+01

(c)
Fig. 3. Velocity p a t t e r n s and i s o t h e r m s for Rm = 106 and (a) RaF = 106, (b) Ray = l0 t, (c) Rm- = 104

B a s e d o n t h e a b o v e d e f i n i t i o n s , t h e r e is a s i n g u l a r i t y defined will be quite irregular in the neighborhood of

a t RaE = 0 . 9 2 5 R a i (Xh = 0 . 0 7 5 ) f o r Nun i n t h e Ray = 0.925Rai.
geometry considered. At this value, the theoretical In the absence of an internal energy source, the
conduction heat flux due to the externally imposed average heat flux ratio along the hot wall (Null)
temperature difference and the internal heat sources increases monotonically with RaE. T h e flow adjacent
cancel each other. It should be noted that because of t o t h e h o t w a l l is a l w a y s u p w a r d a n d t h e d i m e n s i o n l e s s
this singularity the behavior of the heat flux ratio as t e m p e r a t u r e g r a d i e n t a t t h e h o t w a l l is h i g h e r t h a n
 920 Y.S. SUN and A. F. EMERY

30

.,...-0 i .,p
20

Nul0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

" ~ "% "N

U
o10 I

1E+3 IE+4 IE+5 1E+6 1E+7 1E+8

Fig. 4a. Average Nusselt number along the hot wall for cavity flow with an internal heat source,

[ i
l •" " ~ - - g a , ] [ = l e 4

/ i i,4
/ /
15 ~-......... l - lg" Ra,l--le 7 .....................................~ .................
Nu

to )(

0 i t
IE+3 IE+4 1E+5 1E+6 1E+7 1E+8
Sa,E
Fig. 4b. Average Nusselt number along the cold wall for cavity flow with an internal heat source.

that of the pure conduction case. In the presence of influence of the internal heat source on Null also
internal heat sources, the value of Null is governed by becomes unremarkable.
the direction and strength of the flow adjacent to the Figure 4b shows the variation of the average heat
hot wall. At each Raj depending on RaE, part of the flux ratio (Nuc) along the cold wall. For all the values
interior hot fluid flows downward along the hot sur- of Ra~, the value of Nuc increases monotonically with
face forming a counter-direction eddy at the upper RaE because the convective motion always carries
corner near the hot wall (Fig. 3c). If RaE is increased heated fluid to the cold wall. For Ra~/Ra~:> 1, the
at each Ra, the flow pattern changes from one where interior temperature is generally smaller than the cor-
the flow adjacent to the hot surface is downward, to responding conduction temperature since the interior
one where the flow moves in an upward direction over fluid is cooled by the downward flow along the cold
the surface. In Fig. 4a, the influence of internal energy wall. Hence, the increase in heat transfer at the cold
sources becomes progressively smaller for increasing surface due to the convective motion is offset by the
RaE values. In particular, it may be seen that for lower interior temperature. Therefore, the cor-
Ra~/RaE < 0.1, the influence of the internal heat responding Nuc are lower than those for the case
source on Null is negligible. For Ra~/RaE> 100, the without internal heaters. For RadRaE > 100, the
 Effects on natural convection heat transfer 921

10

..Ra.=_l,g6 .........

l~ffil.E5
Nu

Ra=I.E4

k*=lO, ......... k*=O.1 - ~

0 i i /

0 0.2 0.4 0.6 0.8 1

height of baffle
Fig. 5. Influence of baffle height on average Nusselt number for various Ra and k*. Cavity flow with
conductive baffle, without internal heat source.

value of Nuc approaches a constant for each Ray center of the cavity has been destroyed. Figure 6 shows
For Ra~/Rac < 1, the external temperature difference the average Nusselt number variation on the cold
dorainates the fluid motion. The cold wall receives the wall. The influence of increasing baffle height is less
fluid directly from the hot wall, so the Nuc values important for higher RaE and smaller Raw. This is
approach those for the case without internal heat sour- shown in Fig. 6 by the relative insensitivity of Nu to
ces. baffle height, and by the similarity to Fig. 5. However,
when the internal heating increases and dominates the
E[[ect ~f conductive baJfles flow pattern (such as RaE = 105 and Ra~ = 10;), an
To understand the interactions among the solid increase in baffle height significantly affects the heat
baffle, fluid flow and internal heat sources, numerical transfer rate. Figure 7 illustrates that the baffle blocks
results have been given by assuming non-conductive the flow and forces strong secondary eddies in the
walls in Fig. 1. hot chamber in the case with higher internal heating.
Sun and Emery [2] examined the effect of a con- These eddies push warmer fluid through the gap and
ductive baffle in an enclosure without internal heat deliver more heat to the cold chamber (RaE = 105 and
sources (Fig. 5). With an increasing baffle height, the RaE/Ra~ = 100 in Fig. 6). The effect of the centrally-
driving force for the fluid remains while the resistance mounted baffle conductivity on heat transfer is very
to the flow keeps increasing. For a centrally-mounted marginal except for high baffles or a high level of
baffle, as the height of the baffle increases with fixed internal heating.
RaE and k*, the heat transfer decreases. The influence It is of interest to understand how the heating ratio
of the increasing baffle height is less important for affects the average heat transfer. Figure 8 shows the
higher RaE, except when the height is so large that the conjugate effect of the heat source ratio and the height
top and bottom ends interfere with the main velocity of the baffle on the average Nusselt number at a Ray-
stream. It also shows that the effect of baffle con- leigh number of 105. It shows that, for a small ratio
ductivity on overall heat transfer is very marginal. (Ra~/RaE = 10 or 15), the presence of a baffle is similar
However, it is noted that for significantly high baffles, to the case without internal heating--the external
the, effect of conductivity becomes more obvious and Rayleigh number still dominates the flow. Once the
the'. velocity vectors and isotherms are significantly ratio increases to Ra~/RaE= 30 (a critical value for
different for the different conductivities. RaE = 105), the Nu is almost a constant no matter
In the presence of internal heating (Ra~ > 0), the what the baffle height is. This critical value is the result
highest temperature occurs near the internal heater. of the interaction of the effect of internal heating,
This makes the fluid rise along the hot wall and then baffle height and baffle conductivity. Above the criti-
turn into the central region of the enclosure. This cal ratio, the baffle effect becomes more significant. In
phenomenon increases the interaction between the Fig. 6, at Ra~/RaE = 100 for RaE = 106, the trend is
baffle and fluid, since the wide stagnation zone in the still the same as the case without internal heating.
922 Y.S. SUN and A. F. EMERY

2 ............ ,..II~IR~.I~,~I 0 . . . . . . . . . . . . . . . . . . . . . . i................. ~'- . . . . . . . . . . .~

. . . . . . . . . . k.~l I ~" "

0 i i
0.2 0.4 0.6 0.8

height of baffle
Fig. 6. Influence of baffle height on average Nusselt number for various Ra~, Raf and k*. where the ratio
is Ral/RaE.

Obviously, the higher the external Rayleigh number, window on which the measurements are to be made.
the higher the value of the critical ratio will be. A schematic diagram of the mimic box is shown in
It is also of interest to observe the effect of baffle Fig. 10. The test section is a rectangular box of 57, 20
when it is located in different locations inside the and 57 cm length, width and height, respectively. The
cavity. For a cavity with a fixed length of baffle walls are made of 2.54 cm thick styrene foam (k = 0.32
(h~ = 0.9) and no internal heat source, Fig. 9a W mC ') on five sides, with the glass window (k = 1.4
(RaE = 105) shows that a conductive baffle increases W mC ~) on the sixth side. There is a baffle located
the overall heat transfer of the enclosure when the 8.26 cm from the rear, with a 5.08 cm air space at the
baffle is located close to either of the vertical walls, top and bottom to create a natural convection flow of
while an adiabatic baffle decreases the heat transfer. air washing down over the cold window and up over
For a high Rayleigh number (Ra~ = 10°), the con- the heater coil. The baffle is made of 1.27 cm thick
vection totally dominates the fluid flow and a large styrene foam.
stagnation zone occurs in the cavity. The effect of A resistance wire heater heats the box when the
baffle location and conductivity is significant only in inside temperature of the box is lower than the room
the boundary layers. In the presence of an internal air temperature outside the mimic box. Since the
heat source at the left lower corner of the cavity, the difference between the temperature inside the mimic
heat source increases the flow strength and almost all box and room temperature is kept small and the styr-
the heat transfer is by convection in the fluid instead ene foam construction is assumed to be a good thermal
of conduction through the baffle. Figures 9c and 10b resistance, little heat is expected to be lost from the
show that the effect of baffle conductivity is negligible box to the room. Most of the heat lost from the box
in the presence of internal heaters. must then be lost through the window. The mimic
boxes measure only the heat lost to the outside, but
do not measure any heat gained from the outside.
COMPARISON OF EXPERIMENTAL DATA WITH
Only night time data are used for this study, so that
PREDICTIONS
heat gain from the outside or solar heat gain are not
Experiments a problem. The heating coil corresponds to a line
The experimental apparatus used is part of the source of length 2.74 m. Alternating current was sup-
H O U S E project at the University of Washington [15]. plied to the line source. The voltage rating was mea-
The special apparatus used, called mimic boxes, are sured for each individual mimic box: nominal value
controlled heating calorimeters and are used to mea- was 50 W.
sure the window heat flux. Since glazing materials The temperature distributions in the mimic boxes
transmit solar radiation and their thermal resistance were measured using the two terminal integrated cir-
is small, the technique employed tbr these measure- cuit temperature transducers manufactured by Anal-
ments is to utilize a calorimeter within a closed box- ogy Devices, Inc. and designated as AD590s. The
like construction, one side of which is formed by the locations of the AD590s are provided in Fig. l la.
 Effects on natural convection heat transfer 923

rectangular cavity with an aspect ratio of 2.86. con-

ductive walls and baffles, and three internal heat sour-
;:;C::: i ; " ; 7; ; . ; ; : : .: ". c e s - t w o in the lower cavity and one in the upper
J~l//~~'''" ' ~ x . . . . . . . . . ~ill
cavity, all located behind the baffle. The average
,~, x x \ \ x ,
lq~i . '. \ \ -,
%
i values of the dynamic temperature measurements are
pl~,l~t' . . .
\
. .
",
.
\
.
\
. . . . . . . . . l ill assumed to approximate steady state restllts. The
q~lu ~-" . . . . . . . . . . . ' I tI i
_. \ \ \ . . . . . . . . . , oscillation of the measured temperature data shows
l
iH that these average values have a maximum error of
I # C . A large temperature difference was found
between the inside glass surface and the outdoor
. J i l l environment [14]. Different temperatures are mea-
, \"~ .- I / /,t! i ...... i llllL] sured at the glass surface and obviously the outdoor
temperature cannot be used as a coristant temperature
~, . . . . I J ,t I l l boundary condition. Instead, a simulation of heal
~m~, z/'," ~ ' ,t ".. . . . Z.2 -" " '/ /z l
. ~ ~', transfer through the exterior air film and the glass
itself is needed. The exterior air film is assumed to be
a solid zone, as is the glass, but with the conductivity
based on the free convection heat transfer coefficient
and thickness of the air layer. This idealized boundary
(a) condition does not give the exact temperature profile
on the surface of the glass, but this method does give
a reasonable temperature profile on the cold surface
of the cavity. An example of the comparison between
the measured temperatures and predicted values on
the glass surface is shown in Fig. t2a. Since the
exterior condition is not measured, variable triai and
error processes are used to give an approximation to
the measured surface temperatures. Even though the
,,,,,mr i z " - t
measured surface temperatures are not very accurate
..... ,zt/lt., . . ., ~,' .,~¢ r r , , ~ ~ [ ~ l~ due to the difficulty of attaching sensors onto the
,otl/17T/l ," ~i I I I surface, the approximation is more reasonable than
.,mtll]7// " ,~ I t 7 the use of constant outdoor temperature.
A sample of" resulting comparisons is shown in Fig.
12b. Note that there are very large deviations between
the measurements and numerical predictions using the
" 1 i 1' outdoor temperature as the boundary condition of the
cavity (without the simulation of exterior air layer),
but the numerical results are improved by simulating
the exterior air layer as a solid wall with effective
conductivity. The agreement between the improved
prediction and measurement is good. The deviation is
primarily due to the boundary condition modeling
(b) approximations at the glass surface. The velocity vec-
Fig. 7. Velocity patterns for Rau = 10" and (a) Ra I = 1()?, tor patterns and isotherms are shown in Fig. 11 b and
(b) R a l - i06. c. Temperature stratification in front of the baffle
shows that the baffle successfully creates a natural
convection flow in the frontal cavity, except in the
regions near the top and bottom.
The differences between the predictions and
The AD590s were positioned approximately 4.13 cm measurements are the net result of several factors:
behind the glass pane. The indoor and outdoor tem- such as the three-dimensional effect, the effect of radi-
peratures were not controlled, but were allowed to ation, the effect of the edge and frame of the window,
vary with the environment. The indoor temperature the exterior forced convection due to the wind, the
was maintained at a nearly constant value by use of a dynamic weather conditions, the temperature strati-
thermostat, while the outdoor temperature depends fication on the vertical walls, the imperfection of the
totally upon the weather conditions. insulation at the horizontal walls, the difficulty ol
holding the temperature sensors in the air and the
Comparison g / e x p e r i m e n t a l data with predictions uncertainties in thermophysical properties etc.
To find the two-dimensional natural convection at Several test nights with various temperature differ-
lhe central plane, the mimic box is simulated as a ences are examined (Table 1). The temperature differ-
924 Y.S. SUN and A. F. EMERY

, i [] =

Nu

- o - ratio=50 ~

•~ e - ratie=30

-.. ~ . ratio=15 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

ratio=lO

0
0 0.2 0.4 0.6 0.8
height of baffle
Fig. 8. Effect of heating ratio on the average Nusselt number (cold wall) at RaE = I(F.

ence from l0 to 25'C causes the external driving force void fraction increases the heat transfer. For high
equivalent to RaE from 5 x 107 to 5 × 108 and the ratio conductivity walls, the average Nusselt number
of internal-external heating (Ral/RaE) from 5.5 to 7.5. decreases with an increase in void fraction.
It is noted that a large percentage of heat loss (q) is For an enclosure with a conductive baffle, results
contributed by the hot wall, not only from the heaters indicate that heat transfer is strongly influenced by
as assumed. This result indicates that the conduction the coupling effect among baffle conduction, fluid con-
of the cavity walls is not negligible. However, the vection and the strength of internal heating when the
percentage of heat loss through the hot wall is in the baffle is located near walls. The increasing resistance
range 3 3 4 0 % based on the range of temperature to the fluid due to the baffle decreases the overall heat
differences. This might be used to correct the present transfer. When the baffle is conductive, heat can still
calculation of window U-values by replacing Q with be transferred to the rest of the enclosure by conduc-
qc ( = Q / 1 - ~ / ) . tion. When the baffle is adiabatic or when the con-
vection dominates the flow due to a high level of
internal heating, heat is transferred by convection no
CONCLUSIONS
matter what the value of the baffle conductivity. The
The qualitative and quantitative understanding of influence of the baffle become marginal when it is
the influences of conjugate conduction-convection located centrally, since it then sits in a wide stagnation
heat exchange and internal-external heating are pre- zone in the center of the cavity for higher Ra flow.
sented in this study. When internal heating is used A practical window calorimeter is simulated as a
inside the cavity, variations of the numerical solution parametric study. Qualitative agreement has been
from the benchmark solution (without internal heat found between the simulations and measurements. In
sources) show up only when the heating ratio (Ra,/ quantitative comparison, deviations were generated
RaE) is in the range of 1 10. For a ratio less than 1, due to uncontrollable and dynamic boundary
internal heating does not affect the flow pattern and conditions. By modifying the simulation of the
heat transfer, and the external temperature difference exterior air film and glass pane, better boundary con-
is still dominant. For a ratio greater than 10, the ditions (temperature profiles) have been derived, and
average Nusselt number tends to reach a constant the deviations have been reduced to less than 25%.
value. The 25% maximum relative error is quite good con-
When using conductive walls, the average Nusselt sidering the many factors which affect the accuracy of
number is found to increase when the conductivity the experiment. Calculation of the window U-value
ratio (solid-fluid) increases. The effects of the void should be corrected by considering the heat loss
fraction depend on the conductivity of the wall through the walls, but it is not practical to actually
material. For low conductivity walls, an increase in calculate the amount of heat loss for each case. The
 Effects on natural convection heat transfer ~25

......../ ...............................................................i .... i

Nu 2 Nu 4 .... ............................. • k*=10

A ""~"" k*=0.1
. . . . . . . . . . . . ~........... ---~-.- k*=O.l ...................................

• k*=10

0 i

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

location of baffle location of baffle

(a) without internal heat source, Ra E= 10' 0eft), Ra E= 106 ( l i g h t )

0.2

-0.2 i ..... "'-- - = "i --

Nu Nu 4
-0.6
....... ii ...... 1 k--lO t il
-1 i i _

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

location of baffle location of baffle
(b) low level of internal heating, Ra~ = 104, ~/RaE=10 , left(hot wall), right (cold wall)

0.75
...........
................~ ..~. f.._ .e.=:=}
_ 7~_ _......a.........~=.:..:.~.:.:
~ ! ........
Nu 0.5 'iii ii!' Nu4

0.25 t . k*=10
--~--- k,=0.1
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
location of baffle location of baffle

(c) high level of internal heating, Ra E= 106, Rar/Ra~=100, left(hot wall), right (cold wall)
F i g . 9. E f f e c t o f baffle l o c a t i o n s o n t h e a v e r a g e N u s s e l t n u m b e r , w i t h h~ = 0.9.
926 Y.S. SUN and A. F. E M E R Y

front view

I
spacing blocks i
i
•15zm
h, _ng ~'\
w: baffl

heating wire

-F7 / FT
2.5 Ic]

J >
57.15cm

window surface
/\

7.94cm
/
/ 20.0cm

8.26cm

top v i e w
Fig. 10. Schematic diagram of mimic box.
 Effects on natural convection heat transfer 927

57.15cm

-~ 53.34cm

I + 49.s~
~- 44.45cm

, ~- ] 38. l O c m
I
I
i
+ 3 .7sc.,

"4- 25.40cm

~- 19.05cm

•4- 12.70cm

~ -- - - 6.99cm
5.84cm
3,81cm
2.54cm

4. 1 3 c m
Fig. 11. Locations of temperature sensors (a), sample of el velocity pattern (b) and (c) isotherms of mimic
box (window' calorimeter).

Table 1. Results of numerical simulation of window calorimeter

1 2 3 4 5

'1,~ [C] 22.17 20.02 20.51 25.03 19.64

~/~ [C] 1.66 0.14 7.51 0.00 7.06
&T [C] 20.51 19.88 13.00 25.03 12.58
RaF ( × 10 ~) 3.346 4.505 0.64 5.4 3.83
Ral/Ra~ 5.519 6.532 5.813 7.268 5.884
k* 12.698 12.78 12.55 12.76 12.74
q~ 8.272 1 1.244 I 1.901 13.475 10.702
q~ 24.83 30.84 29.34 35.279 28.234
q (%) 33.3 36.46 40.56 38.2 37.9

qH = - i}O/@ at the hot wall, positive to the exterior.

q~. . . . . (?0/2 Y at the cold wall, positive to the exterior.
 928 Y.S. SUN and A. F. EMERY

0.6

predicted ~,.

0.3 ................................... • measured " i .......................................i.................................

0
0

-0.3

Y
0 0.2 0.4 0.6 0.8 1

(a)
0.8

0 0.4 ~'"'--2~
o .................. iI .. " ../
.................... . ~, " ~ ......................i. . . . . predicted(2)

. . . . i. " measured

'" " " predicted(l)

-0.4

0 0.2 0.4 0.6 0.8 1

Y

(b)
Fig. 12. Comparison of predicted and measured data with TH = 20C, 7"(-- 2C and A T = 18c. (a) Tem-
perature profile on a glass surface, (b) temperature profile inside the cavity: predicted data (l) without, (2)
with air layer simulation.

range o f h e a t loss p e r c e n t a g e (t/) given by the p r e s e n t natural convection in enclosures with a conductive baffle,
investigation m i g h t be valuable for p r e s e n t p a r a m e t r i c Numer. Heat Transfer 25, 575-592 (1994).
3. G. de Vahl Davis, Natural convection of air in a square
cases.
cavity : a bench mark numerical solution, Int. J. Numer.
Meth. Fluid 3, 249-264 (1983).
REFERENCES 4. D. R. Chenoweth and S. Paolucci, Natural convection
in an enclosed vertical air layer with large horizontal
1. M. W. Wong, An experimental study of window man- temperature differences J. Fluid Mech. 169, 173 210
agement of nine different types of window products. (1986).
M.Sc. Thesis, Department of Mechanical Engineering, 5. R. A. W. M. Henks and C. J. Hoogendoorn, On the
University of Washington, Seattle, WA (1990). stability of the natural convection flow in a square cavity
2. Y. S. Sun and A. F. Emery, Multigrid computation of heated from the side, Appl. Sci Res. 47, 195220 (1990).
 Effects on natural convection heat transfer 929

6. F. P. lncropera, Convection heat transfer in electronic 11. D. M. Kim and R. Viskanta, Effect of wall heat con-
equipment cooling, J. Heat Transfer 110, 1097--,1111 ductance on natural convection heat transfer in a square
(1988). cavity, J. Heat Transjer 107, 139 146 (19851.
;7. S. Acharya and R. J. Goldstein, Natural convection 12. S. V. Patankar, Numerical Heat Tran,ffer and Fluid
in an externally heated vertical or inclined square box Mechanics. Hemisphere, Washington, DC (t980).
containing internal energy sources, J. Heat Transfer 107, 13. A. Pollard and A. L. Siu. The calculation of some lami-
855 866 (1985). nar flows using various discretisation schemes, Comput.
8. J. Davalath and Y. Bayazitoglu, Forced convection cool- Meth. Appl. Mech. En,qn9 35, 293 313 (1982).
ing across rectangular blocks, J. Heat Transfer 109, 321 14. Y. S. Sun, Conjugate natural convection heat transfer in
328 (1987). a rectangular cavity containing internal heat sources.
0. M. Afrid, and A. Zebib, Natural convection air cooling Ph.D. thesis, Department of Mechanical Engineering,
of heated components mounted on a vertical wall University of Washington, Seattle, W A (1994).
Numer. Heat Transfi, r 15, 243 259 (1989). 15. A. F. Emery, C. J. Kippenhan, G. B. Varey, J. L.
10. Y. Joshi. L. O. Haukenes and S. B. Sathe, Natural con- Garbini, J. H. Heerwagen and D. R. Heerwagen, 1"he
vection liquid immersion cooling of a heat source flush Thermal Interaction of Envelope Components in Slandard
mounted on a conducting substrate in a square enclos- and Improved Houses in the Pacific Northwest, HTD-Vol.
ure. hit. J. Heat Mass Tran,~[br 36, 249 263 (1993). 123, pp. 153-162. ASME, New York (1989).