SERI/TP-252-2812

UC Category: 59b
OE86004424

Natural Convection in an
Enclosure with Discrete
Roughness Elements on a
Vertical Heated Wall

S. Shakerin (Colorado State University)

M. S. Bohn (SERI)
R.I. Loehrke (Colorado State University)

February 1986

Prepared for
8th International
Heat Transfer Conference
San Francisco, California
17-22 August 1986

Prepared under Task No. 3751.21

FTP No. 01-527-86

Solar Energy Research Institute

A Division of Midwest Research Institute

1617 Cole Boulevard

Golden, Colorado 80401-3393

Prepared for thA

U.S. Department of Energy
Contract No. DE-AC02-83CH10093
 NOTICE

This report was prepared as an account of work sponsored by the United States Government. Neither the
United States nor the United States Department of Energy, nor any of their employees, nor any of their
contractors, subcontractors, or their employees, makes any warranty, expressed or implied, or assumes any
legal liability or responsibility for the accuracy, completeness or usefulness of any information, apparatus,
product or process disclosed, or represents that its use would not infringe privately owned rights.

Printed in the United States of America

Available from:
National Technical Information Service
U.S. Department of Commerce
5285 Port Royal Road
Springfield, VA 22161

Price: Microfiche AOi

Printed Copy A02

Codes are used for pricing all publications. The code is determined by the number of pages in the publication.
Information pertaining to the pricing codes can be found in the current issue of the following publications.
which are generally available in most libraries: Energy Research Abstracts, (ERA); Government Reports
Announcements and Index ( GRA and I); Scientific and Technical Abstract Reports (STAR); and publication,
'
NTIS-PR-360 available from NTIS at the above address.
55�1 TP-2812

NATURAL CONVECTION IH AH ENCLOSURE WITH DISCRETE ROUGHNESS

ELEHEHTS OH A VERTICAL HEATED WALL

s. Shakerin , M. Bohn*, and R. I. Loehrke

+ +

ABSTRACT average wall heat flux

1 p
Natural convection flow next to a heated
ii J0
wall with single and repeated, two-dimensional,
rectangular roughness elements is studied numeri­ (TH - Tc) LJ
cally and experimentally. The objective is to Ra1 g S va Rayleigh number
s
determine how these roughness elements influence
spacing between roughness
heat transfer rates from the wall. Each rough­
elements
ness element consists of a thermally conducting, TH + Tc
horizontal cylinder of rectangular cross section Ti initial fluid temperature for
2
attached to the heated, isothermal wall of an transient calculation
enclosure. The height of roughness is on the u horizontal velocity
order of the boundary layer thickness. Dye flow u uL/a dimensionless horizontal
visualization in water confirms the numerical velocity
prediction that the steady flow over these ele­ v vertical velocity
ments does not separate. Only at high Rayleigh v vL/a dimensionless vertical velocity
numbers, when the boundary layer below the rough­ X vertical dimension
ness is unsteady, is local instantaneous flow X = x/L, dimensionless space
reversal observed. Although steady flow rever­ variable
sals near the wall are not predicted or observed, y horizontal dimension
nearly stagnant regions are formed, particularly y y/L, dimensionless space
between closely spaced cylinders. The surface variable
heat flux in these stagnant regions is relatively
Subscripts
low, so the total heat transfer rate may be
nearly the same as for a smooth wall in spite of c cooled wall
the increased surface area. H heated wall
1 lower roughness element
2 upper roughness element
NOMENCLATURE
Greek letters
A aspect ratio = H/L
a thermal· diffusivity
S
H height of test cell
coefficient of thermal
k thermal conductivity
expansion
L distance between heated and
T - T·
cooled surfaces in the test
cell
9
TH - C
�, dimensionless

height of roughness element temperature

v
local Nusselt number kinematic viscosity
II ljl stream function,
. qw L/k (TH - Tc)
a
ljl - u -aljl = v
average Nusselt number ay - ' ax
IT
qw L/k (TH - Tc) 1• IMTRODUCTIOH

p distance along wall

An understanding of natural convection heat
p length of heated surface
including roughness transfer is essential for proper design of build­
Pr Prandtl number ing energy systems, particularly for those build­
ings where passive heating and cooling techniques
II
local heat flux

+
Mechanical Engineering Department, Colorado State University, Ft. Collins, CO, 80523.
*Solar Energy Research Institute, Golden, CO, 80401.

1
 !II-II
s=�• *
� TP-28 12

are employed. Most of the past experimental and 2.1. Numerical

analytical work on natural convection heat trans­
fer from surfaces has focused on smooth surfaces. "The unsteady, two-dimensional Boussinesq
However, the surfaces of buildings are seldom equations were solved for the flow in a rec­
smooth. Based on the results of experiments and tangular enclosure. The equations for vorticity,
analysis in forced convection, one might antlcl­ stream function, and temperature were cast in
pate that the roughness associated with surface finite-difference form using forward time differ­
texture and with isolated elements, such as win­ ences, centered differences for the diffusive
dow frames, may significantly influence the over­ terms, and upwind differences for the convective
all heat transfer rate from a wall. If this terms. Each calculation started with the fluid
influence can be predicted with a reasonable isothermal and at rest. The two side walls were
degree of confidence, then surface treatment may maintained at uniform, constant temperatures
become a design variable for heat transfer con­ above and below the initial fluid temperature and
trol. More fundamental data on the influence of the floor and ceiling of the enclosure were adia­
surface roughness. on heat transfer coefficients batic. The calculations proceeded in time until
in natural convection are required to assess the steady-state conditions were achieved. A nonuni­
magnitude of this effect. form, zonal mesh was used with grid points spaced
Anderson and Bohn ( 1] identified two dis­ more closely in the boundary layer and around the
tinctly different ways that the natural convec­ roughness element than in the core. The results
tion heat transfer coefficient may be altered by were checked for grid independence and, for the
roughness. The roughness may introduce distur­ case of smooth side walls, against a previously
bances into an overlying laminar boundary layer, published "benchmark" solution (de Vahl Davis
which cause that boundary layer to prematurely [ 3] ) , for enclosure flows. A detailed descrip­
undergo transition to turbulence. Thus, that tion of the numerical technique is contained in
part of the wall exposed to turbulent flow, which reference [4].
in absence of the roughness would have been in
laminar flow, experiences heat transfer augmenta­ The roughness element height � (see Figure 1
tion. Another way in which heat transfer may be for nomenclature) was ch,osen to be a multiple of
altered is through various mechanisms associated the boundary layer thickness at the center of the
with the locally altered flow near the roughness smooth-walled enclosure at the same Rayleigh num­
element. The heat transfer in fully separated ber. The boundary layer thickness was defined as
zones may be reduced compared to that at the same the distance from the wall to the first zero
location on a smooth surface. On the other hand, crossing of the velocity profile. All of the
local flow acceleration in the attached regions numerical results presented here are for
and unsteady vertical flow originating in the �/o = l; however, this study has been extended
unstable separated shear layer downstream of the to 1/2 $ �/o $ 2 and the results are qualita­
roughness element may enhance heat transfer. tively the same. Likewise, most of the results
An early work in this area is that of Eckert presented here are for an nclosure with aspect
�
et al. (2 ] who used smoke flow visualization to ratio A= H/L = 1, RaL = 10 , Pr= 0.7, but the
study the effect of a single, rectangular rough­ study has included l $ A $ 3, lOS $ RaL $ 108,
ness element on the apparent position of transi­ and Pr= 0. 7 and 7.0 with qualitatively the same
tion on a vertical heated plate in air. They results.
reported a decrease in the distance from the
leading edge of the plate to the beginning of
transition of up to 15% when an element with a
height of about 3/4 of the boundary layer thick­
ness was added to a smooth plate.
T
The present work deals with the same funda­
mental roughness geometry but focuses on the flow
and heat transfer processes in an enclosure in
the immediate vicinity of single and repeated
rectangular elements. Our interest in enclosures
stems from the ultimate application mentioned
earlier, building heat transfer. Due to strati­
H

1-t
fication and flow recirculation in enclosures,
results from classical vertical-plate-in­ s
infinite-medium
applicable.
studies are not immediately
T
x

ill
2. APPROACH

This study was carried out with comple­ L....-----.1

mentary numerical and experimental efforts.
t-----L-----1
Quantitative predictions were obtained over a
range of parameters from the numerical part,
while the qualitative checks on these predictions
and on the validity of the two-dimensional model Figure 1. Nomenclature for enclosure with
employed were provided by the experimental part. roughness elements on heated wall

2
55�1 •
11-11
� TP-2812

2.2. Experimental The flow over a single roughness placed at

X1/H = 1/4 (see Figure 1) remained steady and
Experiments were performed in two different two-dimensional. The local at g
Ray eigh number
test cells. Dye visualization of the flow over this position was Rax = 1.3 x 10 • Dye injected
the roughness was carried out in a water-filled into the boundary la ]6
r below the rod rose along
one-foot cubical test cell. Opposing, smooth the wall and followed the contour of the rough­
vertical aluminum walls. were heated and cooled to ness. Separation, with reverse flow at the wall
drive the flow. The other two vertical walls or at the surface of the roughness, was never
were made of lucite to permit viewing. The observed in the experiments provided that the
roughness elements for this test were 6.3-mm­ element was located where the flow over the
square aluminum rods that were glued to the smooth wall was steady. Instantaneous flow
heated wall using RTV silicone adhesive. The reversal and shedding from the rod were observed
axis of each $Od was horizontal, and the rod only when the flow approaching the rod was
spanned the entire heated surface. Details of unsteady. Numerical calculations confirm the
the test cell are presented in reference [5]. absence of flow. reversal in steady state for
The floor of the test cell was insulated and the elements attached to the wall of an enclosure.
top was open to allow insertion of a 1.5-mm­ Instantaneous flow reversal above the roughness
diameter dye injection probe. The temperature is predicted only at early times in the start-up
difference between the hot and cold walls was transient.
maintained at 14°C giving a Rayleigh number based The tenacious nature of the boundary layer
9
on the test cell dimensions of 8.3 x 10 . is vividly demonstrated in the flow over two
closely spaced elements. The flow with S = 21 is
shown in Figure 2. In Figure 2a the dye intro­
The second experimental apparatus used in
duced near the wall below the elements follows
this study included an air-filled enclosure and
a Mach-Zehnder interferometer with a 150-mm­ the wall and indicates the unidirectional nature
of the flow. The dye in Figure 2b was introduced
diameter field of view. The opposing, active
aluminum walls of the enclosure are 51 em wide near the maximum velocity point and shows that
the main flow dips in near the wall between the.
and 111 em high and are separated by 31 em. The
remaining surfaces of the enclosure are insulated two elements. A similar set of streak lines is
shown in Figure 3 for rods set at S = 1. Here,
lucite. Moveable optical windows, placed at
the main flow bridges the gap between the rods
opposite sides of the heated plate, expose the
boundary layer to one of the interferometer but the flow inside the gap does not reverse.

beams. A detailed description of this apparatus Similar quantitative results from the

is contained in reference [ 6]. For the purpose numerical calculations are shown in Figure 4 for
6
of .the testing reported here the enclosure side the flow in a square enclosure at Ra1 = 10 for
S = 21 in Figure 4a and S 1 in Figure 4b. The
walls (except for the windows) were insulated
=

externally while the floor and ceiling were boundary layers are relatively thick, so the

insulated internally. The internal rigid foam roughness elements are large. The flow in the

insulation reduced the height of the walls to entire enclosure is steady; the flow in the

91 em, creating an enclosure with an aspect ratio vicinity of the roughness is similar to that

of 3.0; the hot wall and cold wall of the visualized in the water tank.

enclosure were maintained at 40°C and 11°c,·

respectively, yielding a Rayleigh number of f, l
= 5.3
7
Ra1
active plates.
x 10 based in the span between the
..T .

The roughness elements for this test were

12.5-mm-square aluminum bar stock and·attached to
the heated wall with RTV silicone adhesive. The
heated wall was maintained at a uniform temper­
ature to within ±O.S°C using individually
controlled electric heater strips. Thermocouples
attached to a roughness element indicated that
heat conduction through the adhesive was suffi­
cient to keeg the element at the same temperature
(within ±0.5 C) as the hot wall.

3. RESULTS

3.1. Fl ow Field

The flow fields around single and repeated

roughness elements were studied numerically and
experimentally in the water tank. Dye injection
at the base of the heated wall showed that the
Dye injected near wall Dye streak near velocity
-
flow, in absence of the roughness element, was a. b.
steady in the lower half of the enclosure. Some­ maximum
what above the center of the wall the dye streak
lines become sinusoidally distorted and rolled up Figure 2. Dye visualization of flow over two
into vortices before reaching the upper surface. bars; S = 21

3
 TP-2812

!
'l CD

�·
'\""'
..
'c
;
---- 0.10 --
--

....
a

N
0

0
a. Dye Injected near wall b. Dye streak near velocity a �------�--���--�-===�
maximum 0.2 o.� o.6 o.a 1.0

Figure 3. Dye flow visualization for two bars;

s = II. Figure S. Calculated isotherms for flow over
�
single roughness elem nt in an enclo­
sure; A = 1, RaL = 10 , Pr = 0.12

3.2. Heat Transfer

The influence of a single roughness con the wall of the air-filled enclosure is shown in the
boundary layer heat transfer can be seen from the interferogram in Figure 6. This element is
distortion of the isotherms. The calculated attached at x /H = 1/4 so Rax = 2.2 x 107. Both
1
isotherm pattern in an enclosure of A = 1 and the calculated and measured! isotherm patterns
Ra = 108 is shown in Figure S. A closeup view indicate that the surface heat flux is reduced
L along the wall just below and just above the
of the roughness element and isotherms on the

(0
c::i
<D
0

X X

... ....
a a

Figure 4. g
Calculat d streamlines for flow over two roughness el ements in an enclosure; A = 1,
RaL = 10 , Pr = 0.72

4
55�1 TP-2812

'I

element compared with what it would be on a

smooth wall. The temperature gradient at the
wall
outer surface of the element appears to be high,
especially near the lower, outside corner.
These observations are quantified in Fig­
ure 7, where the local Nusselt number, based on
the calculated isotherms, is plotted as a
Nu
function of position along the surface. The
magnitude of the Nusselt number on the rough
surface is indicated by the distance between the ----- Nu for smooth wall
---- Nu for wall with roughness
surface and the solid curved lines. The dashed
line indicates the Nusselt number along the
smooth wall at the same Rayleigh number. It can
be seen that the influence of the roughness is
mainly localized to within about two roughness
heights above and below the roughness location.
It also appears from this plot that the heat
transferred through the extra surface area added
by the roughness (the horizontal surfaces) should
just about balance the reduction in heat trans­
ferred from the wall immediately below and above
the element. This observation turns out to b g
roughly correct, as can be verified for Ra = 10 70 0
L
by comparing the first two entries in Table 1. Nuc
Here, the average Nusselt number Nu over the
L
entire heated surface is presented for several
geometries. This Nusselt number is based on the Figure 7. Calculated local Nusselt number for
total heat transfer rate from the heated surface flow on heated wall in an enclosure
and can therefore be used as a direct measure of with and withou � a roughness element;
this influence of the roughness on overall heat A = 1, Ra 10 , Pr 0.72
1
= =

transfer from a surface. The

Nu ' increase in
L
shown in Table 1, for a single roughness compared
to a smooth wall is 12%. This is much smaller The interferograms for double roughness
than the 32% increase in surface area due to the elements shown in Figure 8 indicate that for
addition of a single roughness at this Rayleigh closely spaced elements the heat transfer from
number. the surfaces that make up the cavity between the
elements should be relatively low. Similar
conditions are shown in Figure 9 for the
calculated isotherms for two elements in an
enclosure Ra = 106•
with The addition of a
1
second roughness w1th S = 1 results in no change
in Nu f om l
that for a single roughness at
= 10 •
L
Ra This can be verified by comparing the
L
second and third entries in Table 1. If the
spacing between the two elements is increased to
S = 21, then Nu increases slightly.
L
The average NuL for a single rectangular
roughness element with an overall length (up the
plate) equal to the sum of the two square rough­
ness elements plus the gap between them is also
shown in Table 1. The near equality of the aver­
age NuL for the cases of two elements with S = 1
and for the rectangular element with length = 31
shows that at this close spacing the total heat
transfer rate from the three surfaces of length t
forming the cavity between two roughness elements
is essentially equal to that from a single sur­
face of length t positioned near the boundary
layer.

4. DISCUSSION AND CONCLUSIONS

Eckert et al. [2] interpreted the existence

of steady smoke streak lines over their obstacle
as an indication that a flow-separation bubble
Figure 6. Interferogram showing isotherms
around a single roughness element n � existed with a steady laminar rotation within the
separated flow region. This phenomenon was not
an enclosure; Ra 2.2 x 10 ,
x
Pr = 0.71 1

5
55�1 ·
S - � · 11-11 TP-2812
�

s = 2/ s =f

Figure 8. Interferograms for two roughness elements in an enclosure; Ra 2.2 107, Pr 0.71
1
= x =
x

a ,------r--�Trrm

<C "'
a 0

"' "'
a a

X X

... 0·Do ...

a a

N N
a 0

0 0
0 �����----�--�� a ���--�----��-��
0.0 0.2 o . .; 0.5 0.8 1.0 0.0 0.2 0.4 0.6

'{ '{
s = 2f s =f

Figure 9. Calculated isotherms in an enclosure with two roughness elements; A = 1, Ra = 106,

0.72
L
Pr =

6
 TP-2812

Table 1. Influence of geometry on the the turning of the boundary layer. With conduc­
total heat transfer rate from the ting, closely spaced, multiple roughness elements
g
heated surface o an enclosure the total heat transfer is not significantly
(A= 1; Ra = 10 , Pr= 0.72) different from that for a smooth surface.
L
A reviewer of this paper brought to the
authors' attention a related paper, Bhavnani and
Geometry of Heated Surface NuL
Bergles [8], which generally corroborates the
findings of the present work.
Smooth surface 9.2

Single roughness 10.3 5. ACKNOWLEDGMENTS

Double roughness
s = 1!.
Computer time on the CYBER 205 was gener­
10.3
s= 21!.
ously provided by the Institute for Computational
10.7
Studies at CSU. This work was partially funded
Rectangular r9ughness by the Office of Solar Heat Technology of the
length 31!. 10.2 U.S, Department of Energy as part of the Solar
length= 41!. 10.3 Buildings Technologies Program. The third author
received support for participation in the experi­
mental part of this work that was performed at
SERI from the Associated Western Universities,
observed in the present study: steady separation
Inc.
with reversal of wall shear stress was never
observed in the two-dimensional flow over square
REFERENCES
rdds. The steady natural convection boundary
layers studied here tend to follow the wall con­
tour very closely. The main boundary-layer flow 1. Anderson, R., Bohn, M., Heat-Transfer En­
skips over the gap between closely spaced rough­ hancement in Natural Convection Enclosure
nesses, but even here the slowly moving fluid Flow, ASME Paper 84-Y-955. Accepted for
between elements follows the main boundary layer. publication in J. Heat Transfer.
A recently published, two-dimensional, numerical
study by Oosthuizen and Paul [7] indicates that 2, Eckert, E.R.G., Hartnett, J.P., Irvine, T.F.,
this conclusion concerning the absence of separa­ Jr., Flow-Visualization Studies of Transition
tion may hold even for large surface interrup­ to Turbulence in Free-Convection Flow, ASME
tions that extend horizontally nearly from wall paper 60-WA-250, 1960.
to wall. However, boundary 1ayers flowing over
large heated horizontal surfaces (or under cooled 3. de Vahl Davis, G., Natural Convection of Air
surfaces) are likely to become three-dimensional in a Square Cavity: A Bench Mark Numerical
so the range of validity of these calculations Solution, Int. J. for Numerical Methods in
needs to be further defined. Fluids, Vol. 3, pp. 249-264, 1983.
Heat transfer augmentation may be obtained
in forced convection by placing a two-dimensional 4. Shakerin, S., Laminar Natural Convection on a
Vertical Surface with Discrete Roughness
:
roughness element, such as one of the rods used
in this study, in a laminar boundary layer. The Ph.D. Thesis, Mechanical Engineering Dept.
highly unstable, separated shear layer created in Colorado State University, 1986.
the flow over a single rod undergoes rapid
trans1t1on to turbulence and may augment heat 5. Bohn, M., Kirkpatrick, A., Olson, P., Experi­
transfer for a considerable distance down;;tream mental Study of Three Dimensional Natural
from the obstacle. The experiments reported here Convection at High Rayleigh Numbers, J. Heat
indicate that this mechanism for heat transfer Transfer, Vol, 106, pp. 339-345, 1984.
augmentation is not so powerful in natural
convection. 6, Bohn, M., Fisher, E., Anderson, R., Develop­
The effect of a roughness element on the ment of an Interferometer for Natural Convec­
downstream flow field and heat transfer wa-s not tion· Enhancement Research, SERI /PR-252-2598,
studied in the present work. However, Eckert Solar Energy Research Institute, Dec. 1985.
et al. [2] have concluded, on the basis of flow
visualization, that the transition point may be 7. Oosthuizen, P. H., Paul, J. T., Free Convec­
displaced upstream by about 15% because of a two­ tion Heat Transfer in a Cavity Fitted with a
dimensional roughness. The influence of this �orizontal Plate on the Cold Wall, Advances
displacement of the transition point on heat 1n Enhanced Heat Transfer--1985, Shenkman,
transfer needs further clarification. S.M. et al., eds., ASME HTD Vol. 43, 1985.

Heat transfer may also be augmented by the 8. Bhavnani, S., Bergles, A. E., Interferometric
addition of discrete roughness elements to a wall Study of Laminar Natural Convection from an
simply because the total surface area is Isothermal Plate with Transverse Roughness
increased. The results of the calculations Elements, Ramachandran Festschrift Vol. '
reported here show that in laminar flow, a single Indian Society for Heat and Mass Transfer
'
horizontal bar is a poor fin even if it is Dec. 1985.
perfectly conducting. The gain caused by added
surface area is just about cancelled by the
[Link] in heat transfer coefficient because of