AN ABSTRACT OF THE THESIS OF

Thamir Rauf Al-Alusi for the degree of Doctor of Philosophy

in Mechanical Engineering Presented on October 3, 1989
Tilt le: An Experimental Study of Natural Convection Heat
Transfer From a Horizontal Cylinder Array Vertically
Aligned to and Confined by a Single Wall and Two Walls

A hdract nvins7nRwl
Redacted for privacy
Dwight J. Bushnell
v7

Convection heat transfer was experimentally investigated

for two different geometries. One case had three horizontal
cylinders in a vertical plane placed between two vertical walls.
The second case had three horizontal cylinders in a vertical plane
with only one vertical wall. Several different cylinder center-to-
center spacings were investigated. The wall spacings for the two
wall and single wall cases were also varied.
The cylinders were placed in a still air medium at
atmospheric pressure and were maintained at a constant heat
flux. The modified Rayleigh number, based on the diameter of the
cylinders, ranged from 6.2 x 104 to 1.2 x 106. A scale analysis
was performed to predict the relationship between the Nusselt
number and the modified Rayleigh number. A standard finite-
difference code was employed to show the temperature
distribution and velocity vectors distribution around the
cylinders.
The results showed that there was a maximum heat transfer
from each cylinder at a specific wall-array spacing and a specific
center-to-center spacing. Comparisons of heat transfer results
with a single free cylinder and a free array of horizontal
cylinders were made and discussed. Empirical equations were
proposed to predict the effects of the experimental parameters
on the heat transfer as expressed by the average Nusselt number
of each cylinder or the average Nusselt number of the whole
array.
Flow visualization was accomplished using laser sheets. The
resulting studies showed that the presence of a single wall or the
asymmetrical placement of the array between the two walls
eliminated the presence of the vortices in the spaces between the
cylinders and displaced the stagnation points. The asymmetrical
placement of the array between two walls created a reversed
current between the two walls.
An Experimental Study of Natural Convection Heat Transfer
From a Horizontal Cylinder Array Vertically Aligned to and
Confined by a Single Wall and Two Walls

by

Thamir R. Al-Alusi

A THESIS

submitted to

Oregon State University

in partial fulfillment of
the requirements for the
degree of

Doctor of Philosophy

Completed October 3, 1989

Commencement June 1990

APPROVED:

Redacted for privacy

Professor echanical Engineering in charge of major

Redacted for privacy

Head of Department of Mechanical Engineering

Redacted for privacy

Dean of Graduate/ hool

Date thesis is presented October 3, 1989

Typed by Cathryn Westbrook for Thamir R. Al-Alusi .

 ACKNOWLEDGEMENTS

I wish to thank my major advisor, Professor Dwight Bushnell,

for his guidance and encouragement throughout my graduate
studies. His patience and support over the past five years have
been greatly appreciated. I would also like to express my
appreciation to each member of my committee, Professor
Jonathan Istok, Professor Milton Larson, Professor Alan Robinson,
and Professor Charles Smith, for their time and assistance.
I would like to thank Dr. Donald Trent, visiting faculty from
Battelle Laboratories, for his guidance and input in the use of
Tempest, a computer software program which he co-authored.
Finally, I would like to thank the members of my family for
their unflagging moral support and encouragement. First, I wish
to thank my father, Rauf Al-Alusi, for making it possible for me
to pursue my graduate studies. Secondly, I wish to thank my
aunts, Sabriah and Fatima Al-Alusi, and my brothers and sisters,
for their ongoing support over the past several years. And last,
my love and appreciation to my wife, Cathryn Westbrook, for her
time and assistance throughout the preparation of my thesis.
 TABLE OF CONTENTS

Chapter Page

1. INTRODUCTION AND THEORETICAL ANALYSIS 1

1.1 Introduction 1

1.2 Scope of the Study 3

1.3 Governing Equations and Scaling Analysis 3

2. LITERATURE REVIEW 10
2.1 Single Cylinder without confining walls(s) 10
2.2 Array of cylinders without confining wall(s) 11
2.3 Single cylinder confined by two parallel
walls 14
2.4 Array of cylinders confined by two parallel
walls 18
2.5 Cylinder(s) confined by a single wall 21

3. EXPERIMENTAL APPARATUS AND PROCEDURES 27

3.1 Introduction 27
3.2 Components 29
3.2.1 Heaters 29
3.2.2 Thermocouples 29
3.2.3 Cylinders 31
3.2.4 Wall(s) 37
Main Frame and Enclosure
3.2.5 40
Electrical Systems
3.2.6 42
3.3 Flow Visualization 46
3.4 Procedures 50
3.4.1 Data Collection 50
3.4.2 Data Reduction 53
3.4.3 Uncertainty 56
4. SINGLE WALL: RESULTS AND DISCUSSION 57
4.1 Free Single Cylinder 57
4.2 Temperature distribution along the array
with various wall spacings: 60
4.3 Heat Transfer Coefficient Results 68
4.3.1 The effect of the wall spacing on the
lowest cylinder of the array 68
4.3.2 The Effect of the wall spacing on the
heat transfer from the cylinders of the
array 76
4.4 Data Correlation 105

5. TWO WALLS: RESULTS AND DISCUSSION 112

5.1 The effects of the right wall spacing, (S/D)R,
on the temperature distribution along the
array 112
5.2 The effects of the right wall spacing, (S/D)R,
on the heat transfer from each cylinder in
the array 118
5.3 The effects of right wall spacing on the
average Nusselt number of the whole array,
Nua,,,Rw 147
5.4 Data Correlation 156

6. NUMERICAL ANALYSIS AND FLOW

VISUALIZATION 169
6.1 Single Wall Cases 170
6.2 Two Wall Cases 177

7. CONCLUSIONS AND RECOMMENDATIONS 183

7.1 Conclusions 183
7.1.1 Single wall cases 184
7.1.2 Two wall cases 185
7.2 Recommendations 187
BIBLIOGRAPHY 18 8

APPENDICES
A DATA ACOUISITION PROGRAM 19 2
B RADIATION CORRECTION 19 6
B.1 Radiation correction from a free single
cylinder and an array of cylinders without a
wall 19 6
B.2 Radiation correction for an array with a
single wall 19 8
B.3 Radiation heat transfer for two wall cases 205
C DATA REDUCTION PROGRAM 213
D UNCERTAINTY ANALYSIS 2 21
 LIST OF FIGURES

Figure Page
1.1 System of coordinates. 5

3.1 Pictorial view of the testing section. 28

3.2 Test section assembly for the cylinders. 32

3.3 Thermocouple disk. 34

3.4 End-Caps. 36

3.5 End-Block. 38

3.6 Side wall. 38

3.7 The main frame. 41

3.8 The enclosure. 41

3.9 Power supply system. 43

3.10 Temperature monitoring system. 44

3.11 Laser illumination system. 47

3.12 Smoke generation system. 49

4.1 Nusselt number, Nui, versus Ra* for a free

single cylinder. 59

4.2 The effect of wall spacing on the normalized

temperature at q= 49.338W/m2. 61

4.3 The effect of wall spacing on the normalized

temperature at q= 149.014 W/m2. 62
4.4 The effect of wall spacing on the normalized
temperature at q= 493.380 W/m2. 63

4.5 The effect of wall spacing on the normalized

temperature at q= 986.762 W/m2. 64

4.6 The effect of wall spacing on the normalized

temperature at q= 1480.143 W/m2. 65

4.7 The effect of cylinder's position in the array on

the normalized temperature , 0, at S/D=
infinity (no wall condition) . 67
4.8 The average Nusselt number of the lowest
cylinder, Nui , versus Ra*, at CC=1.5D. 69

4.9 The average Nusselt number of the lowest

cylinder, Nui , versus Ra*, at CC=2D. 70

4.10 The average Nusselt number of the lowest

cylinder, Nui , versus Ra*, at CC=4D. 71

4.11 The average Nusselt number for the array's

cylinders without a wall, Nlli,f , Vs. Ra* at
CC=1.5D . 73
4.12 The average Nusselt number for the array's
cylinders without a wall, Nlli,f , Vs. Ra* at
CC=2D 74
4.13 The average Nusselt number for the array's
cylinders without a wall, Nui,f , Vs. Ra* at
CC 4D. 75
4.14 The effect of the wall spacing on the average
Nusselt number of each cylinder, Nui , at
CC=1.5D . 77
4.33 The effect of the wall spacing on the Nusselt
number ratio, Nui/Nui,f , at CC=4D and
q=986.762 (W/m2). 98

4.34 The effect of the wall spacing on the Nusselt

number ratio, Nui/Nui,f , at CC=4D and
q=1480.143 (W/m2). 99

4.35 The effect of the wall spacing on the average

Nusselt number of the whole array at CC=1.5D .
102
4.36 The effect of the wall spacing on the average
Nusselt number of the whole array at CC=2D. 103

4.37 The effect of the wall spacing on the average

Nusselt number of the whole array at CC=4D. 104
4.38 The measurement of Yi dimention as used in
equation 4.7. 106
4.39 The experimental Nui values and correlated
Nui values Vs. Ra* at S/D=0.5 and CC=1.5D . 10 8

4.40 The experimental Nui values and correlated

Nui values Vs. Ra* at S/D=0.5 and CC= 2D . 10 9

4.41 The experimental Nui values and correlated

Nui values Vs. Ra* at S/D=0.5 and CC=4D . 1 10

5.1 The effect of right wall spacing on the normalized

temperature at q= 49.338 W/m2. 113

5.2 The effect of right wall spacing on the normalized

temperature at q= 149.014 W/m2. 114

5.3 The effect of right wall spacing on the normalized

temperature at q= 493.380 W/m2. 115
5.28 The effect of the wall spacing on the Nusselt
number ratio, Nui,Rw/Nui,f , at CC=4D. 144

5.29 The effect of the wall spacing on the Nusselt

number ratio, Nui,Rw/NuLf , at CC=4D. 14 5

5.30 The effect of the wall spacing on the average

Nusselt number of the whole array at
CC=1.5D. 14 8

5.31 The effect of the wall spacing on the average

Nusselt number of the whole array at CC=2D. 14 9

5.32 The effect of the wall spacing on the average

Nusselt number of the whole array at CC=4D. 15 0

5.33 The effect of the wall spacing on the Nusselt

number ratio, Nui,Rw/Nui,f , at CC=1.5D. 15 2

5.34 The effect of the wall spacing on the Nusselt

number ratio, Nui,Rw/Nui,f , at CC=2D. 15 3

5.35 The effect of the wall spacing on the Nusselt

number ratio, Nui,Rw/Nui,f , at CC=4D. 15 4

5.36 The effect of the wall spacing on the Nusselt

number ratio, Nui,Rw/Nui3O.5 , at CC=1.5D. 15 7

5.37 The effect of the wall spacing on the Nusselt

number ratio, Nui,Rw/Nui3O.5 , at CC=2D. 15 8

5.38 The effect of the wall spacing on the Nusselt

number ratio, Nui,Rw/Nui3O.5 , at CC=4D. 15 9

5.39-a The experimental Nui,Rw values and the

correlated Nui,Rw values Vs. Ra* at CC=1.5D . 161
 -b The experimental Nui,Rw values and the
correlated Nui,Rw values Vs. Ra* at CC=1.5D. 162

5.40-a The experimental Nui,Rw values and the

correlated Nui,Rw values Vs. Ra* at CC= 2D . 163

-b The experimental Nui,Rw values and the

correlated Nui,Rw values Vs. Ra* at CC= 2D. 164

5.41-a The experimental Nui,Rw values and the

correlated Nui,Rw values Vs. Ra* at CC=4D . 165

-b The experimental Nui,Rw values and the

correlated Nui,Rw values Vs. Ra* at CC=4D . 166

6.1 The mesh model for TEMPEST for CC=2D array. 171

6.2 Temperature, Velocity, and Flow fields for

CC=1.5D array with a single wall at S/D=0.5 . 172
6.3 Temperature, Velocity, and Flow fields for
CC=2D with a single wall at S/D=0.5 . 173

6.4 Temperature, Velocity, and Flow fields for

CC=4D with a single wall at S/D=0.5 . 174

6.5 Flow visualization for arrays symmetrically

placed between two walls at S/D=(S/D)R=0.5 178

6.6 Temperature, and Velocity fields for CC=1.5D

array with two walls at S/D=0.5 and
(S/D)R=2.0 . 180

6.7 Temperature, Velocity, and Flow fields for

CC=2D array with two walls at S/D=0.5 and
(S/D)R=2.0 . 181
 LIST OF TABLES

Table Page

3.1 Spacings of the cylinders and of the walls for

the experiments. 51

4.1 Coefficients of equation 4.7, Nui/Rai*°.2 = Al +

A2 Exp[ -A3 (Yi/CC)]. 107

4.2 Correlation coefficient, B1, for equation 4.8, Nuav

= B1 Rai*0.2 111

5.1 The correlation coefficient, B1 , in equation 5.4. 160

5.2 The Correlation coefficient, B1, for Ntlav,Rw=

B1 Ra1,Rw*0.2 168
 LIST OF APPENDIX FIGURES

Figure Page,

B.1 The wall numbers for the view factors. 19 9

B.2 Crossed-String method to calculate F3-4. 201

B.3 The dimensions of equation B.5 . 203
B.4 The numbers of the walls for two walls cases. 208
 LIST OF APPENDIX TABLES

Table Page

B.1 View factors for a single wall case at CC= 2D and

S/D= 0.5 204

B.2 Radiation corrections for the array at CC= 2D and

S/D=0.5 206

B.3 View factors for two walls case at CC= 2D with

S/D=0.5 and (S/D)R= 2.0 . 210

B.4 Radiation corrections for the array at CC= 2D and

two walls at S/D= 0.5 and (S/D)R =2.0 . 212
 NOMENCLATURE

A = Surface area of the cylinder , D L), m2 .

Ac = End cap cross section area, m2

CC = Cylinder center-to-center spacing, m.
Cp = Specific heat, W Hr/(Kg K).
D = Cylinder diameter, m.
e = Emissivity.
Gr* = Modified Grashof number, equation 3.3 .

g = Acceleration of gravity, m/s2 .

h = Average heat transfer coefficient, W/m2 K.
I = Electrical current, amp.
K = Thermal conductivity, W/m K.
L = Cylinder length.
Nu = Average Nusselt number of the cylinder,=((h
D)/K)
Nuav = Average Nusselt number of the whole array for
single wall cases, equation 4.4 .

Nu avif = Nusselt number ratio for single wall cases,

equation 4.5 .

Nilav,RW = Average Nusselt number of the whole array for

two walls cases, equation 5.2 .
Nu av,RW/f = Nusselt number ratio for two walls cases,
equation 5.3 .

Nu ay/a = Nusselt number ratio for single wall cases,

equation 4.6.
Nui = Average Nusselt number for cylinder #i .

Nu i, f = Average Nusselt number for cylinder #i in a free

array .

Nui,Rw = Average Nusselt number for cylinder #i for two

walls cases.
Nus = Average Nusselt number for a free single
cylinder.
p = Pressure, N/m2 .
P = Electrical power, W.
Pr = Prandtl number.
Q = Total power input, W.
Qcv = Heat flux by convection, W.
Qcd = Heat flux by conduction, W.
Qr = Heat flux by radiation, W.
q = Heat flux per unit area, W/m2 .
R = Electrical resistance, ohm .
R a* = Modified Rayleigh number, equation 3.4 .

Rai* = Modified Rayleigh number for cylinder #1 for

single wall cases.
Rai; = Modified Rayleigh number for cylinder #i for free
array cases.
RaiRw* = Modified Rayleigh number for cylinder #i for two
walls cases.
R1 = One ohm electrical resistance.
S = Spacing between the array and the wall, m .
Sb = Stefan-Boltzmann constant = 5.6696 x 10-8
wit-112.1(4

S/D = Array to left wall spacing for single wall cases.

(S/D)R = Array to right wall spacing for two walls caces.
T = Temperature, °C .
Tw = Surface temperature of the cylinder, °C.
Tw,i = Surface temperature of cylinder #i , °C.
Tinf = Ambient temperature, °C.
U = Fluid velocity in Y-direction, m/s.
V = Voltage, Volt.
v = Fluid velocity in Y- direction, m/s.
X = X- coordinate, m.
Y = Y- Coordinate, m.
Yi = Distance from the center of the lowest cylinder to
the center of cylinder #i .

a = Themal diffusivity, m2/Hr .

= Coefficient of thermal expansion, K-1.
8T = Thermal boundary layer thickness, m.
11 = Dynamic viscosity, Kg/m Hr.
= Kinematic viscosity, m2/Hr.
P = Density, Kg/ m3.
= Angular location measured from 6 o'clock
position, radians.
0 = Normalized temperature, equation 4.3.
 An Experimental Study of Natural Convection Heat
Transfer From a Horizontal Cylinder Array Vertically
Aligned to and Confined by a Single Wall or Two Walls

CHAPTER 1

INTRODUCTION AND THEORETICAL ANALYSIS

1.1 Introduction

Natural convection processes occur both in natural and

technological applications when the fluid flow is driven by
buoyancy forces resulting from inhomogeneities in fluid density.
These gravitational forces arise when a body is placed in an
otherwise motionless medium having a higher or a lower
temperature than that of the body.
Establishment of heat transfer by natural convection takes
place in three stages [1, 2]. During these stages, the temperature
and velocity changes are assumed to be confined to a small
boundary layer next to the body's surface. In the first stage, the
heat is transferred by pure conduction from the body to the
adjacent fluid particles. During the second stage, the heat is
transferred by a combination of pure conduction and convection.
Finally, in the third stage, the velocity and temperature profiles
are independent of time and the steady convection flow is fully
established. Parsons and Arey [2, 3] show Mach-Zehnder
 2

photographs of these stages for natural convection heat transfer

for a single horizontal wire and for two horizontal wires vertically
aligned.
Although natural convection has a lower heat transfer
coefficient than forced convection, heat transfer engineering
designers prefer the natural convection mode because it is more
reliable due to the elimination of the cooling/heating fluid
circulation parts [ 4 ]. The convection heat transfer coefficient is
a function of the fluid flow, the thermal properties of the fluid
medium, and the geometry of the system. Many industrial heat
components have horizontal cylinders stacked vertically near a
wall or positioned asymmetrically between two walls. These
components are found in electronic equipment (i.e., computers
and power supply equipment), heat exchangers, nuclear power
equipment, and power transmission cables which are aligned next
to a wall.
When horizontal cylinders are stacked vertically near a wall,
the heat transfer from the lower cylinder behaves much like free
natural convection from a single cylinder. Due to the plume
rising from the lower cylinder and the boundary layer buildup by
the wall, the flow around the upper cylinders is no longer a free
convection flow. For a steady state case, the temperatures of the
cylinders above the lower cylinder are a result of the balance
between the surrounding fluid temperature increases (due to the
lower cylinder(s) ) and the velocity of fluid which is induced by
 3

the raised plume [5].

The exact solution for natural convection for such an
arrangement is complicated by complex geometries and by the
nonlinear nature of the problem. The nonlinearities arise from
the fluid acceleration terms and the coupling of the momentum
and energy equations. Due to the wake effect(s), the cylinder
spacing is a critical and not fully understood variable [3].
Consequently, cylinder spacing has begun to receive more
attention [3]. The lack of quantitative information concerning
cylinder and wall spacing and the need to have more
understanding of the flow around the cylinders has prompted
this study.

1.2 Scope of the Study

The objective of this study is to investigate heat transfer by

natural convection from three horizontal cylinders, at constant
heat flux, aligned vertically parallel to a vertical wall(s). During
the course of the experiment, the effect of the wall spacing on
natural convection from the cylinders was studied. The effect of
cylinder-to-cylinder spacing on natural convection was also
investigated. Increased understanding of the fluid flow around
the cylinders was accomplished by videotaping and taking
photographs of the flow field. Laser sheets were used to
illuminate smoke particles in the testing section. The
 4

temperature and velocity fields were studied (for some cases) by

using Tempest (computer software from Battelle Laboratories) on
the Floating Point system super computer at O.S.U.
Pilot experiments were conducted with a single isolated
cylinder in order to compare readings with the results of
previous studies where the average Nusselt number from a
horizontal cylinder at constant heat flux was expressed as a
function of modified Rayleigh number, Ra*. The results of the
pilot experiments were also compared to the scale analysis which
will be shown in the next section of this chapter.

1.3 Governing Equations and Scale Analysis

In the cases of steady state, two-dimensional flow and

laminar free convection, the boundary layer theory concepts
were employed. These concepts are governed by the
conservation of mass, momentum, and energy equations. With
the assumption of constant fluid properties, except in the case of
the density in the buoyancy force term, the governing differential-
equations appear [6] as follows:

au av
+ =0
DX DY

aU au a2u a2u
(1.2)
u -arc + g {3 Tind sin + +
ax 2 aY2 /
 5

av av a2v a2v
(1.2b)
U-57 + V = g Tind cos A +
ax2 ay2

aT ,,aT a2T a2T

(1.3)
U
+ a 2 4- ay2

The system of coordinates is shown in figure 1.1 .

y is of
Following the scale analysis rules as outlined by Bejan [7],
the same order of magnitude as the thermal boundary layer,ar
This will be written as Y ST. Since x=f (D, 8),
X D
ar2 « D2
2
-ay2
a >> a2
axe

Gravity field

2X
e=
D

Figure 1.1 System of coordinates.

 6

a2
Thus the terms in the momentum and energy
axe
equations (equations 1.2a, 1.2b and 1.3, respectively) are
neglected. The scaling of the continuity equation, equation 1.1,
gives
V U 8T (1.4)
75-

and the scaling of the energy equation gives

AT AT
AT
V a
D ST

convection I I conduction

By substituting for V from equation 1.4 in the last equation, both

convection terms are of the order U AT Therefore,

U a D (1.5)
T

The momentum equations, equation (1.2a) and (1.2b), give the

following scale balance:

iJ 2 , V or g 13 AT sin 8
ST

I_ inertia I I_ friction _I I_ buoyancy _I

UVor ,
V2
8T
V
g AT cos 0

I_ inertia I I_ friction I I_ buoyancy _I

 7

Hence the ST is ruled by an Inertia Buoyancy balance or

by a Friction Buoyancy balance. For Pr greater than one, the
thermal boundary layer thickness, ar , is much smaller than the
velocity boundary layer thickness and ST is ruled by the Friction
Buoyancy balance.
By adding the square of the friction terms and the square of
the bouyancy terms in the above equations, the following
relationship is obtained:

2 U2 V 2 132 AT2 (sin2 8 + cos2 8 )

(7-
8T
+
8T
g

Equation (1.4) implies that V2 << U2 . Therefore, V2 can be

neglected. By substituting for (sin2 0 + cos2 = 1 and rearranging
the terms, the above relationship can be written as,
vU
1
g 13 AT 4

By substituting for (U) from equation 1.5,

v aD
1
gI3AT4 (1.6)

The no-slip condition implies that the heat transfer adjacent to

the wall at 0< Y <0 is by pure conduction and governed by
Fourier's law [8].
 8

q=K aT (1.7)
aY 1(D

AT
Therefore, q K gives AT q 81.
oT
K
Sustituting for this value of AT and for a = p Cp and

E. equation (1.6) becomes

p. K2 D5
2 1
p gi3CpqD4 5
8.r

*-1/5
Thus ST Ra D (1.8)

where Ra* is modified Rayliegh number. The wall-heat flux

equation, q=h(T-Tinf) = h AT , and equation (1.7) give

AT
q-K oT
h AT

K
hence h
ST

Using equation (1.8) and rearranging the terms

*1/5
Nu = hD Ra (1.9)
K
Therefore, the Nusselt number for free natural convection heat
transfer from a single isolated horizontal cylinder is a function of
the modified Rayleigh number to the 1/5th power. Further, the
 9

modified Rayleigh number, shown by equation 1.8, is of the same

order of magnitude as the ratio of D to the thermal boundary
layer thickness, 8.
 10

CHAPTER ER 2

LITERATURE REVIEW

2.1 Single Cylinder without confining walls(s)

Numerous studies on natural convection heat transfer from a

horizontal cylinder have been published. Morgan [9] tabulated
sixty-four references which presented the natural convection
heat transfer results from an isolated cylinder or array of
horizontal heated cylinders (wires). The ratios of cylinder length
to diameter in these studies ranged from 2.8 to 9000. The
Rayleigh number, Ra, based on the cylinder diameter and on the
film temperature, T f= (T w + T inf )/2, ranged from 7 x 10 to
10 . Only one of these studies: Dyer [10], was conducted' at
uniform heat flux. Based on the data from these references and
on their correlation equations, Morgan [9] proposed the
correlation for natural convection from a horizontal cylinder as
M
Nu = B (Ra)
where B and M are constants depending on the range of Ra. Ra
is based on the cylinder diameter and the fluid properties at the
film temperature, Tf
In a theoretical and experimental study, Kim, Pontikes, and
Wollersheim [11] studied the local and the average Nusselt
number, Nu, from a horizontal cylinder to a Newtonian fluid
 11

(mineral oil). They obtained the following relations for the

average Nusselt number:

.19
Isothermal cylinder, Nu i = 0.89 (Gri . Pr)

.2
Constant heat flux, Nu = 0.57 (Gre. Pr)
where Gr is the Grashof number and Pr is the Prandtl number.
The subscripts i and c mean Gr based on temperature difference
or uniform heat flux, respectively.

2.2 Array of cylinders without confining wall(s):

Lieberman and Gebhart [12] investigated the heat transfer

from an array of heated wires. The wires had a length to
diameter ratio equal to 1450. (The Rayleigh number was of order
10 and the array took different angle positions. The results were
inferred from interferograms of the temperature field and the
rising plume. They found that the temperatures of the wires
increased up the array when the spacing between the wires was
37.5 diameter (the closest spacing). But, for higher spacing (75
diameter), the temperatures decreased as the position of the wire
ascended the array. This was due to the increase of the air
velocity in the plume, which resulted in lowering the plume
temperature while the plume rose up in the array.
Marsters [13] studied the natural convection for vertical
 12

arrays of heated horizontal cylinders. In his work the effect of

cylinder-to-cylinder spacing was experimentally studied for
three, five, and nine-cylinder arrays. The spacing took values
from 2 to 20 cylinder diameters. Air was the working medium
and the Grashof number, based on cylinder diameter and
temperature difference, ranged from 750 to 2000. He found that
the Nusselt numbers for narrow spacing are smaller than the
Nusselt numbers for a single cylinder (about 50% smaller), while
for wide spacing, the Nusselt numbers are about 30% higher than
for a single cylinder. The temperature distribution along the
array agreed with the results of Lieberman and Gebhart [12].
Furthermore, the lower cylinder in any array behaves much like
a single cylinder. Marsters suggested that the Grashof number
should be based on the characteristic length of the distance from
the lower cylinder rather than on the cylinder diameter to
explain the temperature distribution along an array.
The effect of vertical separation distance and cylinder-to-
cylinder temperature imbalance on natural convection for a pair
of horizontal cylinders was examined by Sparrow and
Niethammer [5]. In these experiments, the lower cylinder wall-
to-ambient temperature difference ranged from zero to three
times the upper cylinder wall-to-ambient temperature
difference. The values of Rayleigh numbers, based on the
cylinder diameter, for the upper cylinder were from 20,000 to
200,000.
 13

The results were presented as the Nusselt number ratios and

the temperature ratios. The Nusselt number ratio was the
Nusselt number of the upper cylinder divided by the Nusselt
number of an isolated cylinder of the same Rayleigh number,
while the temperature ratio was the ratio of the wall-to-ambient
temperature difference of the upper cylinder divided by that of
the lower cylinder. At a specified temperature ratio and Rayleigh
number, the Nusselt number ratio increases sharply at a small
separation distance, S. The slope of the Nusselt number ratio
becomes flatter at larger spacing and the ratio takes a maximum
value in the range of S/D between seven and nine. At low
spacing ratio the increase of the temperature ratio degrades the
Nusselt number ratio.
Sparrow and Boessneck [14] studied the effects of transverse
misalignment on natural convection from a pair of parallel,
vertically stacked, horizontal cylinders. At several fixed
separation distances (cylinder-to-cylinder distance) the
transverse offset was varied. The transverse misalignment had
eight values ranging from zero to three cylinder diameters. The
separation distance took several values between 2 and 9 cylinder
diameters. The cylinder diameter was 1.49 inches and its
length/diameter ratio was 20. Rayleigh numbers were based on
the cylinder diameter and the cylinder-to-ambient temperature
difference. During the course of the experiment, Rayleigh
numbers ranged from 20,000 to 200,000 and the testing fluid
 14

was air.
Since the lower cylinder is not affected by misalignment (i.e.,
it behaves like a single cylinder), the study focused on the upper
cylinder Nusselt number ratio. This ratio was the misaligned
upper cylinder Nusselt number to the perfectly aligned upper
cylinder. The offsetting enhanced the Nusselt number ratio (up
to 27%) at small cylinder-to-cylinder vertical separation. While at
large separation distances the offsetting decreased the Nusselt
number ratio (up to 22%). For specific offset, the Nusselt number
enhanced at lower Rayleigh number. This enhancement
decreased as the vertical separation increased. In all the cases
the Nusselt number reached a single value, which was about 3%
higher than that of a single cylinder when the offset was large.
The results, when all the cylinders were aligned, agree with the
results of [5], where the Nusselt number of the upper cylinder
was enhanced as the vertical separation between the cylinders
was increased.

2.3 Single cylinder confined by two parallel walls:

Marsters [15] studied the effects of adiabatic walls

confinement on the heat transfer from a horizontal heated
cylinder. During the course of the experiments, Rayleigh number
values ranged from 10 to 500,000. This range was achieved by
using air, water and freon 113 as working fluids and two
 15

different cylinder diameters to give cylinder length-to-diameter

ratios of 70 and 67.9. The wall spacing was varied between 2 to
20 cylinder diameters and the wall height-to-diameter ratio had
values between 5 and 128. The study consisted of two parts.
First, the cylinder was tested with no-wall (isolated cylinder) and
a correlation equation for the average Nusselt number, Nu, was
presented

(0.14+0.013 log Ra)

Nu = 0.88 Ra for 10 < Ra < 500,000
where Ra was based on the cylinder diameter and the
temperature difference. In the second part, the cylinder was
confined by the walls. The results showed that the heat transfer
characteristics were enhanced significantly by the presence of the
walls. Even at large wall spacing (20 cylinder diameters), there
was a 30% increase in the heat transfer over the no-wall case. On
the other hand, Marsters found that there was no effect of the
cylinder elevation (the distance from the bottom of the walls to
the center of the cylinder) as long as the cylinder lay between the
walls. Analysis of the data gave the following correlation
equation for the Nusselt number:
a b
Nu = 0.82 Ra . 10
where a = 0.17 (1 + 0.05 Log Ra)
b = 0.02 (11/w) 0-0.018 (H/t) - 9.2 (d/11))
H = height of the walls
w = spacing between the walls.
 16

This equation indicates that the higher order of the geometry

terms was unimportant. Also, it shows Nu decreased when
H/w>28 for fixed values of w, D and Ra.
Sparrow and Pfeil [16] determined the heat transfer
characteristics of a heated horizontal cylinder situated at the
mid-height of a vertical channel with adiabatic channel walls.
The cylinder length-to-diameter ratio was 20/1. The experiment
was carried out in air with Rayleigh number, Ra (based on
cylinder diameter and temperature difference), between 15,000
and 200,000. The effects of the channel height, the interwall
spacing (the distance between the walls), and the wall materials
on the heat transfer were studied. It was found that the Nusselt
number, relative to the no-wall condition, was enhanced as the
channel height increased. Thus, the vertical placement of the
cylinder in the channel enhances the heat transfer. This
enhancement can reach up to 40% at small interwall spacing,
while lower at large interwall spacing.
Sparrow and Pfeil's [16] conclusion does not agree with that
of Marsters [15], where the vertical placement of the cylinder
does not affect the heat transfer characteristics. However, both
studies agree that the Nusselt number was enhanced at the small
interwall spacing. This enhancement decreases as the spacing
increases. In Sparrow and Pfeil's [16] experiment, the spacing
effect decreased as the height of the channel increased. They
used three different shroud types (the walls parallel to the
 17

cylinder axis) including: highly conducting, highly conducting

with insulation on the back side, and non-conducting shrouds.
The results revealed that the Nusselt number was insensitive to
the various types of shrouds. They presented a correlation
equation for the Nusselt number in the following form,
n
Nu = C Ra
where C and n are constants depending on the ratios of the
channel height and interwall spacing to the cylinder diameter.
Karim et al. [17] investigated the effects of the interwall
spacing and the vertical placement on the natural convection heat
transfer from a horizontal isothermal cylinder symmetrically
placed between two parallel adiabatic walls. Two different
cylinder diameters were used with length-to-diameter ratios of
17.4 and 25.98. Average Nusselt numbers were determined for a
Rayleigh number of 2,000 to 300,000. The results showed that
the Nusselt number degraded with increasing Rayleigh number
for all the wall spacing ratios (wall spacing to cylinder diameter).
Their experiments agree with [15] , where the confinement
enhances the Nusselt number and there is no significant effect
from the cylinder's vertical placement on the heat transfer
characteristics. But, their findings disagree with [15] when there
is an optimal spacing for maximum heat transfer. A correlated
equation to calculate the average Nusselt number was presented
by Karim et. al. as
.25
Nu = (0.481 + 0.172 exp (-0.258 (W/D))) . Ra
 18

where W is the spacing between the walls and Ra is the Rayleigh

number based on cylinder diameter, D, and on cylinder-ambient
temperature difference.

2.4 Array of cylinders confined by two parallel walls:

Marsters and Paulus [18] examined the effects of confining

walls on a vertical array of horizontal cylinders. The array
consisted of nine heated cylinders with cylinder's length-to-
diameter ratio equal to 70 and with cylinder-to-cylinder spacing
equal to 6 cylinder diameters. They found that the walls
influenced the heat transfer characteristics of an individual
cylinder in the array, but the walls had less effect on the overall
heat transfer of the array. They presented curves for normalized
temperature (the cylinder-to-ambient temperature difference to
the bottom cylinder-to-ambient temperature difference). The
curves illustrate that the normalized temperature increased as
the cylinder was elevated in the array and as the wall spacing
decreased. The experiments were carried out with/without
ventilating walls (the side walls perpendicular to the cylinder
axes). They also examined the normalized temperature for a
single wall spaced two cylinder diameters distance from the
center of the array. Again, in this case, the temperature
increased as the cylinders were elevated in the array.
Natural convection characteristics for a vertical array of
 19

heated cylinders with/without confining walls were

experimentally investigated by Tokura et. al. [17]. The
experiments were carried out for 2, 3, and 5-cylinder arrays with
cylinder lengths equal to 20.8 cylinder diameters. For non-
confining arrays, their results agreed with [5 and 13] that the
average heat transfer coefficients increase as the center-to-center
distances increase. This enhancement reaches a maximum value
when the spacings are approximately five times the cylinder's
diameter. The following equations were predicted:
(1) The average Nusselt number for the second cylinder
in an array as a function of Rayleigh number, Ra, and
distance from the center of the bottom cylinder to the
center of the second cylinder, CC, and the cylinder
diameter, D.

3/4 1/4
Nu = 0.26 (CC/D)3/4 [1-exp{-2.22/((CC/D) ) -1))] Ra

This equation produced close results to those obtained in

the case of a single cylinder when CC extends to infinity.
The above equation can be used for the other cylinders in
the down stream of the array when the spacing between
the cylinders is large.
(2) The average Nusselt number for a whole array as a
function of Ra and b/D where b is the diameter-to-
diameter cylinder separation distance and N is the
 20

number of the cylinders in the array.

1/4 .055N)
Nu = 0.41 Ra Ln ([(b/D)/1.3] +0.434

.055N=
The error in this equation is ± 10% for [(b/d)/1.3] 0.7
to 1.2 and Gr = 40,000 to 400,000.
For an array between two parallel plates, they found that
the Nusselt number for a whole array was enhanced. This
enhancement reached its maximum when the spacing between
the plates was three times the cylinder diameter, S/D = 3. When
S/D = 30, the effects of parallel plates on the average Nusselt
number was insignificant. The average Nusselt number for each
cylinder was almost the same for all the cylinders above the
bottom one when the spacing between the cylinders was 6D (i.e.,
b/D=6). But, when b=D, the average Nusselt number for the
cylinder above the bottom one was much smaller than the
average Nusselt number for the bottom cylinder. This difference
decreases as the S/D increases. In general, the effect of the
spacing between the cylinders is greater than the effect of the
distance between the plates. Relative to an array without
confining walls, the effect of the walls increased the heat transfer
by 10 to 15% when the separation distance between the walls
was 2 to 6 times the cylinder's diameter.
 21

2.5 Cylinder(s) confined by a single wall:

Five references are cited under this category. Four of them

were published with Sparrow as a co-author. The articles were
published between 1981 and 1987. Three of these articles, as
shown below, investigate the natural convection heat transfer
from a horizontal cylinder(s) fixed perpendicularly on a vertical
heated wall.
The natural convection heat transfer characteristics from an
isothermal horizontal cylinder attached to an isothermal vertical
plate at the same temperature were investigated experimentally
by Sparrow and Chrysler [20]. During the course of the
experiments, the cylinder axes were perpendicular to the plate
surface. The effect of the cylinder position from the leading edge
was studied by attaching the cylinder to one of the three
positions along the height of the plate. Two cylinders with
length-to-distance ratio, L/D, equal to 1 and 1/2 were used. The
Rayleigh number, Ra (based on cylinder diameter, D), ranged
from 14,000 to 140,000. Sparrow and Chrysler found that at a
given Ra the Nusselt number was insensitive to the cylinder's
vertical position. In addition, the Nusselt number was enhanced
at the higher position and it was lower at the lower position for a
given Ra. However, at the middle position the Nusselt number
was lower than that at the lowest position. This demonstrated
nonmonotonic variation of the Nusselt number with the elevation
 22

of the cylinder. The presence of the vertical plate degraded the

Nusselt number due to the boundary layer build-up by the plate.
This effect was higher on the shorter cylinder than on the longer
cylinder due to the plate-cylinder interaction. From the data
obtained in these experiments, the following empirical equation
was presented to find the average Nusselt number as a function
of Ra:

1/4
Nu = C Ra for 14,000< Ra < 140,000.
where C depends on the cylinder's length-diameter ratio and the
cylinder position from the lower edge of the plate.
The above work, [20], was extended by Sparrow et. al. [21]
to investigate the natural convection heat transfer for a vertical
array of horizontal cylinders perpendicular to a vertical plate.
Both the cylinders and the plate were at the same constant
temperature. Two arrays comprised of two and three cylinders
were used. Using a two-cylinder array allowed examination of
the effects of wider cylinder-to-cylinder distance and two array
elevations (distance from the leading edge of the plate to the
lower cylinder). Four parameters were studied. These include
the fin (cylinder) length-to-diameter ratio, the interfin spacing,
the position at which the fin is attached to the host vertical plate,
and the Rayleigh number, Ra.
For an individual cylinder in the array, the results showed
that the presence of a cylinder (or cylinders) below it lower the
 23

Nusselt number(s) at higher Rayleigh number. The lowest

cylinder of the array was not affected by the presence of the
cylinder(s) above it. For the two-cylinder array, the presence of
the array at the lower part of the plate gave the upper cylinder a
higher Nusselt number than the array at the higher part of the
plate. Higher interfin spacing also produced higher Nusselt
numbers for the upper cylinder. For the three-cylinder array,
the Nusselt number degraded as the cylinder elevation increased
at low Rayleigh number. While at higher Ra, the cylinder in the
middle obtained the highest Nusselt number.
Sparrow et. al. [22] extended the investigation of natural
convection heat transfer from an isothermal horizontal cylinder
attached perpendicularly on a vertical plate at equi-temperature
to include the duct-flow effects. The walls of the duct (except the
vertical plate) were adiabatic. During the experimental work, the
spacing, S, between the isothermal vertical plate and the opposite
wall was changed from S/L = 1.4 to S/L = 5.4, where L = cylinder
length. At large S/L the condition reached external flow mode.
This also occurred in the cases discussed by Sparrow and Chrysler
[20].
The enhancement or the degradation of heat transfer was
compared as a percentage of the external flow mode (large S/L).
The results showed that the heat transfer from the lower cylinder
position was enhanced by 60% to 20% as the S/L changed from
1.43 to 5.4 This enhancement was lower for the middle cylinder
 24

position (about 40 to 50%). In both cases (lower and middle

cylinder positions) the effect of the Rayleigh number was
insignificant. For the upper cylinder position, the results were
sensitive to the Rayleigh number at a small S/L ratio. Conversely,
the results were less sensitive for the Rayleigh number at a
higher S/L ratio. At low S/L the enhancement decreased as the
Rayleigh number increased. At high S/L the enhancement was
zero. These results suggest that the external flow is more
preferable for the upper cylinder position.
The effects of the radiation conditions of the unheated walls
on the heat transfer were examined during these experiments by
using black body walls and reflective walls. The reflective walls
showed lower enhancement of the heat transfer than the black
body walls. This was due to the temperature difference of the
black body surfaces being higher than the temperature difference
of the reflecting surfaces. The temperature difference was
defined as the difference between the wall temperature and the
ambient temperature. In the black body case, the higher
temperature differences helped to induce more flow inside the
duct in addition to the induced flow by the isothermal vertical
plate. This higher velocity enhanced the heat transfer.
The characteristics of natural convection heat transfer from
a heated isothermal horizontal cylinder parallel to an adiabatic
wall(s) were examined by Sparrow and Ansari [23]. The
experimental configurations included a single vertical wall
 25

situated to the side of the cylinder, a horizontal wall beneath the

cylinder, and a corner formed by a vertical and a horizontal wall.
The cylinder length was equal to 20 times its diameter. The ratio
of the distance between the cylinder and .the wall(s), in one of the
specified conditions, to the cylinder diameter, S/D, was changed
systematically from S/D = 1/12 to S/D = 4/3. The results were
presented as the ratio of the dissipated heat transfer from the
cylinder in the presence of the wall(s) to the dissipated heat
transfer with no-wall(s) (isolated cylinder). The experiments
were carried. out in air with Rayleigh numbers from 20,000 to
200,000.
In general, the presence of the wall(s) degraded the cylinder
heat transfer relative to the isolated cylinder heat transfer. In
the case of a vertical wall, there was a 20% reduction for S/D =
1/12. This reduction was negligible when S/D = 1/4 or greater.
For a horizontal wall beneath the cylinder, the reduction in the
heat transfer was 5% greater than the vertical wall case. The
effect of the horizontal wall was eliminated when the spacing
ratio, S/D, equalled 1.33 or greater.
The degradation in heat transfer for the corner case was 40%
at the closest spacing and 20% for 1/4 cylinder diameter spacing
or greater. The experiments also demonstrated that the adiabatic
wall(s) experienced a temperature rise relative to the ambient
temperature due to wall(s)-to-cylinder interaction. The highest
temperature rise reported was approximately 86% of the
 26

cylinder-ambient temperature difference, which occurred at the

closest cylinder-wall spacing of the corner-cylinder interactions.
McCoy [24] studied the convection heat transfer behavior
from an isothermal heated horizontal cylinder parallel to an
isothermal vertical wall. The enhancement or degradation of the
Nusselt number relative to the Nusselt number of an isolated
cylidner, Nu/Nus, was examined with different cylinder positions.
These positions included cylinder-wall spacings and the elevation
of the cylinder from the leading edge of the vertical wall. Water
was used as a working fluid and the Rayleigh number ranged
from 400,000 to 10,000,000. It was found that Nu/Nus ratio
increased slowly as the wall-cylinder distance was decreased.
This enhancement reached its maximum at approximately S/D =
0.2. Then Nu/Nus dropped off sharply for closer spacings. This
was due to the interaction between the cylinder and the wall
boundary layer. This interaction was higher at low Rayleigh
numbers where the boundary layer was thick.
 27

CHAPTER 3

EXPERIMENTAL APPARATUS AND PROCEDURES

3.1 Introduction

The overall design of the experiments was devised in order

to satisfy the following requirements:
1) Two dimensional heat transfer conditions. These were
achieved by specifying proper cylinder dimensions and
cylinder's end conditions.
2) Steady state heat transfer conditions. These were
accomplished by allowing adequate time for the
experimental set-ups and by providing stable
environmental conditions.
3) Accurate data readings. Proper devices and
computerized techniques to supply and collect data were
used to satisfy this condition.
4) Rapid and efficient re-arrangement of the geometrical
experimental set-up (especially wall(s) spacing). Figure
3.1 shows a pictorial view of the testing section.
The above requirements will be discussed in detail in the
following sections, where the experimental components and
instrumentation are discussed. At the end of this chapter the
experimental procedure is presented.
 ACRYLLIC
LEFT WALL RIGHT WALL
INSULATION FOR 2 WALL
CASE ONLY
CYLINDER 3

CYLINDER 2

CYLINDER 1

Figure 3.1 Pictorial view of the testing section.

 29

3.2 Components

3.2.1 Heaters

Etched-foil heaters were used to provide a uniform surface

heat flux and to ease the fabrication process of the testing
cylinders. Two etched-foil heaters were used for each cylinder.
The heaters were manufactured by MINCO Products, Inc. The
heated foils were insulated by 0.2 mm (0.008 inch) glass
reinforced silicone rubber. Their dimensions were 7.62 Cm X
12.7Cm (3 inches x 5 inches )with a 0.5mm (0.02 inch) maximum
thickness. The working temperature range of the insulation was
from -62 degrees Centigrade to 235 degrees Centigrade.
Each heater had 30.48 Cm (12-inch) teflon insulated wires,
size 24 AWG. The unheated ends of the heaters were trimmed to
give a final dimension of 7.42Cm x12.38Cm (2.92 inches x 4.875
inches). This allowed for a better fitting of the heaters inside the
cylinders and reduced the unheated area of the heaters to supply
uniform heat flux to the cylinder surfaces.

3.2.2 Thermocouples

Constantan and copper-nickel (type-T) thermocouples were

employed in the experiments. Type-T was chosen because it
has a low rate of error across the temperature range of the
 30

experiments. This type of thermocouple has a temperature range

from -200 degrees Centigrade to 350 degrees Centigrade with a
limit of error of 1.0 degree Centigrade or 0.75%, whichever of
these is greater [25].
Small thermocouple wires of .254mm (0.01 inch) in
diameter were selected to reduce the conduction heat loss from
the cylinders. The thermocouples were fabricated by using a
Hot-Spot thermocouple welder from DCC corporation and
following the instructions in [26]. After obtaining a spherical,
homogeneous thermocouple bead, the bare lead wires close to the
bead were insulated by thermocouple epoxy to prevent their
contact. Then, the thermocouple resistance was measured and
compared to the resistance of the lead wires to insure a good
electrical contact in the junction (bead). All thermocouples were
calibrated by using boiling water and ice water as reference
temperatures. The calibration process also included measuring
the voltage drop across the thermocouples and comparing it to
the tables in reference [25]. Those thermocouples which passed
the calibration test were selected for use in the experiment. A
total of 26 thermocouples was employed in the experiments. The
location and manner by which the thermocouples were attached
to each component will be discussed in later sections.
 31

3.2.3 Cylinders

Three cylinders comprised the main experimental apparatus.

These cylinders were fabricated to be identical in all respects (i.e.,
dimensions and surface radiation conditons). The test section
assembly for the cylinders is shown in Figure 3.2. The cylinders
were fabricated from aluminum tube with a 2.54 Cm (1.0 inch)
outside diameter and a 2.36Cm (0.93 inch) inside diameter.
Aluminum was chosen as the heat transfer surface because,
when polished to a mirror-like finish, it reduces the radiative
heat loss which competes with natural convection [14]. The high
thermal conductivity of aluminum is another important factor as
aluminum provides a more uniform heat flux surface.
Each cylinder had a length-to-diameter ratio equal to 10,
allowing for suppression of the axial heat transfer effect. The
cylinders were polished to a mirror-like finish by using a fine-
metal polish in order to reduce the radiation effect.
Eight thermocouples were used for each cylinder. Two of
them were used for the end-cap temperature gradient. The other
six thermocouples were used to measure the temperature of the
cylinder surface, Fig. 3.2. Two thermocouples were cemented to
the interior surface of the cylinder by using thermocouple epoxy
and placed 2.54 Cm (1.0 inch) from each cylinder's end, below
the top stagnation point of the cylinder. The epoxy, which was
manufactured by OMEGA Engineering, Inc., uses aluminum
 10.0'
THERMOCOUPLE 1.0'

1.0'

COPPER
CYLINDER
HEATER .125 DIA.
ALUMINUM. STEEL ROD
CYLINDER
THERMOCOUPLE DISK
END-CAP FIBER GLASS

Figure 3.2 Test section assembly for the cylinders.

 33

powder to enhance the thermal conductivity of the joint. The

other four thermocouples were radially positioned at the mid-
length of the cylinder at 90 degree intervals from the top
stagnation point around the inner circumference of each cylinder.
These thermocouples were hosted in a thermocouple disk.
Each disk, Figure 3.3, was made of 0.635 Cm (0.25 inch)
aluminum plate. The disk's outer diameter was 2.286 cm (0.9
inch). This left enough clearance for the heaters to be inserted
between the disk and the cylinder wall. Each disk had two axial
holes, one located at the center and the other located off-center.
The latter was used to pass the lead wires of the front heater and
the front end thermocouple to the back end of the cylinder (the
end through which all the lead wires were passed). The center
hole was used to pass the steel support rod of the end-caps. The
radial thermocouples were hosted in four radial holes, 90 degrees
apart. Each radial hole was made by two drilling stages.
In the first stage, a 5 mm (0.2 inch) diameter with a 6.35 Cm
(0.25 inch) depth hole was drilled. A thermocouple pin and a
3.17 mm (0.125 inch) outside diameter spring were hosted in this
hole. The pin was made of 3.81 mm (0.15 inch) copper tube and
was 4.57 mm (0.18 inch) in length. The pin was used to support
the thermocouple bead, which was threaded through the pin and
cemented to it by the epoxy. The thermocouple beads were also
threaded through the spring, which pressed the pin and the bead
against the cylinder's surface. The thermocouple beads were
 THERMOCOUPLE BEAD
A

COPPER PIN
RADIAL HOLES
AXIAL HOLE
SPRING

.15'
THERMOCOPULE
LEADS

\-J
A
-411w
- .15'
SECTION AA

Figure 3.3 Thermocouple disk.

 35

flattened at the point of contact with the cylinder to insure a

larger contact area. To insure a safe passage for the radial
thermocouple leads, a 1.02 mm (0.04 inch) diameter second-stage
hole was drilled to connect the radial holes to the center hole of
the disk.
The thermocouple disks were placed at the midpoint of each
cylinder. Then, two foil heaters were inserted from each end of
the cylinder. These heaters were backed up by 2.223 mm (0.875
inch) diameter copper cylinders with a length of 12.065 Cm (4.75
inches). The copper cylinders were used to support the heaters
against the aluminum cylinders and to prevent hot spots on the
heating foils due to loss of contact between the heaters and the
cylinders. In order to prevent an internal natural convection in
the cores of the cylinders, the cores were filled with pressed fiber
glass insulation. End-caps were used to seal the cylinders.
Solid delrin and artificial cork were used to fabricate the
end-caps. As shown in figure 3.4, each cap consisted of two disks.
The first disk was made of delrin and hosted two thermocouples,
one on each axial surface. The thermocouples were glued to the
surfaces with thermocouple epoxy. A 3.81mm (0.15 inch)
diameter hole was drilled at the center of each disk to allow
insertion of a 3.175 mm (0.125 inch )diameter steel rod used to
hold the caps at the ends of each cylinder. The rear end-caps had
two extra holes drilled to thread the lead wires of the heaters and
thermocouples through. This left the front ends free of lead
 THERMO-
COUPLES

DELR I N

CORK
SECTION A-A

Figure 3.4 End-Caps.

 37

wires to give clear flow visualization pictures. The second disk of

the end-caps was made of artificial cork and was glued with
silicone rubber onto the external surface of the delrain cap. After
the cylinders were installed on the main supporting frame, the
end-caps of the cylinders were hosted in a block of styrofoam
with three holes. The thickness of the styrofoam blocks was 5.08
Cm (2.0 inches) and each block extended 2.54 Cm (1.0 inch)
beyond the surfaces of the array, Figure 3.5

3.2.4 Wall(s)

Two types of walls were used in the experiment: the side

wall(s), which were parallel to the cylinders and the end walls
(the baffles), which were perpendicular to the cylinders' axes.
The side walls were made from 1.27 Cm (0.5 inch) thick acryllic
and were constructed with a height of 63.5 Cm (25 inches) and a
width of 25.4 Cm (10 inches). They were backed with 2.54 Cm (1
inch) thick styrofoam insulation which was glued on the back
surface of each wall. The surfaces of all the walls facing the
cylinders were painted with a flat black paint in order to achieve
a uniform radiation condition, [18].
Aluminum bars 1.27 Cm x 1.905 Cm x 35.56 Cm (0.5" x 0.75"
x 14") were fixed at the top and the bottom of each wall. The
1.27 Cm x 35.56 Cm (0.5" x 14") surface of the bars lay on the
same plane as the surface of the walls facing the cylinders. The
 38

A
r

STYRO
FOAM
La' DIA.
HOLES

Leg 2- A
SECTION AA
Figure 3.5 End-Block.

ALUMINUM BAR

ACRYLLIC

10'

Figure 3.6 Side wall.

 39

bars extended 5.08 Cm (2.0 inches) from the sides of the walls to
form an "I" shape as shown in figure 3.6. To facilitate the wall-
array spacing adjustments, each extension had a 1.27 Cm (0.5
inch) hole to host a 0.953 Cm (3/8" )diameter adjusting screw.
The adjusting screws were bolted on the main frame above and
below the baffle supports. Each side wall was equipped with
seven thermocouples. The thermocouples were led in from the
back surface of each wall through 1/32" holes and their junctions
lay in the same plane as the front surface of the walls.
In order to approximate the experimental conditions of two-
dimensional cases [23], end walls made of 3.175 mm (1/8") thick
acryllic were used. The end walls assisted in preventing the
transverse inflow of air toward the cylinders. Three sets of end
walls were fabricated to accommodate the three settings (CC =
4D, 2D and 1.5D). The positions of the top cylinder (cylinder #3)
from the top end of the baffles were the same for all the sets.
This left the positions of the lower cylinders (#2 and #1)
dependent on the cylinder-to-cylinder spacings.
The surfaces of the baffles that faced the cylinders were
painted with a flat black paint. There were three 2.54 Cm (one-
inch) diameter holes on each baffle to host the ends of the
cylinders. There was no direct contact between the cylinders'
surfaces and the baffles, since the baffles supported the end-caps
of the cylinders. For each set, two end walls were placed 25.4 Cm
(10 inches) apart on the main frame to form a C-shape channel
 40

for a single wall case and a closed vertical duct for a two-walls
case.

3.2.5 Main Frame and Enclosure

An illustration of the main frame is shown in Figure 3.7.

Angle-aluminum 2.54 Cm x 2.54 Cm (1" x 1") was used to make
the 40.64 Cm x 50.8 Cm x 182.88 Cm (16" x 20" x 6') frame. The
baffle supports were made of 1.905 Cm (3/4" x 3/4") wooden
bars. The lower edges of the baffle supports (which were at the
same level as the lower edges of the side walls) were three feet
above the floor. The aluminum frame was mounted on the frame
base, which was made of 5.08 Cm x 10,16 Cm (2" x 4") wooden
bars. Four screws (1.27 Cm x 12.7 Cm) were used on the four
corners of the base to support the main frame assembly. By
changing the height of the screws below the base, the vertical
plane and the horizontal level of the array could be adjusted.
To reduce the air movement around the testing section, the
main frame was placed in a 1.1 m x .765 m x 2.134 m (3.5' x 2.5'
x 7') enclosure. The enclosure was made of 1.27 Cm (1/2")
plywood sheets. The top and bottom ends of the enclosure were
open, Figure 3.8. To prevent air stratification in the enclosure, a
2.54 Cm (one-inch) gap between the floor and the bottom end of
the enclosure was left. The test section was made accessible
through a 91.44 Cm x 76.2 Cm (3' x 2.5') wooden door on one side
Figure 3.7 The main frame. Figure 3.8 The enclosure.
 42

of the enclosure. The door was fitted with a 40.64 Cm x 22.86 Cm

(16" x 9") acryllic window. While the experiment was running,
the window was covered from the inside with a black surface.
The black window surface was removed during the flow
visualization procedure.
The main frame was positioned in the middle of the enclosure
where its 50.8 Cm x 182.88 Cm (20" x 6') side was parallel to the
plane of the door. The array plane was placed perpendicular to
the 50.8 Cm x 182.88 Cm (20" x 6') side of the main frame. In
order to maintain a uniform radiation condition, the main frame
and the inside surfaces of the enclosure were painted with a flat
black paint.

3.2.6 Electrical Systems

There were two electrical systems used in this experiment,

each having different functions. The purpose of the first system
was to supply power to the heaters in the cylinders and to
measure the power supply for each heater, Figure 3.9. The
second system was used to monitor the output of the
thermocouples, Figure 3.10. A regulated power supply, Model
62-121 manufactured by Dressen-Barnes Corporation, was used
to supply a direct current to heat the cylinders.
Each heater was connected in a series with an ammeter,
rheostate and one ohm resistance. The one ohm resistance was
 al
CYLUWEll

CYUXIMIU el

Ml- 1 ohm realstanoo. ilohootato. Am. Ampcnotor. Bogor

V1., Voltaire drop across Ml. Who Voltage drop ammo H
Figure 3.9 Power supply system.
 THERMOCOUPLE LEADS
FROM THE CYLINDERS
DT707T
SCREW
TERMINAL
PANEL

SCREEN
DT2085
ANALOG
TO PCLAB
DIGITAL
CARDS

HP VECTRA PERSONAL COMPUTER PRINTER

Figure 3.10 Temperature monitoring system.

 45

used to measure the exact current through each heater by

measuring the voltage drop across the resistance. Two rotary
switchs with six channels for each of them, two channels for each
heater, were used. At each position of the rotary switchs, two
voltage terminals were activated. One terminal was used to
measure the voltage drop across the one ohm resistance and the
other terminal was used to measure the voltage drop across the
heater. The voltage drops were measured by a Tektronix DM
5010 programmable digital multimeter and a Tektronix DM 502A
autoranging DMM.
The temperature measurements were accomplished by using
data acquisition boards from Data Translation, Inc., PCLab
software and a Hewlett-Packard Vectra personal computer, as
shown in Figure 3.10. The hardware and software were
calibrated and adjusted according to the specifications in their
manuals [27, 28]. The thermocouple leads were connected to
DT707-T screw terminal with a cold junction compensation circuit
board. When the voltage across the thermocouples were sensed
by the DT707-T's barrier strip, a second cold junction
thermocouple was formed. The cold junction thermocouple is the
same type as the actual thermocouple, but with opposite
electrical polarity, and is referenced to the ambient temperature
of the DT707-T screw terminal panel. This panel has a
thermocouple cold-junction compensation (CJC) circuit to
determine the temperature of the DT707-T [27]. The output
 46

signals from the DT707-T were fed into DT2085 analog-to-digital

cards. These cards were installed in the Hewlett-Packard Vectra
(IBM compatible) personal computer.
A computer program, as shown in Appendix A, was written
to employ the PCLab subroutines. This program converts the
digital voltage output from DT2085 into temperature centigrade.
It prints out the average of sixteen readings for each
thermocouple channel. Then it waits for twenty seconds and
repeats the process.

3.3 Flow Visualization

Flow visualization was accomplished by illuminating smoke

particles with laser sheets perpendicular to the cylinders' axes as
shown in figure 3.11. The patterns of the flow fields were
recorded on a video tape and on slides. Some modifications were
made to accommodate this procedure. A 6.35 mm (1/4") wide
strip of the back insulation along one of the 1.27 Cm (1/2")
acryllic side walls was removed. This strip was extended
vertically at the mid-width of the wall. A strip of the black paint,
on the opposite side of the wall, was removed to provide a clear
strip on the wall. This clear strip was used as an entrance for the
horizontal laser sheet. The source of the horizontal laser sheet
passed through a 6.35 mm x 5.08 Cm (1/4" x 2") slot on the 0.762
m x 2.134 m (2.5' x 7') side of the enclosure. A hole, on the same
 ENCLOSURE
SIDE WALL
WALL

CYLINDRICAL

LASER BEAIE
BEAM LASER
SHEET SPLITTER HEAD

ER #1

0 CYLINDRICAL
LENS

LASER BEAM
BEAM BEAU
REFLECTOR REFLECTOR

Figure 3.11 Laser illumination system.

 48

side of the enclosure, with a 1.27Cm (1/2") diameter was made at

a level 15.24 Cm (6.0 inches) below the bottom edge of the side
walls of the main frame. This hole was used to pass the laser
beam through to generate the vertical laser sheet.
A three-watt Argon and Krypton Ion Laser, Model 85, from
Lexel Corporation, was utilized. A beam splitter, two beam
reflectors and two cylindrical lenses were used to generate two
laser sheets, Figure 3.11. These sheets illuminated one vertical
plane at the mid-width of the array. The vertical plane, except
those sections blocked out by the cylinders themselves, was
illuminated by the horizontal sheet. The horizontal sheet was
created from the horizontal laser beam. Therefore, a vertical
laser sheet created from the vertical laser beam was used to
illuminate an opaque area and to enhance the illumination of the
other parts of the illuminated plane.
Figure 3.12 shows the fabricated smoke generation system
designed to produce the smoke particles. This system consisted
of two chambers. The core chamber was used as a burner, while
the exterior chamber was employed to filter the large particles
from the ashes. Tobacco was used to generate the smoke. First,
the generated smoke passed through a 0.92 m (3') pipe of
aluminum, which was cooled by wet cloth. This reduced the
temperature of the smoke substantially. Then, the smoke was
passed through a thirty-foot, thin-wall teflon hose with a 9.53
mm (3/8") diameter. At the end of this line the smoke was at
 ALUMINUM
PIPE
--> SMOKE
BURNER
OUTLET
0 0
0.
0 was. 0
Q0
004 0
0 30' ENCLOSURE
AIR TEFLON WALL
SUPPLY HOSE

Figure 3.12 Smoke generation system.

 50

room temperature. This system was placed outside the

experimental room, but the teflon hose was passed into the room
and through the wall of the enclosure.

3.4 Procedures

3.4.1 Data Collection

The aim of each experimental run was to establish a thermal

equilibrium. For each set-up, the cylinders were aligned
vertically by using a plumb bob. Then, the wall(s) was placed
vertically at a specific wall spacing, S, by using feeler gauges and
a plumb bob. This was achieved by using the adjusting screws on
the main frame base. Next, the enclosure door and the testing
room door were securely closed prior to beginning each
experimental run and opened only at the end of each run. The
power supply and rheostats were adjusted in order to supply the
same power for each cylinder.
The power dissipation from each heater was calculated from
P=IV
where V is the voltage drop across the heater and I is the current
measured from the voltage drop across the one ohm resistance.
The manner in which I is measured overcomes the error in the
power dissipation if it is calculated from P = 12 R or P = V2/R
due to heater resistance changes when the temperature changes.
 51

Once the power was adjusted, it was not changed until all the wall
spacings results at fixed set-up (i.e., fixed center-to-center
spacing, CC) were recorded. For all the set-ups, the power
dissipation was adjusted to give the following heat flux: 49.338
W/m2 , 149.014 W/m2 , 493.38 W/m2 , 986.762 W/m2 , and
1480.143 W/m2. The wall spacing settings were started from the
smallest to the largest spacing. These values are shown in Table
3.1 for each case.

Table 3.1 Spacings of the cylinders and of the walls for the
experiments

CC/D Number of Variable wall Fixed wall Total

Cylinders Walls spacing, S/D spacing

non 1 non -- -- -- -- -- ---- 24

1.5, 2, & 4 3 1 .081 .155.25 .50 .75 1.0 1.5 2.0 inf ---- 135

1.5.2. & 4 3 2 .50 .75 1.0 1.5 2.0 2.5 3.5 -- 0.5 105

After the power was adjusted, the temperature monitoring

program, shown in Appendix A, was started on the Hewlett-
Packard computer. The program displayed the temperature
readings on the screen every twenty seconds. When the
variation in the average temperature was less than 0.2% for ten
 52

minutes, steady-state conditions were considered to be

established.The data were recorded one hour after the
establishment of the steady-state conditions in order to
accommodate the slow thermal response of the 1.27 Cm (1/2")
thick acrylic wall. As shown in Table 3.1, a total of 264
experimental runs were recorded.
Since the wall thickness of the cylinder(s) is very small,
0.889 mm (0.035 inch), and the thermal conductivity of
aluminum is much higher than the thermal conductivity of the
air, the temperature readings of the thermocouples were
considered to be the same as the temperatures of the cylinder
surfaces. The difference between the two temperatures, the
inner and the exterior cylinder surfaces, was found to be less
than 0.0066 degrees Centigrade, as calculated from equation 3.1.
For a one-dimensional steady state heat conduction from a
hollow cylinder, the total heat transfer, Q, can be represented as:

2 II KAI L (Tin Tw)

Q= Ln( rw )
(3.1)

For the present case:

rs =0.0127 m
rin=0.011811m
L =0.254m
KAL=204 W/m °C from [8].
 53

Therfore, Tw=Tin- (0.000222 Q) where Q= total input power in

watts.

3.4.2 Data Reduction

Once the average temperature of the cylinder(s) and the

ambient temperature were recorded and the power supply was
determined, the total heat transfer by convection, Qcv, could be
calculated from:

Qcv = Q Qr- Qcd

Qcd is the conduction heat loss from the cylinders' end-caps. This
heat loss was calculated from the Fourier's low as:

Qcd= K A dT
dX

Where dT is the temperature difference of the thermocouple

readings between the inner and outer surface of the end-caps,
dX is the end-caps' thickness, Ac is the end cap cross section area,
and K is the thermal conductivity of the end-caps' material (K =
0.0023 w/m °C). The maximum heat loss by conduction was
about .02 % of the total input power. This occurred when the
total input power, Q, was equal to 30 watts at center-to-center
spacing, CC = 1.5 D, and wall spacing ratio, S/D = 0.081 .

The radiation heat loss, Qr, ranged from 6% to 8% of the total

 54

input power for no wall cases and from 4% to 7% of the total

input power for both single and double wall cases. Since the
cylinders were polished to mirror-like surfaces, the emisivity was
considered equal to 0.05, [8, 22, 5, 16]. Appendix B shows the
methods that were used to calculate the view factors and the heat
loss by radiation from a free cylinder (within an infinite
medium), from an array of cylinders without walls, and from an
array of cylinders with wall(s).
Once the convection heat transfer, Q,, was determined, the
average heat transfer coefficient was obtained from:

Qcv
h
A (Tw Ting )

Then the average Nusselt number, Nu, was determined from:

hD Qcv D (3.2)
Nu = =
K A K (Tw Tiff )

For correlating the data, the modified Grashof number was

calculated from:
Qcv 4
p-2 D
Gr* (3.3)
K2 1.1.2

The modified Rayleigh number was calculated from:

 55

G13 p2 Cp Acv) D4
Ra*= Gr* Pr = (3.4)
K2 11

All the air properties were calculated at the film temperature,

Tf=:(Tw+Tinf)/2

from the following equations:

(1) The density was calculated by considering the air as an
ideal gas:

P Rg T= 0 Kg /m3 (3.5)

where p is the atmospheric pressure and Rg is the gas

constant.
(2) The thermal conductivity was calculated from the
following equation which is recommended for
temperatures up to 550 degrees C, [29]:

0.6325 x 10-5 T" 360.

K (3.6)
12) ( 0.86042 )
[1.0 + x 104 ]
K =() W/m.K

(3) The specific heat of the air was calculated from

equation 3.7. This equation is valid for a temperature
between 260 degrees K and 610 degrees K, [30]:
 56

Cp=(0.249679-7.55179x10-5 T+ 1.69194x10-7 T2
6.46128x10-11 T3) 1.162 = 0 W Hr/(Kg K) (3.7)

(4) The coefficient of thermal expansion, 13, was

calculated as:

= inf =0 K-1 (3.8)

(5) Hilsenrath et. al. [31] show that the viscosity of the air,
, at atmospheric pressure can be calculated from:

145.8 T(312)
P- (360.0 x 10-7) = 0 Kg/m Hr (3.9)
(T + 110.4)

Equations 3.5 through 3.9 were implemented as a subroutine in

the data reduction program, shown in Appendix C.

3.4.3 Uncertainty

The uncertainty in the Nusselt number was calculated as

shown in Appendix D. It was found that there is a maximum of
5.0% uncertainty in the calculated Nusselt number from the
experimental results.
 57

CHAPTER 4

SINGLE WALL: RESULTS AND DISCUSSION

The results of the heat transfer experiments and the

correlation equation for the Nusselt number for an isolated single
cylinder will be discussed at the beginning of this chapter.
Following will be a discussion of the experimental results for a
three cylinders array with a single wall. The correlation equation
representing the heat transfer coefficient in a known
dimensionless form (i.e. Nu) will be discussed for the three
cylinders' array with a single wall case at the end of this chapter.

4.1 Free Single Cylinder

The three cylinders that were used in the array were tested
separately as isolated single cylinders in an infinite expanded
medium (the surrounding air). The data were collected for each
cylinder for eight different heat flux conditions. These conditions
were as follows: 49.338 w/m2, 149.014 w/m2, 197.352 w /m2'
493.38 w/m2, 789.41 w/m2, 986.762 w/m2, 1480.143 w/m2, and
1973.525 w/m2. There were two primary reasons to study the
free single cylinder case. First, it was necessary to verify the
data-taking process and to check the related equipment setup.
This was accomplished by comparing the Nusselt number from
 58

this experimnent with the available data in the literature. The

second reason was to find a correlation equation to represent the
average Nusselt number for a single free cylinder, Nus, as a
function of the modified Rayleigh number, Ra* . This equation
would serve as the datum from which the heat transfer in the
other cases would be considered enhanced or degraded.
Figure 4.1 shows the average Nusselt number for each
cylinder versus the modified Rayleigh number for the eight heat
flux values. The functional relationship between Nus and Ra* for
these data is shown in equation 4.1:

Nus = 0.571 Ra*0.2027 3x104 < Ra* < 106 (4.1)

The r2 , defined as the percentage of the variability in the

dependent variable which is explained by the independent
variable, for equation 4.1 is 99%. The solid line in figure 4.1
represents equation 4.1, while the dashed line represents
equation 4.2 which is the correlation equation of Dyer [ 1O].

Nus = 0.6 Ra*0.2 104 < Ra* <106 (4.2)

Equation 4.1 lies about 1.6% below Dyer's equation. Equation

4.1 also agrees with the scale analysis results in Chapter 1 where
Nu is in the order of Ra*0.2 as shown in equation 1.9.
 10

7
z
6
Cylinder I
Cylinder 2
Cylinder 3
5 Author; Equation 4.1
Dyer : Equation 4.2

0 2 4 6 10
(X 100000)

Figure 4.1 Nusselt number, Nut, versus Ra* for a free single cylinder.
 60

4.2 Temperature distribution along the array with various wall

spacings:

The data are presented as the normalized temperature, 0,

which is defined as the ratio of excess temperature from the
ambient temperature for each cylinder divided by the excess
temperature from the bottom temperature, as shown in equation
4.3 .

Tw i Tinf
(4.3)
Tiff

These data are shown in Figures 4.2 through 4.6. Each figure
shows the data for constant heat flux and for the three center-to-
center spacings, CC, of the array. These values are shown in each
figure.
In these figures the normalized temperatures, 0, for CC =
1.5D and CC = 2D have a fixed pattern where the highest cylinder
always has a higher normalized temperature than the second
cylinder. In both cases the normalized temperatures are greater
than one. This indicates that the upper cylinders, numbers 2 and
3, are at a temperature higher than that of the lowest cylinder in
the array. This increase in cylinder temperature for cylinders
higher in the array is attributed to the balance between the
temperature increase of the surrounding air of the upper
cylinders and the increase of the plume velocity due to the
density change from adding heat from the lower cylinders to the
1.4

1.3

1.2

Li
1

0.9
CC-1.5D CC-2D CC-4D
0.8 Cylinder 3 0 4).
Cylinder 2

0.7
0 0.5 1 1.5 2 2.5

S/D
Figure 4.2 The effect of wall spacing on the normalized temperature at
q= 49.338W/m2.
 1.4
IV, ars V Vi1. SI, 11
1.3
di lb
1.2 . ,cl
)10K

1.1 -t
Ni

0.9
CC-I.5D
0.8 Cylinder 3 *
CC -2D
0
CC-4D
o
Cylinder 2
0.7 aiita1 gm lila tali. a11_ I
0 0.5 1 1.5 2 2.5
S/D
Figure 4.3 The effect of wall spacing on the normalized temperature
q= 149.014 W/m2.
 ir
1.4

1.3 ++
+
1.2 +
00 4. +
1.1
olE 11
)1(
* 0 0
iii 1ii 9 AE

0.9
CC-1.5D CC-2D CC-4D
0.8 Cylinder 3 * 0 0,
Cylinder 2

0.7 A . ltaIs a a slabs L 1 . * a I

fr.

0 0.5 1 1.5 2 2.5

S/
Figure 4.4 The effect of wall spacing on the normalized temperature at
q= 493.380 W/m2.
 1.4

1.3

1.2 + +
02 + + :
ME Iii + o 0
1.1 x(
.1111

0 Mal

4)---ox
At A(
ii w )*
1

* x
0.9

0.8

0.7
0 0.5
Cylinder 3
Cylinder 2

li
CC-1.5D

1
CC-2D
0
CC-4D

1.5
al l I
2 2.5
01

rem

S/D
Figure 4.5 The effect of wall spacing on the normalized temperature at
q= 986.762 W/m2.
1.4

1.3
*+
1.2
+
0
Li W )1( a
)1( )1(

xi IF-FM x
0.9
CC-1.50 CC-2D CC-4D
0.8 Cylinder 3
Cylinder 2

0.7 at I la 1__ al 11
0 0.5 1.5 2 2.5

Sip
Figure 4.6 The effect of wall spacing on the normalized temperature at
q= 1480.143 W/m2.
 66

plume [13,18 and 19].

The cylinders are more sensitive to the wall spacing ratio,
S/D, at a higher input power than at a lower input power. At the
higher input power, the normalized temperature data for CC = 1.5
D and CC = 2D decreases sharply as the S/D increases from 0.081
to S/D = 0.5. Then the normalized temperature increases as S/D
increases from 0.5 to 2.0, and attempts to reach the value of 0
when S/D is infinity. This can be observed by comparing Figures
4.2 to 4.6 with Figure 4.7. Also, at S/D = 0.5, the normalized
excess temperature difference between cylinders 2 and 3 is very
small and sometimes reaches zero. This indicates that the
behavior of these cylinders is the same at wall spacing ratio, S/D
= 0.5. As the wall spacing ratio diverts from S/D = 0.5, in either
direction, the value of (03 02) increases.
For CC = 4D, the normalized excess temperature, 0 , is less
than one for both cylinders (cylinders 2 and 3). The behavior of
these cylinders was not consistent. In general, the maximum
difference between the normalized excess temperatures for
cylinders 2 and 3 is less than 0.03% from that of the lowest
cylinder, where q1= 49.338 w/m2. This indicates that the upper
cylinders attain a lower temperature than the lowest cylinder.
This condition was also reported by Marsters [13], Marsters et. al.
[18], and by Lieberman et. al. [12]. At higher input power than q1
and at S/D < 1.0, this difference, (03 02), reaches zero. This
indicates that both cylinders 2 and 3 behave in the same manner.
 1.4
1.3
1.2
1.1

0.9
q (W/m2)
0.8 49.338
149.014
0.7 493.380
986.762
1480.143
0.6 a

1 2 3

Cylinder number
Figure 4.7 The effect of cylinder's position in the array on the normalized
temperature , 0, at S/D= infinity (no wall condition) .
 68

The dropoff in the normalized temperature is minimal and almost

constant for S/D < 1.0, but the dropoff increases as S/D increases
above 1.0. In all the cases (CC = 1.5 D, CC = 2D, and CC = 4D), the
upper cylinders behave more like the lowest cylinder at S/D
0.5.

4.3 Heat Transfer Coefficient Results

The heat transfer capability of each cylinder is characterized

by its Nusselt number. At the beginning of this section, the effect
of the wall spacing on the results of the lowest cylinder, cylinder
1, will be compared to the Nusselt number of a free single
cylinder. Then the results of all the cylinders for all the cases (CC
= 1.5 D, 2D, and 4D) will be presented and compared to the case
where there is no wall, S/D = infinity .

4.3.1 The effect of the wall spacing on the lowest cylinder of

the array

The Nusselt number versus the modified Rayleigh number

for the lowest cylinder at different wall spacings was plotted on
Figures 4.8 to 4.10. Also the Nusselt numbers from equation 4.1
for a free single cylinder were superimposed on these figures. In
all the cases, the Nui values at S/D = 0.5 lie higher than the
values for the free cylinder. This enhancement in Nu is very
 I
10

6 IMENO,

5 NINO

4 S/D-0.081 - - -0 S/D- 1.0

5/D-0.155 S/D-1.5
3 S/D-0.250 --- - - - S/D-2.0
S/D-0.500 x S/D-Int in.
S/D-0.750 Equation 4.1
2

10000 100000 1E6

Ra
Figure 4.8 The average XT.
ilusselt number of the lowest cylinder, Nui , versus
Ra*, at CC=1.5D.
 10

Af

./*

5
4e/
'/
--- - - - S/D-0.08I --- - - - S/D-I.0
it S/D-0.155 S/D-1.5
--x- S/D-0.250 - - - - S/D-2.0
3 S/D-0.500 x S/D-In1in.
S/D-0.750 Equation 4.1
2

10000 100000 1E6

R
Figure 4.9 The average Nusselt number of the lowest cylinder, Nu l , versus
Ra*, at CC=2D.
 10

4 + S/D-0.081 --- - - - S/D-1.0

-
S/D-0.155

2
- x - S/D-0.250
S/D-0.500
S/D-0.750
x
---- ---
S/D-1.5
S/D-2.0
S/D-Inlin.
Equation 9.1

10000 100000 1E6

Wa
Figure 4.10 The average Nusselt number of the lowest cylinder, Nu l, versus
Ra*, at CC=4D.
 72

small. It is about 5% maximum. For the rest of the values of the

wall spacing, with the exception of S/D = 0.081, the Nui values lie
between S/D = 0.5 and the values of the free cylinder. The
maximum percentage of the Nui enhancement is less than the
uncertainty in the values of Nu. Therefore, the effect of the wall
spacing, where S/D > 0.155, on the lowest cylinder is minimal and
the lowest cylinder can be regarded as a free single cylinder.
The Nusselt number at S/D = 0.081 is degraded from that of
the free cylinder by 10 to 12% for CC = 1.5 D and 2D and by 0.03
to 10% for CC = 4D. From these plots (Figures 4.8 - 4.10), it can be
seen that there is a degradation in Nui as S/D increases from 0.5
to infinity where the Nui values reach the free single cylinder
values. This is also shown in Figures 4.11, 4.12 and 4.13 where
Nui versus Ra* were plotted for all the cylinders at S/D = infinity.
When equation 4.1 is superimposed on these figures, it becomes
apparent that the lowest cylinder in a free array (i.e., S/D =
infinity) behaves the same as a free single cylinder and it is not
affected by the presence of the cylinders above it. This fact was
also reported by Marsters [13], Marsters et. al. [18], and Tokura
et. al. [19]. The presence of the lowest cylinder degrades the
Nusselt number for the cylinders above it at the small center-to-
center spacings, CC = 1.5 D and CC = 2D. On the other hand, at CC =
4D, the presence of the lowest cylinder enhances the Nusselt
number for cylinders 2 and 3. At CC = 4D, the plume has the
opportunity to blend with the surrounding air particles and reach
 I 11-- "1- 11-11--r-T
1
10 !Roo

9 Ow.

Ow.

6 1. IN 4.1

11

110

5 aN
* Cylinder 3
0:1 Cylinder 2
+ Cylinder I
4 i-1... L...-....-1,- 2
*
ss-1_1.-1. .41-.-......1- ....4.......-1........2 k I 1

10000 100000 1E6

R5
Figure 4.11 The average Nusselt number for the array's cylinders without a
wall, NI , Vs. 12a* at CC=1.5D .
 rr r-i r--

oc' Cylinder 3
qv Cylinder 2
Cylinder I

10000 100000 1E6

R
Figure 4.12 The average Nusselt number for the array's cylinders without a
wall, , Vs. 12a* at CC=2D .
 ti

gr
OJ

*Cylinder 3
Cylinder 2
Cylinder I

10000 100000 1E6 1E7

Ra
Figure 4.13 The average Nusselt number for the array's cylinders without a
wall, Nui, f , Vs. Ra* at CC=4D .
 76

the upper cylinder(s) at a temperature lower than that when CC =

1.5 D or CC = 2D.
The enhancement of the upper cylinder(s) Nusselt numbers
were noticed by Marster [13] and Tokura et. al. [19]. For the
three cylinder array without a wall, Marsters showed that the
enhancement in Nu starts at CC > 4D, and for CC = 4D, the Nu is
almost the same as for the free cylinder. Tokura et. al. showed
that Nu for the upper cylinder(s) is higher than that of the lowest
cylinder when CC = 3D.

4.3.2 The Effect of the wall spacing on the heat transfer

from the cylinders of the array

The Nusselt numbers versus S/D, for all the cases and for
every cylinder, at specific heat flux are plotted in Figures 4.14
through 4.19. At CC = 1.5 D and CC=2D, the Nusselt numbers for
the upper cylinders, cylinders 2 and 3, increase rapidly as S/D
increases from 0.081 to 0.5. Then as the S/D values increase, the
Nu values decrease and approach the Nu values for the no wall
case. The degradation in the Nu values for the third cylinder is
greater than that for the second cylinder. In both cases, CC = 1.5D
and CC = 2D, the Nu values degraded as the cylinder position
becomes higher in the array. This degradation reaches its
minimum at S/D = 0.5, where Nu values for cylinders 2 and 3 are
almost the same and are closest to the Nu values for the lowest
 10

4>

44*fr *
Ai* sf . .......

rf
I.
Illk

fs q(w/m2). 49.338 493.380 1480.143

Cylinder 3
Cylinder 2
Cylinder
I All ILIL1A . . I

0 0.5 2 2.5

Figure 4.14 The effect of the wall spacing on the average Nusselt number of
each cylinder, Nui , at CC=1.5D .
 10

4
4 (W/M2) -
Cylinder 3
3 Cylinder 2
Cylinder I
2

0.5 1.5 2 2.5

SiD
Figure 4.15 The effect of the wall spacing on the average Nusselt number of
each cylinder, Nui , at CC=1.5D .
 w..7"-----,....17 7-11-7
...-- .
fir" --- -"- --- -- ---- --. ----. -41
.,"- .
..,
Z.al
q (W/m2) - 49.338 493.380 1480.143
Cylinder 3 .......
Cylinder 2
Cylinder I -0
- --

0 0.5 1 1.5 2 2.5

S/D
Figure 4.16 The effect of the wall spacing on the average Nusselt number of
each cylinder, Nui , at CC=2D .
 10

5 6

4
(W/m2) - 149.014
3 Cylinder 3
CYlinder 2 - -
Cylinder I
2

0 0.5 1 1.5 2 2.5

S/D
Figure 4.17 The effect of the wall spacing on the average Nusselt number of
each cylinder, Nui , at CC=2D .
 12

11

10

6
q (W/m2) - 49.338 493.380 1480.143
5 Cylinder 3

0
I.. t
0.5
Cylinder 2
Cylinder I
-

A A
-
A

1.5
I Ra
2 2.5
S/D
Figure 4.18 The effect of the wall spacing on the average Nusselt number of
each cylinder, Nui , at CC=4D .
 12

11

10

13

6
q (W/m2) -
5 Cylinder 3
Cylinder 2
Cylinder I
4 . .

0 0.5 1 1.5 2 2.5

S/D
Figure 4.19 The effect of the wall spacing on the average Nusselt number of
each cylinder, Nui , at CC--.---4D .
 83

cylinder.
For CC = 4D, Nu values for all the cylinders increase sharply
as S/D increases from 0.081 to 0.25. When S/D becomes higher
than 0.25, the Nu values of the lowest cylinder stays the same.
While for the upper cylinders, the Nu values are the same up to
S/D = 1. At S/D > 1, the Nui values for cylinders 2 and 3 increase
in an attempt to reach the Nuu values at S/D= infinity (i.e. no
wall cases).
The ratios, Nui/Nui,f, of the Nusselt number for each cylinder
at various S/D to the Nusselt number for that cylinder (no wall
condition) at the same input heat flux are plotted in Figures 4.20
through 4.34. First, the discussion will be focussed on those cases
where center-to-center spacings are small, CC = 1.5D and CC = 2D,
shown in Figures 4.20 to 4.29. In these figures all the cylinders
experience the same percentage of degradation, which is about
20% at S/D = 0.081. This degradation decreases as the S/D
increases and the values of Nui/Nuif reach unity at 0.155 < S/D <
0.25 for low heat flux cases and at S/D = 0.155 for higher heat
flux cases. At S/D > 0.25, the effect of the wall spacing on the
heat transfer of the cylinder depends on the position of the
cylinder in the array. For all the heat flux values, except the
lowest values (q = 49.388 w/m2), the upper cylinders have the
highest enhancement at S/D approximately equal to 0.5. At the
peak, the highest cylinder in the array has an enhancement
between 15% to 22% above the no wall case, while the second
 I;

1.3 an
U U II II II
O EM

1.2 am.

. 1 .* *4
e
lean

Wane

z 0.9
0 Cylinder 3
Cylinder 2
0.8 Cylinder I
Equation 4.4

0.7 &I 1 g a ag I

0 0.5 1 1.5 2 2.5

S/D
Figure 4.20 The effect of the wall spacing on the Nusselt number ratio,
Nui/Nui,f , at CC=1.5D and q=49.338 (W/m2). 03
41.
 1.3

1.2

1.1

z 0.9
<> Cylinder 3
* Cylinder 2
0.8 Cylinder I
Equation 4. 4

0.7
0 0.5 1 1.5 2 2.5
S/D
Figure 4.21 The effect of the wall spacing on the Nusselt number ratio,
Nui/Nui f , at CC=1.5D and q=149.014 (W/m2).
 1.3

1.2

1.1

0.9
Cylinder 3
0.8 t Cylinder 2
Cylinder I
Equation 4. 4

0.7
0 0.5 1 1.5 2 2.5
S/D
Figure 4.22 The effect of the wall spacing on the Nusselt number ratio,
Nui/Nui,f , at CC=1.5D and q=493.380 (W/m2).
 1.3

1.2

1.1

0.9
4() Cylinder 3
0.8 * Cylinder 2'
*Cylinder
Equation 4.4
0.7
0 0.5 1.5 2 2.5

Figure 4.23 The effect of the wall spacing on the Nusselt number ratio,
Nui/Nui,f , at CC=1.5D and q=986.762 (W/m2).
L3

1.2

Li

0.9
0 Cylinder 3
Cylinder 2
0.8 Cylinder 1.
Equation 4. 4

0.7
0 0.5 1 1.5 2 2.5
S/D
Figure 4.24 The effect of the wall spacing on the Nusselt number ratio,
Nui/Nui,f , at CC=1.5D and q= 1480.143 (W/m2).
 1.2

1.1

0.9

0.8 Cylinder 3
* Cylinder 2
Cylinder I
Equation 4. 4
0.7
0 0.5 1 1.5 2 2.5
S/D
Figure 4.25 The effect of the wall spacing on the Nusselt number ratio,
Nui/Nui,f , at CC=2D and q=49.338 (W/m2). co
 1.2

1.1

5 0.9

0.8 0t Cylinder 3
Cylinder 2
Cylinder I
Equation 4.4

0.7
0 0.5 1 1.5 2 2.5
S/D
Figure 4.26 The effect of the wall spacing on the Nusselt number ratio,
Nui/Nui,f , at CC=2D and (1=149.014 {W /m2).
 1.2

1.1

0.9

0.8 0. Cylinder 3
* Cylinder 2
Cylinder
Equation 4. 4
0.7
0 0.5 1 1.5 2 2.5
SAD
Figure 4.27 The effect of the wall spacing on the Nusselt number ratio,
Nui/Nui,f , at CC=2D and q=493.380 (W/m2).
 I III V ig ir I I
1.2 II

0 0
1.1 .3 .

.
.
t
JD
*. . . .
. . ,1%,.
.*

z5
4:.
0
. .

0.9
0
Cylinder 3
0.8 Cylinder 2
Cylinder i
Equation 4. 4

0.7 I I I IL I I 11 I

0 0.5 1 1.5 2 2.5

S/D
Figure 4.28 The effect of the wall spacing on the Nusselt number ratio,
Nui/Nui,f , at CC=2D and q= 986.762 (W/m2).
 i.2 1
V
1

O 0
1.1 0
.

z
5 0.9

0 Cylinder 3
0.8 * Cylinder 2
Cylinder I
Equation 4. 4

0.7 . . I a 1 1 IL I . I

0 0.5 1 1.5 2 2.5

S/D
Figure 4.29 The effect of the wall spacing on the Nusselt number ratio,
Nui/Nui,f , at CC=2D and q=1480.143 (W/m2).
 94

cylinder has an enhancement between 9% to 15%. The lowest

cylinder in the array reaches the peak with an enhancement of
about 5% maximum at S/D = 0.25 at all the heat flux values
except the lowest value. At the lowest heat flux value (q =
49.338 w/m2), all the cylinders reach the peak at S/D = 1 for
both of the cases (CC = 1.5D and CC = 2D). Cylinder 3 at CC = 2D
has the highest enhancement (i.e., 10%) of any cylinder. The
second cylinder at CC = 1.5D shows the highest enhancement (i.e.,
14%) compared to the no wall case at S/D > 0.75. For CC = 1.5D
and CC = 2D, all the cylinders' enhancements decrease as the S/D
value increases from the S/D values at the peak in an attempt to
reach the Nui/Nuu value of unity.
Next, the discussion will be focussed on the effect of the wall
spacing, S/D, on Nui/Nuu for CC = 4D, as shown in Figures 4.30 to
4.34. The lowest cylinder shows a 10% maximum degradation at
S/D = 0.081. As S/D increases higher than 0.155, the Nui/Nuu of
the lowest cylinder increases to its maximum of 10% at S/D = 1.0
at the lowest heat flux, q = 49.338 w/m2. For the higher heat flux
values, the maximum enhancement for the lowest cylinder was
5% at S/D = 0.5.
For the lowest heat flux values, the three cylinders have
similar ratios of Nui/Nui j up to S/D = 0.5 where the middle
cylinder shows the lowest enhancement. As S/D increases from
0.5 at lowest heat flux, the highest cylinder shows higher
enhancement than the lowest cylinder, while the middle cylinder
 I V I I 1-
1.2 mo
V V I ff

....... ,11 9
.
1.1 -f-

0.9

0.8 Q Cylinder 3
* Cylinder 2
Cylinder I
Equation 4.4

0.7 I A a A A A. I
A A A A I

0 0.5 1 1.5 2 2.5

SID
Figure 4.30 The effect of the wall spacing on the Nusselt number ratio,
Nui/Nui,f , at CC=4D and q= 49.338 (W/m2).
 1.2

1.1

z
0.9

Cylinder 3
0.8 Cylinder 2
Cylinder I
Equation 4,4

0.7
0 0.5 1 1.5 2 2.5
S/0
Figure 4.31 The effect of the wall spacing on the Nusselt number ratio,
Nui/Nui , at CC=4D and (1=149.014 (W/m2).
 I II f 1 I I .
II I I
1.2

1.1

z 0
0
z 0.9 :0

0.8 O Cylinder 3
t Cylinder 2
0 Cylinder I
Et/Wilton 4.4

0.7 IA A A 1144_411,44.4_41 AA A 1 11 I

0 0.5 1.5 2 2.5

S/D
Figure 4.32 The effect of the wall spacing on the Nusselt number ratio,
Nui/Nui,f , at CC=4D and q=493.380 (W/m2).
 1.2

1.1

Cylinder 3
0.8 Cylinder 2
Cylinder
Equation 4.4

0.7
0.5 1 1.5 2.5

Figure 4.33 The effect of the wall spacing on the Nusselt number ratio,
Nui/Nui,f , at CC=4D and q=986.762 (W/m2).
 I . . V
I
1.2 iv Ir
I I

1.1

4-
....... .....
0.9 O
O 0 0
:o 0, Cylinder 3
0.8 * Cylinder 2
Cylinder I
O Equation 4.4

0.7 I a a i I

0 0.5 1.5 2 2.5

SiO
Figure 4.34 The effect of the wall spacing on the Nusselt number ratio,
Nui /Nuif , at CC=4D and q=1480.143 (W/m2). l.0
 100

stays at a lower enhancement than the lowest cylinder. At the

lowest heat flux, the maximum enhancement of the highest
cylinder is 18% at S/D = 1.0, while for the second cylinder it was
6% at S/D = 1.0.
When the heat flux is higher than 49.338 w/m2, the
maximum enhancement in the lowest cylinder was 5% at S/D
values between 0.25 and 0.5. For the upper cylinders, the
Nui/Nuu values were less than one for q > 49.338 w/m2. These
degradations are between 12% to 24% at S/D = 0.081 and they
decrease as S/D increases. The reduction of Nu3/Nu3,f from unity
is twice the reduction of Nu2/Nu2,f from unity. These values,
Nui/Nuu (i=2 & 3) , increase as S/D increases.
The average Nusselt number ratio for the whole array,
Nuavg, is superimposed on Figures 4.20 to 4.34, where Nuvag is
shown by the dotted line, as calculated from:

3
Nui
Nuav
i=1 (4.4)

Nuav (4.5)
Nuav /f=
E Nui f
i =1
3.0

Although Nuav/f is plotted as a line, the plot is a point to

point curve. At q = 49.338 w/m2, Nuavif shows an 8%
 101

enhancement for CC = 1.5 and CC = 2D at an S/D between 0.75 and

1.0. For CC = 4D, Nuav has about 14% enhancement at an S/D
between 1.0 and 1.5. For CC = 1.5D and CC = 2D with q > 49.338
w/m2, the Nuavif values have peaks, with 8 12% enhancement,
at .5 < S/D < 0.75. The Nuavg decreases sharply as S/D decreases
from 0.5 and reaches its lowest value about 0.82 at CC = 1.5 and
about 0.84 at CC = 2D at S/D = 0.081. When S/D increases to more
than .75, the Nuavj values decrease slowly in an attempt to reach
unity at S/D = infinity.
The Nuav/f values for CC = 4D are always less than one when
q > 49.338 w/m2. These values increase sharply as S/D increases
from 0.081 to 0.5. Then the increments in these values of Nuavif
become smaller as Nuavg attempts to reach unity.
Figures 4.35 to 4.37 show the Nusselt number ratios, Nuavis
of the Nuav values to the Nusselt numbers of a single cylinder,
Nus, as calculated from equation 4.1 at Ra* of the lowest cylinder
of the array. The Nuavis is defined in the following equation:

3
Nui
i=1
Nuav 3.0
Nuav Is = (4.6)
Nus 02027
[ 0.571 Rai*

The enhancements of Nuav from Nus were insignificant and

Nuavis are mostly equal or less than one at CC = 1.5D and CC = 2D.
In both cases, there were peak values at .5 < S/D < .75, except at
 I
i.1 T
1

640

tn
z
0.9

q- 49.338 Wha2
0.8 q- 149.014 W/m2
q- 493.380 W/m2
q- 986.762 W/m2
-- ----- q- 1480.143 Wha2
0.7 a A 1 IL L t 4 1 a

0 0.5 1 1.5 2 2.5

S/D
Figure 4.35 The effect of the wall spacing on the average Nusselt number of
the whole array at CC=1 .5D .
 1.1

0.9

q- 49.338 W/m2
0.8 q- 149.014 W/m2
O q- 493.380 W/m2
q- 986.762 W/m2
--- q- 1480.143 W/m2
0.7
0 0.5 2 2.5

Figure 4.36 The effect of the wall spacing on the average Nusselt number of
the whole array at CC=2D .
 1.2

1.1

q- 49.338 W/m2
0.9 q- 149.014 W/m2
q- 493.380 W/m2
q- 986.762 W/m2

0.8
0 0.5 1 1.5 2 2.5
S/0
Figure 4.37 The effect of the wall spacing on the average Nusselt number of
the whole array at CC=4D .
 105

the lowest heat flux, where the peaks were at .75 < S/D < 1.0. For
CC = 4D, there was a maximum of an 8% enhancement in Nua,
values at .25 < S/D < .5 in all the heat values, except the lowest
heat flux where the maximum enhancement was 10% at 1.0 < S/D
< 1.5.

4.4 Data Correlation

The experimental data will be represented in emperical

equations in this section. The average Nusselt number for each
cylinder of the array was fitted in a single equation at a specified
S/D value and a specified CC value. In these equations, the
average Nusselt values, Nut, were represented as a function of
Rat *, cylinder position in the array (Yi), and center-to-center
spacing (CC). Figure 4.38 shows the (Yi) dimensions. The general
form of this relation is:

Nui
= A1 + A2 Exp[ A3 ( (4.7)
Rai

The values of A1, A2, and A3 are shown in Table 4.1. The r2
for each case was also shown in this table. The Nui values, at S/D
= 0.5 , as calculated from equation 4.7 were superimposed on the
experimental data, as shown in figures 4.39 to 4.41.
 106

Side
wall

Figure 4.38 The measurement of Yi dimention as

used in equation 4.7.
 Table 4.1 Coefficients of equation 4.7, Nui/ltai".2 = Al + A2 Exp[ -A3 (Yi/CC)].

CC = 1.5D CC = 2D CC = 4D
S/D A A2 A3 r2% Al A2 A3 r2% Al A2 A3 r2%

0.081 0.38816 0.12099 1.68782 95 0.44023 0.08724 1.23577 86 0.53390 0.00448 -0.86297 40

0.155 0.42637 0.15017 1.15017 89 0.46097 0.14512 0.62651 94. 0.49841 0.10969 -0.02341 6.4

0.250 0.47994 0.12880 1.08535 88 0.48866 0.13475 0.46858 93 0.49055 0.12841 -0.05972 44

0.500 0.53707 0.07132 1.97020 83 0.57548 0.04468 0.84132 74 0.52552 0.08878 -0.11651 74

0.750 0.52015 0.08865 1.43274 83 0.86924 -0.25232 -0.06193 61 0.59267 0.02977 -0.15812 71

1.000 0.50866 0.10537 1.28668 83 0.56256 0.05466 0.83912 80 0.49579 0.12126 -0.10031 74

1.500 0.47515 0.12910 1.05709 94 0.36239 0.24567 0.17936 81 0.39546 0.21512 -0.09113 70

2.000 0.45226 0.14330 0.89824 96 0.36551 0.24209 0.19971 90 0.26659 0.33887 -0.07347 70

INF IN. 0.44692 0.14101 1.13765 96 0.46811 0.12704 0.51142 85 0.09139 0.50926 -0.08129 80
 11.5

.5

7.5

8.5

5.5

4.5

10000 100000 1E8

R6
Figure '1.39 The experimental Nu; values and correlated Nui values Vs. Ra* at
S/DA.5 and CC -1.51) .
 to

5 7

10000 100000 tea

R
Figure 4.40 'Hie experimental Nui values and correlated Nui values Vs. Ra* at
S /D =O.5 and CC= 21) .
 1

Exprimental Equation 4. 7
Cylinder 3 O
Cylinder 2
i0 Cylinder I MI*

10000 100000

RA
Figure 4.41 The experimental Nui values and correlated Nui values Vs. Ra* at
S/D=0.5 and CC=4D .
 111

Table 4.2 shows the coefficient of the curve fitting, which is

defined in Equation 4.8, for the average Nusselt number for the
whole array,Nuay.

3
Nui
i=1
Nuav = 7.0 = B1 Ra*° '2 (4.8)

Table 4.2 Correlation coefficient, B1, for equation 4.8,

Nuav = B1 Rai".2

CC = 1.5D CC = 2D CC = 4D
S/I) B1 r2% B1 r2% B1 r2%

0.081 0.43243 99.5 0.47916 98.8 0.55287 98.2

0.155 0.49607 98.0 0.53926 97.1 0.60844 98.7
0.250 0.54069 99.2 0.56980 97.5 0.62368 98.8
0.500 0.56403 99.7 0.59535 99.2 0.62452 99.5
0.750 0.55490 99.8 0.59425 98.9 0.62771 99.9
1.000 0.55014 99.1 0.58631 99.2 0.63009 99.9
1.500 0.53139 99.3 0.56906 99.2 0.63291 99.7
2.000 0.52108 99.6 0.56284 99.5 0.63612 99.5
INFIN. 0.50840 99.7 0.55027 99.3 0.64887 95.4
 112

CHAPTER 5

TWO WALLS: RESULTS AND DISCUSSION

This chapter will focus on the effects of the asymmetrically

placed cylinders array between two parallel walls. As shown in
Chapter 4, the maximum enhancement in a single wall condition
occurs at S/D 0.5. To investigate whether or not there is any
possibility to enhance the heat transfer coefficient in two wall
cases, the following conditions will be discussed in this chapter.
For two wall cases, the left wall was kept at S/D = 0.5, while the
right wall spacings were varied as follows: (S/D)R = 0.50, 0.75,
1.00, 1.50, 2.00, 2.50, and 3.50 for the same heat flux values of
the single wail (i.e., 49.338 w/m2, 149.014 w/m2, 493.38 w/m2,
986.762 w/rn2, and 1480.143 w/m2). The same cylinder center-
to-center spacings were used in the two wall cases as in the
single wall cases (i.e., CC = 1.5D, 2D, and 4D).

5.1 The effects of the right wall spacing, (S/D)R, on the

temperature distribution along the array

The normalized access temperature, 0 , as shown in

equation 4.3, was used to show the effects of the wall spacing on
the cylinders of the array. 0 versus (S/D)R for all the cases (CC =
1.5D, 2D, 4D) at a specific heat flux value are shown in figures 5.1
through 5.5. From these figures the 0 values for the upper
 I
1.6
1.5
1.4
1.3
i.2
ts)
1.1 .74.

0.9
CC-1.50 CC-2D CC -4D
0.8 Cylinder 3 .
Cylinder 2 * x"
0.7 a a a

0 1 2 3 4

(S /D)R

Figure 5.1 The effect of right wall spacing on the normalized temperature
at q= 49.338 W/m2.
 1.6
1.5
1.4
1.3
1.2
1.1
1

0.9
CC- 13D CC-2D CC -4D
0.8 Cylinder 3 --s-- 0
Cylinder 2 --* x.
0.7
0 1 2 3 4

(SiD)R

Figure 5.2 The effect of right wall spacing on the normalized temperature
at q= 149.014 W /m2.
 1.6
1.5
1.4
1.3
1.2
1.1

0.9
0.8
0.7
0 1 2 3 4

(S/D)R

Figure 5.3 The effect of right wall spacing on the normalized temperature
at q= 493.380 W/m2.
 1.6
1.5
1.4
1.3
1.2
1.1
1

0.9
CC -1.5D CC -2D CC-4D
0.8 Cylinder 3 ---t .0
Cylinder 2 ---- - x
0.7
0 1 2 3 4

(S /D)R

Figure 5.4 The effect of right wall spacing on the normalized temperature
at q= 986.762 W /m2.
 V

1.6
1.5
1.4
a
1.3
1.2
Li x .:9 *-
X- -6z .

0.9
CC- I .5D CC-2D CC-4D
0.8 Cylinder .3 -- -0- -
Cylinder 2
0.7
0 1 2 3 4

(S/MR

Figure 5.5 The effect of right wall spacing on the normalized temperature
at (F--- 1480.143 W/m2.
 118

cylinders, cylinders 2 and 3, are shown to be greater than one

and the 0 values are shown to decrease as the (S/D)R increases.
Also, the reduction in 0 values, as (S/D)R increased, was found
to decrease as the center-to-center spacing (CC) increased.
Therefore, 0 is more sensitive to (S/D)R as the CC values decrease.

5.2 The effects of the right wall spacing, (S/D)R, on the heat
transfer from each cylinder in the array

The average Nusselt number, NUiRw, for each cylinder was

calculated from equation 3.2 as discussed in chapter 3. Figures
5.6 through 5.11 show the effects of the right wall spacing (S/D)R
on the Nusselt number of each cylinder at a specific input heat
flux value. Figures 5.6 and 5.7 were plotted for CC = 1.5D and
Figures 5.8 and 5.9 were plotted for CC = 2D. From these figures,
the effects of (S/D)R on Nlli,Rw for CC = 1.5D are similar to the
effects of (S/D)R on Nlli,Rw for CC = 2D. Figures 5.6 to 5.9 also
show that the Nusselt numbers of the lowest cylinder decrease as
(S/D)R increases. For the upper cylinders, the Nusselt numbers
increase as (S/D)R increases. It can be noticed that the lowest
cylinder is more sensitive to the wall spacing (S/D)R than the
upper cylinders (cylinders 2 and 3) and has the highest Nu values
at a specific heat flux.
The highest cylinder has the lowest Nu value at a specific
heat flux and (S/D)R < 1.5. At (S/D)R > 1.5, NU2Rw and Nu3,Rw
were found to have almost the same values, and at times N113,Rw
 I I
i3

11

.0 --0 .4- ............- _..i.

54 *- --. ---,... *.4.-1-- ---1-
_

Z
''''' **** orr r-1
.111P.` -6
45 . . ihr

q (W/m2) - 49.338 493.380

Cylinder 3
Cylinder 2 -
Cylinder I
. . &

0 i 2 3 4

(S )R

Figure 5.6 The effect of the right wall spacing on the average Nusselt
number of each cylinder,Nui,Rw , at CC=1.5D .
 .4. . ..
4 -4
z
5-4 - 4-*

. ---
41)** it....
ep.

q (W/m2) - 199.019 986.762

Cylinder 3 4

Cylinder 2
Cylinder I --4/
/

0 2 3
(S/D)R

Figure 5.7 The effect of the right wall spacing on the average Nusselt
number of each cylinder,Nui,Rw , at CC=1.5D .
 q (W/m2) -
Cylinder 3
Cylinder 2
Cylinder I

Figure 5.8 The effect of the right wall spacing on the average Nusselt
number of each cylinder,Nui,Rw , at CC=2D
 13

11

7
z
5

q (W/m2)-
3 Cylinder 3
Cylinder 2
Cylinder I
1

0 1 2 3 4

(S/D)R

Figure 5.9 The effect of the right wall spacing on the average Nusselt
number of each cylinder,Nui,Rw , at CC=2D
 14

12

10

04.
5 8
z
6

q (W/m2) - 49.338 493.380 1480.143

4
Cylinder 3
Cylinder 2
Cylinder I
2 A a

0 1 2 3 4

(SID)R

Figure 5.10 The effect of the right wall spacing on the average Nusselt
number of each cylincler,Nui,Rw , at CC=4D .
 14

12

10

'4
6 e

q (W/m2) -
4 Cylinder 3
Cylinder 2
Cylinder I

0 1 2 3 4

(S/D)R

Figure 5.11 The effect of the right wall spacing on the average Nusselt
number of each cylinder,NuiRw , at CC=4D .
 125

became higher than Nu2,Rw. This can be explained as follows. At

(S/D)R < 1.5, the plume was restricted to upward movement, and
the upper cylinders were surrounded by a high temperature
plume. As the right wall spacing (S/D)R increased, the plume at
the right side of the array had a lower velocity than the plume at
the left side of the array. The high velocity of the plume at the
left side of the array created an unbalanced pressure at the top of
the array and caused a reversed current from the ambient to the
spacing between the two walls above the array. This current
removed more heat from the top cylinder, which resulted in an
increase of the Nu3Rw value.
At CC = 4D, the Nusselt numbers for the three cylinders
decrease as (S/D)R increases, as shown in figures 5.10 and 5.11.
At a specific heat flux value, the lowest cylinder has the highest
Nusselt number in the array, the second cylinder has the next
highest Nusselt number, and the highest cylinder (cylinder 3) has
the lowest Nusselt number. These findings are unlike the single
wall case, where Nu3 is the highest and Nui is the lowest.
This is because the right wall restricted all the air particles
to move in the direction of the high temperature raised plume,
which surrounded the upper cylinders. This will be discussed in
detail in Chapter 6.
Three different Nusselt number ratios were used in this
chapter to study the effects of the right wall spacing on the
cylinders of the array. These ratios are:
 126

1) (Nui,Rw/Nui,E ), the Nusselt number of each cylinder to

the Nusselt number as calculated from equation 4.1,
Nu 1,E = 0.571 Rai,f*0.2027, where Rai; is the modified
Rayleigh number value of the lowest cylinder when
there is no wall (i.e., free array) at the same q. This ratio
is shown in the following equation:

Nui jtw Nui,Rw

(5.1)
NuJ,E [ 0.571 Ra * 0.2027 ,

2) (Nui,Rw/Nui3O.5), the Nusselt number of each cylinder to

the Nusselt number of the same cylinder with the
presence of a single wall at S/D = 0.5.
3) (Nui,Rw/Nui,f), the ratio of the Nusselt number of each

cylinder to the Nusselt number of the same cylinder

when there is no wall (i.e., free array).
The deviation of Nui,Rw from the Nusselt number value of a
free cylinder, Nuis, can be noted by the Nui,Rw/Nui,E ratio. This is
attributable to the fact that the lowest cylinder in a free array
behaves the same as a free single cylinder, as shown in Chapter 4.
Figures 5.12 to 5.17 show the Nui,Rw/NuLE versus (S/D)R. At CC=
1.5D and CC = 2D, there was a 30% to 40% enhancement in the
lowest cylinder (cylinder 1) when the right wall spacing was
(S/D)R = 0.5 (i.e., the array was placed midway between the two
walls). For CC = 4D, this enhancement was between 40% and 50%.
Figure 5.12 The effect of the wall spacing on the Nusselt number ratio,
Nui,Rw/Nui,E , at CC=1.5D .
 1.4

1.3 1
Pit
1.2
z
1.1
cc
6
z i
". LA
II

lr"
0.9
e, ;
0.8 4" 4. +.
q (w/m2) 149.014 986.762
Cylinder 3
0.7 Cylinder 2
Cylinder
0.6
0 1 2 3 4

(S /D)R

Figure 5.13 The effect of the wall spacing on the Nusselt number ratio,
Nui.Rw/Nui,E , at CC=1.5D
 1.5

1.4

1.3

1.2

1.1

0.9
0.8
0.7
0 1 2 3 4

(S/D)R

Figure 5.14 The effect of the wall spacing on the Nusselt number ratio,
Nui.Rw/NuLE , at CC=2D.
 1.5

1.4

1.3

1.2

1.1

0.9
(iNilin2)-
0.8 Cylinder 3
Cylinder 2
Cylinder I
0.7

0 1 2 3 4

(SiD)R

Figure 5.15 The effect of the wall spacing on the Nusselt number ratio,
Nui,Rw/Nui,E , at CC=2D.
 1.5 -' I I

1.4 -
14 .

1.2
+. N-- t--- --- ---17.
z----.......
-

i
1:4. ---t---- L.:-.1L. 11---...:4"'ir,;-:71-f=":
6 1.1 .....4. .-F 4- +- ÷
ir-.

e. 'a

0.9? q (W/m2) - 49.338 493.380 1480.143

0.8 Cylinder 3 (>

Cylinder 2 - -0-
Cylinder 1 --
0.7 I . 1 1 I It I I 1 i I . I.

o 1 2 3 4

(S/D)R

Figure 5.16 The effect of the wall spacing on the Nusselt number ratio,
Nui,Rw/Nui,E , at CC=4D.
 1.5
1.4
14
5 1.3
1.2
cc
1.1

0.9
q (W/m2) - 149.014

0.8 Cylinder 3
Cylinder 2

0.7
Cylinder I ----
0 2 3 4
(S/D)R

Figure 5.17 The effect of the wall spacing on the Nusselt number ratio,
Nui,Rw/Nui , at CC=4D.
 133

For all the CC values, the ratio, Nui,Rw/NuLE , decreased as the

(S/D)R increased.
The behavior of the lowest cylinder ratios, Nui,Rw/Nu 1,0.5 (as
shown in figures 5.18 to 5.23) and NuiRw/Nu j (as shown in
figures 5.24 to 5.29) is the same as the behavior of N111,RwiNU1,E.
This is because NuLE represents the correlated function of Nu1,f
(which is the exact value), and N1.11,0.5 is about 5% higher than
Nu LE as shown in chapter 4.
In figures 5.12 to 5.29, the upper cylinders (cylinders 2 and
3) have the same pattern at CC = 1.5D and CC = 2D. While at CC =
4D, the upper cylinders behave differently. The Nusselt number
ratios show three stages for the upper cylinders at CC = 1.5D and
CC = 2D. The first stage is between (S/D)R = 0.5 to 1.0, where the
Nusselt number ratios were almost the same at a specific q. The
second stage range is from (S/D)R = 1.0 to 2.0. In this stage the
ratios (Nui,Rw/NuLE , Nui,Rw/Nui,o.5 , and NuiRw/Nuu ) increase
sharply as (S/D)R increases. The third stage starts at (S/D)R = 2.0
and the ratios are less sensitive to (S/D)R changes.
The second cylinder shows a 2% to 10% degradation in the
first stage of (NuiRw/Nui,E) and (Nui,Rw/Nui3O.5), while the third
cylinder shows an 8% to 20% degradation in this stage. This
means that the (Nui,Rw/Nui,E) and (Nui,Rw/Nui3O.5) for the third
cylinders are lower than that of the second cylinder at CC = 1.5D
and CC = 2D. The second stage is the transition stage where the
Nu ratios for the third cylinder cross the Nu ratios of the second
 1.4

1.3

1.2

1.1

0.9

0.8
0 2 3 4

(S/D)R

Figure 5.18 The effect of the wall spacing on the Nusselt number ratio,
Nui,Rw/Nuo.5 , at CC=1.5D.
 1.4

i.3

if) 1. 2
5
.

c4.

z
0.9

0.8

0 2 3 4

(S /DAR

Figure 5.19 The effect of the wall, spacing on the Nusselt number ratio,
Nui,Rw/Nui.0.5 , at CC=1.5D.
 1 . 4

1. 3

1. 2

q (Whu2) -
0.8 Cylinder 3
Cylinder 2
Cylinder I
0.7 .

0 1 2 3 4

(SA)R

Figure 5.20 The effect of the wall spacing on the Nusselt number ratio,
Nui,Rw/Nuo.5 , at CC=2D.
 1.5 1
11NIN

1.4 awr

1.3

1.2
2
1.1

,.
......
1 D.,
(W/m2) - 199,019 986.762
0.9 .4.. 4: Cylinder 3 ...,... ...tame
Cylinder 2
Cylinder I
0.8
0 1 2 3 4

(S /D)R

Figure 5.21 The effect of the wall spacing on the Nusselt number ratio,
Nui,Rw/Ntio.5 , at CC=2D.
 s
1.5
q (W/m2) -
Cylinder 3
1.4 Cylinder 2
Cylinder I
1.3

1.2
z
1.1
04.

0.9

0.8
0 2 3 4

(S /DAR

Figure 5.22 The effect of the wall spacing on the Nusselt number ratio,
Nui,Rw/Nuo.5 , at CC=4D.
Figure 5.23 The effect of the wall spacing on the Nusselt number ratio,
Nui,Rw/Nui3O.5 , at CC=4D.
 1.4

1.3 o

1.2

II

1.1
trtc
' -0-
!
q (W/m2) - 49.338 493.380 1480.143

0
0.9 Cylinder 3 0*
Cylinder 2 - - a -- "O'
Cylinder I --6--- ____._____

0.8 a a

0 1 2 3 4

(MA
Figure 5.24 The effect of the wall spacing on the Nusselt number ratio,
Nui,Rw/Nuif , at CC=1.5D..
 1 . 5

1. 4

1. 3
. 4. ......... +
..
1 . 2 _-4
4:*
es
1.1 -4:
+ 4:
ao
1
q (W/m2) - 149.014 986.762
Cylinder 3 . +. .
0. 9 Cylinder 2
Cylinder 1
0.8 J
0 2 3 4

(S/D)R

Figure 5.25 The effect of the wall spacing on the Nusselt number ratio,
Nui,Rw/Nuis , at CC=1.5D.
 1.6

1.5

1.4

1.3

1.2

z54 1.i

q (W/m2) -
Cylinder 3
0.9 Cylinder 2
Cylinder I
0.8
0 1 2 3 4

(S/D)R

Figure 5.26 The effect of the wall spacing on the Nusselt number ratio,
Nui,Rw/Nuif , at CC=2D.
 1
1.5

1.4

1.3
0.1

1.2

1.1 .
1

q (W/m2) - 149.014 986.762

Cylinder 3
0.9 Cylinder 2
Cylindtr I
0.8 1 . . a 1 a .

0 1 2 3 4

(MR
Figure 5.27 The effect of the wall spacing on the Nusselt number ratio,
Nui,Rw/Nuis , at CC=2D.
 I
1.6
q (W/n12) - 49.338 493.380 1480.143 -
1.5
1.4
Cylinder 3
Cylinder 2
Cylinder I 0
1.3
z 1.2
1.1
1 .1 0
0.9 0
0.8
0.7
0 1 3

(S/D)R

Figure 5.28 The effect of the wall spacing on the Nusselt number ratio,
Nui,Rw/Nui,f , at CC=4D.
 1.5
1.4
1.3
1.2
1.1
ci4
1

0.9
0.8
0.7
0.6
0 2 3 4

(S/D)R

Figure 5.29 The effect of the wall spacing on the Nusselt number ratio,
Nui,Rw/Nui , at CC=4D.
 146

cylinder at (S/D)R 1.25. The ratio values of the third cylinder

become higher than that of the second cylinder in the second
stage, as shown in figures 5.12 to 5.15 and 5.18 to 5.21. In the
third stage, the degradation in the ratios ((Nui,Rw/Nui,E) and
(Nui,Rw/Nui3O.5)) values were less than those in the first two
stages. In general, the upper cylinders at CC = 1.5D show a higher
degradation than at CC = 2D. There were no significant
enhancements in the upper cylinders' Nusselt numbers from
NULE and NUL0.5

The upper cylinders' Nusselt numbers were enhanced,

relative to Nuu , with the presence of the second wall at CC =
1.5D and CC = 2D. This is shown in figures 5.24 to 5.27. In these
figures, the three stages relative to (S/D)R are also shown. In the
first stage, the second cylinder has an enhancement higher than
the third cylinder. The enhancement of the second cylinder in
the second stage increases as (S/D)R increases from 1.0 to 2.0. In
the third stage, the increment of (NuiRw/Nuif) in the second
cylinder is slower than in the second stage and it reaches the
same enhancement as the first cylinder by the end of this stage.
For the highest cylinder (cylinder 3), (Nu3,Rw/Nu3,f) increases
sharply after the first stage and crosses the lines of the first
cylinder by the end of the second stage. In the third stage the
third cylinder has a higher enhancement than the lowest
cylinder. The enhancement in (Nu3,Rw/Nu3,f) at CC = 1.5D in the
second and third stages (i.e., (S/D)R = 1.0 to 3.5) is higher than
 147

the enhancement at CC = 2D in the same stages.

For CC = 4D, the Nusselt number ratios (NuiRw/Nui,E) ,
(Nui,Rw/Nui3O.5), and (NuiRw/Nui,f) were shown in figures 5.16,
5.17, 5.21, 5.22, 5.28 and 5.29. In these figures, the cylinders'
enhancements have the same pattern in which the first cylinder
has the highest enhancement (between 40% to 55% at (S/D)R =
0.5). The second cylinder enhancement is always higher than the
third cylinder enhancement. The enhancement of the first
cylinder is more sensitive to the wall spacing (S/D)R than the
upper cylinders (cylinders 2 and 3) and decreases as (S/D)R
increases.

5.3 The effects of right wall spacing on the average Nusselt

number of the whole array, Nuav,Rw

The arithmetic mean of the Nusselt numbers of all the

cylinders of the array was calculated, as shown in equation 5.2, at
specific (S/D)R and specific heat flux.
3
E Nui Rw
i=1
Nilav,RW 3.0 (5.2)

Figures 5.30 through 5.32 show the Nuav,Rw versus (S/D)R

for CC = 1.5D, 2D, and 4D, respectively. At CC= 1.5D and CC = 2D,
the Nuav,Rw values are insensitive to the right wall spacing,
(S/D)R. This indicates that the total enhancement of Nlli,Rw - total
degradation of NUiRw is a constant value at a specific heat flux
 q- 49.338 W/m2
* q- 149.014 W/m2
* q- 493.380 W/m2
0 q- 986.762 W/m2
x q- 1480.143 W/m2
A I.
1 3 4

(S /D)R

Figure 5.30 The effect of the wall spacing on the average Nusselt number
of the whole array at CC =l.5D.
 . . . . .
10 X
X X
0 O

)1( )I( )1(

ONO

q- 49.338 W/m2
q- 149.014 W/m2
* q- 493.380 W/m2
q- 986.762 W/m2
x q- 1480.193 W/m2
WEI

1 2 3 4

(S /DAR

Figure 5.31 The effect of the wall spacing on the average Nusselt number
of the whole array at CC=2D .
 . , .
13
Xx
IWO o x X
11 - o X
X
)1( Ni O
* )1( 1

q- 49.338 Who`
5 * q- 119.014 W/m2
QM.

* q- 493.380 W/m2
0 q- 986.762 W/m2
3 x q- 1980.143 W/m2
A A A A I

0 1 2 3 4

(MR
Figure 5.32 The effect of the wall spacing on the average Nusselt number
of the whole array at CC=4D .
 151

value. There is an insigificant drop (approximately 5% maximum)

in the average Nusselt number at (S/D)R between 0.75 and 2.0, as
shown in figures 5.30 and 5.31.

The NUav,Rw for CC = 4D is shown to be decreasing as the

(S/D)R increases, figure 5.32. This is an expected result for CC =
4D since the Nui,Rw values for all cylinders decreases with the
increasing of (S/D)R values, as shown in the previous section. In
all the cases of CC, the Nuav,Rw increases as the heat flux values
increase.
The average Nusselt number for the whole array, Nuav,Rw,
was compared to the correlated Nusselt number of a free cylinder
through the following ratio:

Nuav,RW If =
[*yNuav,Rw

0.571 Ra.2027
(5.3)

Where Ralf* is the modified Rayleigh number of the lowest

cylinder for a free array (i.e., no wall) at the same heat flux value
at which the Nuav,Rw was calculated from equation 5.2. This
ratio, Nuav,Rwif, versus (S/D)R was plotted in figures 5.33 to 5.35
for CC = 1.5D, 2D, and 4D, respectively. These figures show that
the Nuav,Rw values are enhanced by the presence of the right wall
for all the heat flux values except the lowest value, q = 49.338
w/m2, for CC= 1.5D and 2D. In general, the figures of the
 1 . 1

q (W/m2) .
49.338
149.014
493.380
986.762
1480.143
0.8 .

0 1 2 3 4

(S /DAR

Figure 5.33 The effect of the wall spacing on the Nusselt number ratio,
Nui,Rw/Nui , at CC=1.5D.
 1. 2

1 . 1

q (W/m2)
99.338
149.014 -
493.380 -
986.762
1480.193
0.8
0 1 2 3 4

(S/D)R

Figure 5.34 The effect of the wall spacing on the Nusselt number ratio,
Nui,Rw/Nuis , at CC=2D.
 i.5

1.4

1.3

1.2
01. U

1.1
q (W/m2)
49.338
149.014
493.380 -
986.762
1480.143
0.9 . . I

0 1 2 3 4

(S/D)R

Figure 5.35 The effect of the wall spacing on the Nusselt number ratio,
Nui,Rw/Nui,f , at CC=4D.
 155

enhancement of Nu av,RW increase as the center-to-center

spacing, CC, increases. The maximum enhancement occurs when
the walls are placed symmetrically to the array (i.e., (S/D)R = 0.5).
These enhancements were between 2% to 8% for CC = 1.5D, 9% to
13% for CC = 2D, and 20% to 44% for CC = 4D. The exact value of
these enhancements depends on the heat flux value, q. Figure
5.33 shows that there is a degradation of about 4% maximum in
Nuav,Rwif at the lowest heat flux value for (S/D)R > 0.5.
There was a heat transfer degradation in the lowest heat
flux case at CC = 2D and (S/D)R > 2.0. Maximum degradation in
this case was 3% at (S/D)R = 3.5, as shown in figure 5.34. In
general, the enhancement in NUav,Rwg at CC = 1.5D and 2D
decrease as (S/D)R increase from 0.5 to 1.50. At (S/D)R>1.50,
the enhancement in Nuav,RW/f values increase due to the
enhancement in the Nusselt number of the highest cylinder,
, in the second and third stages due to the feed back
NU3,RW

current. The enhancement in Nuav,Rwg at CC = 1.5D and 2D are

insensitive to the right wall spacing at (S/D)R greater than 2.0 .

This is contradictory to the enhancement of NuavRw for CC = 4D,

where the ratio, Nuav,Rwif , decreases continuously as (S/D)R
increases, as shown in figure 5.35. Also, it can be noticed from
figure 5.35 that the enhancement decreased by 10% at the lowest
q and by 24% at the highest q between (S/D)R = 0.5 and
(S/D)R=3 .5 .
The effects of the right wall spacing on Nu,,Rw compared to
 156

the average Nusselt number of the whole array for a single wall
at S/D = 0.5, NUav,0.5, were investigated through the Nllav ,Rw to
Nu av,0.5 , ratio. The ratio (Nuav,Rw/Nuav,0.5 ) versus (S/D)R is
presented in figures 5.36 to 5.38. In general, there was no
degradation in Nllav ,Rw relative to Nua0.5 The enhancements in
(NUav,RW/NUav,0.5 ) at CC = 1.5D were between 4% and 9% at (S/D)R
= 0.5 and between 2% and 7% at (S/D)R = 3.5. For CC = 2D, the
enhancements were approximately 8% at (S/D)R = 0.5 and
approximately 3% at (S/D)R = 3.5 at all q values, except q =
49.338 w/m2, as shown in figure 5.37. Also, figures 5.36 and
5.37 show that the enhancements in Nuav,Rw values are
insensitive to the right wall spacings at (S/D)R > 2.0 .

Figure 5.38 shows the enhancements of Nua,,Rw at CC = 4D,

relative to Nu av,0.5 The pattern of the values is the same as that
in figure 5.35, except that the values in figure 5.38 are shifted
10% lower than those in figure 5.35. Unlike the (Nuav,Rw/Nuav,0.5)
at CC = 1.5D and 2D, the (Nuav,Rw/Nuav,0.5) at CC=4D is more
sensitive to (S/D)R and it decreases as (S/D)R increases.

5.4 Data Correlation

The curve fitting of the data was presented to fit the Nusselt
number as a function of the modified Rayleigh number to the 0.2
power. This relation is shown in equation 5.4:
 i. i .1 VV. V g

0.9
(W/m2)
49.338
149.014
* 493.380
0 986.762
1480.143
0.8 I

0 2 3 4

(S /D)R

Figure 5.36 The effect of the wall spacing on the Nusselt number ratio,
Nui,Rw/Nui.o.5 , at CC=1.5D.
 1.2

1.1

q (W/m2)
0.9 49.338
149.014
493.380
986.762
1480.143
0.8
0 1 2 3 4

(S /D)R

Figure 5.37 The effect of the wall spacing on the Nusselt number ratio,
Nui,Rw/Nuo.5 , at CC=2D.
 1.4

1.3

1.2

1.1

q (W/m2)
49.338

0.9 149.014
493.380
986.762
1480.143
0.8
0 1 2 3 4

(S/D)R

Figure 5.38 The effect of the wall spacing on the Nusselt number ratio,
Nui,Rw/Nuo.5 , at CC.---4D.
 160

NUozw = B1 Rai,Rw*0.2 (5.4)

The correlation coefficient, B1, for each cylinder at specific (S/D)R

is shown in table 5.1 with r2 values for each case.

Table 5.1 The correlation coefficient, B1 , in equation 5.4 .

Cylinder 1 Cylinder 2 Cylinder 3

CC S/D B1 r2% B1 r2% B1 r2%

1.5D 0.50 0.7647 98.9 0.5342 99.4 0.4922 99.3

1.5D 0.75 0.7385 98.8 0.530198.9 0.4882 98.9
1.5D 1.00 0.7190 99.8 0.5276 98.9 0.4911 99.2
1.5D 1.50 0.6906 99.6 0.5117 99.4 0.5326 97.4
1.5D 2.00 0.6717 99.6 0.5386 96.7 0.5622 96.8
1.5D 2.50 0.6648 99.8 0.5556 97.7 0.5667 97.1
1.5D 3.50 0.6542 99.7 0.5667 98.7 0.5631 97.2
2.0D 0.50 0.7797 99.2 0.5793 99.4 0.5302 99.2
2.0D 0.75 0.7572 95.9 0.569199.9 0.5310 99.5
2.0D 1.00 0.739499.1 0.566499.5 0.535398.3
2.0D 1.50 0.6990 99.2 0.5693 99.3 0.5728 98.8
2.0D 2.00 0.6809 98.4 0.5953 98.0 0.5907 97.4
2.0D 2.50 0.6696 98.3 0.6007 97.4 0.5907 97.0
2.0D 3.50 0.6537 98.4 0.6034 96.8 0.5857 96.8
4.0D 0.50 0.8430 99.9 0.728197.5 0.6897 93.5
4.0D 0.75 0.8094 96.5 0.6976 93.4 0.6729 92.3
4.0D 1.00 0.7758 98.9 0.6818 97.9 0.6532 95.3
4.0D 1.50 0.7436 99.3 0.6723 96.8 0.6536 94.6
4.0D 2.00 0.7104 99.5 0.6632 97.8 0.6434 96.2
4.0D 2.50 0.6914 99.6 0.6597 98.7 0.6402 96.7
4.0D 3.50 0.6769 99.4 0.6655 98.8 0.6402 98.7

Figures 5.39 through 5.41 show the calculated Nusselt

 as

la

es

Z 7

$
$
4
4 is
(X 101 wow
Rs' Ra'

ss

la

P
ai
Z 7
Z 7

3
4 all is 4 le la
(X t.00001
Ra' (X MOON
Ra'
Figure 5.39-a The experimental NUi.Rw values and the correlated NUi.Rw
values Vs. Ra* at CC=1.5D .
 Is

SS

II 4 SO ill
Ol 1000001
Ra' IX 14001101

Dig iseatal &evades 5.4

Cylinder 3 Co
Cylinder 2
Cylinder I

0 4 SO
Ot 1000101

Figure 5.39-b The experimental Nui,Rw values and the correlated NUi,Rw
values Vs. Ra* at CC=L5D .
 13
$3

SS
IS

a
Z 7

3
I IS
IS is
DC 100000
Rs' Rs'
01 1000001

13

SS
SS

Zz 7

1 1 SO 1
01 000001
el 1000001

Figure 5.40-a The experimental NUi,Rw values and the correlated I\TUi,Rw
values Vs. Ra* at CC= 2D .
 as

11

a
OD

Z 7

a
I 4 tt lo as a 4 $ so
Rs*
01 MOON a looms
Rs'

as

le
0 Dor Imolai Equation 3.4
a
Z 7
Cylinder 3
Cylinder 2
o
_____
Cylinder I

0 I 4 ao as
a loom)
Ra'
Figure 5.40-b The experimental NUi,Rw values and the correlated NUI,RW
values Vs. Ra* at CC= 2D .
 *4

4 1 IS
et 1000001

Ra' Ra' IX 1000001

Figure 5.41-a The experimental Nui,Rw values and the correlated Ntli,Rw
values Vs. Ra* at CC=4D .
 14

is

is
z

4
4 is is a la
Re mous ix moo*
Rs'

Expr Equation 1.4

Cylinder 3
Cylinder 2
Cylinder I

0 4 in
01 100000)
Ra'
Figure 5.41-b The experimental Nui,Rw values and the correlated Nui,Rw
values Vs. Ra* at CC=4D .
 167

number from equation 5.4 superimposed on the experimental

data, for all the CC values.
In these figures, the Nusselt number for the lowest cylinder,
Nu 1,Rw, was higher than the Nu2,Rw and Nu3,Rw. The Nusselt
numbers for cylinders 2 and 3 are very close when compared to
the Nusselt number for cylinder 1. Also, from figures 5.39 to
5.41, the difference between NU2,Rw and NU3,Rw decreases as
(S/D)R increases, with NU3,Rw taking the lowest value. At high
(S/D)R values, the NU3Rw becomes greater or equal to the NU2,Rw.
This is due to the enhancement of the Nusselt number of the
highest cylinder, NU3,Rw. This enhancement occurs as a result of
the feedback current. The maximum error in this correlation, as
calculated from equation 5.5 is less than 5% for 18x104<
,

R a 1Rw< 1.6x106 when the lowest heat flux value is not included
and about 12% for 6.4x104<Ra1,Rw<1.6x106 when all the heat

flux values are considered.

( Nuexperimental Nucorrelation
Error % = x 100 (5.5)
Nuexperimental

Table 5.2 illustrates the B1 values and r2 for the average

Nusselt number correlation for the whole array, Nuav,Rw This
relationship is the same as the one shown in equation 5.4, except
that the modified Rayleigh number that is used in this
relationship is the modified Rayleigh number of the lowest
cylinder in the array, Rai,Rw . From equation 5.5 , it is found that
 168

the maximum error in this correlation is less than 3% for 18x104<

Rai,Rw< 1.6x106 when the lowest heat flux value is not included
and about 10% for 6.4x104<Ra1,Rw<1.6x106 when all the heat

flux values are considered.

Table 5.2 The Correlation coefficient, B1, for

Nu av,RW= B1 Rai,Rw *0.2

CC =1.5D CC =2D CC =4D

S/D B1 r2% B1 r2% B1 r2%
0.50 0.5843 98.9 0.6184 99.5 0.7474 98.1
0.75 0.5737 99.2 0.6084 98.6 0.7214 96.5
1.00 0.5680 99.7 0.6025 99.1 0.6987 98.8
1.50 0.5694 98.6 0.6067 99.5 0.6859 98.2
2.00 0.5844 98.6 0.6176 98.4 0.6694 98.7
2.50 0.5902 99.1 0.6164 98.1 0.6617 99.1
3.50 0.5898 99.1 0.6111 97.8 0.6561 99.4
 169

CHAPTER 6

NUMERICAL ANALYSIS AND FLOW VISUALIZATION

The temperature and velocity distributions and flow

visualizations for selected cases will be presented in this chapter.
Also, the results of Chapters 4 and 5 will be discussed. In order
to prevent repetitious discussion, only a limited number of
experimental conditions were chosen to represent the three cases
(i.e., CC = 1.5D, 2D and 4D). Further, selection of a limited number
of cases allowed for more efficient use of the available executable
time for the numerical analysis. This discussion will present the
numerical analysis and flow visualization results concurrently.
The numerical analysis was used to present the temperature
fields and velocity vectors around the cylinders. A finite
difference computer program for three dimensional
hydrothermal analysis, Tempest [32], was used to conduct the
numerical analysis. Some computer language modifications of
this program were necessary in order to use Tempest on the
available computing system.
The Floating Point (FPS) super computer was used as the
scientific computer and the IBM 4381 was used as the front end
computer. Both systems are available at the O.S.U. computer
center. Due to the extensive executable time needed for Tempest
to reach a steady state condition for the cases of the present
 170

study, those conditions with q = (149.014) w/m2 for a single wall

with CC = 1.5D, 2D and 4D and for double walls with CC = 1.5D and
2D will be presented.
Because Tempest accepts either cartesian or polar
coordinates but not both, the cylinders were represented by a
series of square cells. Also, multi-grid sizes were used to satisfy
the code requirements and the physical geometry of the set-up.
A sample of the grid system is shown in figure 6.1 for CC = 2D.
The flow visualization was recorded on video tape and
slides. The slides were taken by a 135 mm Pentax with a slide
film of 1600 ASA. The medium was illuminated by laser sheets
as described in Chapter 3.

6.1 Single Wall Cases

The temperature distribution, velocity distribution and flow

visualization for CC = 1.5D, 2D and 4D are shown in figures 6.2
through 6.4. All these cases are at q = 149.014 w/m2 and the left
wall spacing, (S/D), is equal to 0.5. From the output of the
numerical analysis of these cases, the maximum temperature
deviation from the experimental data was less than 5% as
calculated from equation 6.1:
Texperimental Tnumerical (6.1)
Tdeviation 100
Texpeftental
au 'paw ppoui JOJ isacllArai Joj az=3D Arlie
 29.36
111

II
4

1 1 it ). ,,,, , ,
32.72

36.08

t 39.44

42.80
a ,1\ ''', , c c
l* . . .
\
f..r.

\ .. ",- Y , , ,
..
42.80 P
I r tt , ,
V\ 0 , ,
f A
,- ',.., , , , , , ,
1 5 . 8 6 01)
29.36
\k.!. .$)

A F T r ,T 'T A , , , -A

0
1 CI
O
O

0.00 012 0.013 0.12

a) Temperature distribution. b) Velocity distribution. c) Flow visualization .

Figure 6.2 Temperature, Velocity, and Flow fields for CC=-1.5D array
with a single wall at S/D=0.5 .
ry
L_ L_
29.17

35.53
I IIII:c.Su. K , A' 444 A .4 44
38.71
tilirlOrkt. 4. -4 4 4 44 4 -4

I I \\k .0". 'V -4 4 -4 -4 4 4

11 Ay a)
(r)

t I illh.00400,1,,,.
tilt \NN "\ N- -A -A. At -A -S -A A

l'% 1 il IL y -( 4. -A A. 4 A4 O
29.17
32.35 i tIkA \''''k*"Pl'.YrsY 4 -A 4 4 4 44 44
35.53
/3

tt 4 1. -C 4 aL at at. 4. -d d

4 k."*"...-Irre.e.te-e -.4 -4 -a

A A r-rrrt-ff, - - .. . . . - --4 .-

°C) I

a
0_00 012 0.00 012

a) Temperature distribution. b) Velocity distribution. c) Flow visualization .

Figure 6.3 Temperature, Velocity, and Flow fields for CC=2D with a
single wall at S/D=0.5 .
 L

tilLeappote-f f 4 -4

661,4,11,rmweeccc G4

(°C)
0
0 00 0.12 0.00 0.12

a) Temperature distribution. b) Velocity distribution. c) Flow visualization

Figure 6.4 Temperature, Velocity, and Flow fields for CC=4D with a
single wall at S/D=0.5 .
 175

From figures 6.2 through 6.4, it can be seen that the space
between the cylinders and the wall act as a chimney where the
air particles are at a higher temperature than the ambient. The
high temperature zone creates a low density region between the
wall and the cylinders of the array. This leads to a low pressure
in this region. Due to buoyancy forces, this low pressure drives
the air particles horizontally toward the wall through the
cylinder-to-cylinder spacing and vertically through the lower
cylinder-wall space. As the center-to-center spacing, CC,
increases, the raised plume from one cylinder to the next higher
cylinder has more time to be mixed with the ambient air particles
which were directed toward the wall. Also, as the CC spacing
increases, the raised plume from the lower cylinder(s) has fewer
flow restrictions given the geometry of the set-up and has more
time to be accelerated by the buoyancy forces while it rises. This
can be noticed by comparing the velocity vectors of figures 6.2
through 6.4.
From the numerical results at CC = 4D, as shown in figure
6.4(a), the maximum temperature difference between the three
cylinders is 0.18 degrees Centigrade, whereas a maximum
difference of 0.4 degrees Centigrade is obtained from the
experimental results. This indicates that each of the three
cylinders acts like a single cylinder due to the balance between
the increased temperature of the rising plume (which reduces the
heat transfer) and the increased velocity of the plume (which
 176

enhances the heat transfer) as it ascends the array. Furthermore,

the flow visualization of figures 6.2 through 6.4 shows that the
weak regions above the cylinders were shifted towards the wall.
The streamlines of the flow show that the air particles approach
the upper cylinders horizontally. For the lowest cylinder, the air
particles approach the cylinders in a plane rotating toward the
vertical wall. In other words, if there is a single wall at the left
side of the array, the upper and lower stagnation points of each
cylinder were rotated counter clockwise from the vertical plane
(the plane which passes through the axis of each cylinder). This
rotation of the stagnation points was at its maximum at S/D = 0.5
and the rotation decreases as the wall-array spacing decreases or
increases from S/D = 0.5.
The flow visualization showed that the presence of a single
wall induces more air circulation from the ambient around the
highest cylinder in the array. This enhances the heat transfer
from the highest cylinder and thus explains why the highest
cylinder behaves closely to the middle cylinder (cylinder 2). For
a free array (no wall case), the flow visualization showed that the
streamlines were symmetrically arranged about the plane of the
array. Also, the stagnation points of the cylinders lay on the
vertical plane which passes along the axis of each cylinder. The
flow was laminar around the cylinders in the arrays with small
cylinder center-to-center spacings (i.e., CC = 1.5D and 2D). For the
highest center-to-center spacing (i.e., CC = 4D), the flow was
 177

laminar around the first and second cylinders and unstable (or in
a transition condition) above the second cylinder. In this
condition, the streamlines of the flow swung from one side of the
array to the other. This instability of the flow began above the
second cylinder and its frequency increased as it rose in the
array. This explains why the upper cylinders were at a lower
temperature than the lowest cylinder. Furthermore, when the
power was increased, this instability of the flow began even
closer to the second cylinder.

6.2 Two Wall Cases

The flow visualization showed that the flow was laminar and
symmetrical about the vertical plane which passes through the
centers of the cylinders. This is true only when the arrays were
placed symmetrically between the walls (i.e., at S/D = (S/D)R = 0.5
conditions). In this condition, the stagnation points of all the
cylinders lay in the same vertical plane and the weak regions lay
vertically above the cylinders. The weak regions between the
cylinders for CC = 1.5D and 2D were extended to include the
whole spaces between the cylinders, as shown in figure 6.5 (a).
In these spaces, the flow visualization showed that there were
double vortices. The left vortex rotates clockwise, while the right
vortex rotates counter clockwise. For CC = 4D, the weak regions
were present to a limited distance (i.e.,a distance less than the
 a) Flow visualization, for CC=2D. b) Flow visualization, for CC-4D.

Figure 6.5 Flow visualization for arrays symmetrically placed

between two walls at S/D=(S/D)R=0.5 .
 179

cylinder-to-cylinder spacing) above each cylinder, as shown in

figure 6.5 (b).
As the array to the right wall spacing, (S/D)R , increased (i.e.,
the array was placed asymmetrically between the walls), the
streamlines of the flow were no longer symmetrical on both sides
of the array. This result occurred due to the left wall jet. This
jet was established as a result of the higher temperature zone
between the array and the left wall to it as compared to the rest
of the region.
As the right wall spacing, (S/D)R, increased, the velocity
vectors shifted from the vertical direction (for symmetrical walls
placements) toward the closest wall to the array. For the region
above the array, the left wall jet induced a feedback current (or a
reverse current). This current penetrated downward close to the
farthest wall (right wall) from the array and introduced cooler air
particles between the walls. Then, the reverse current circulated
upward along the left wall jet, as shown in figure 6.6 and figure
6.7 .

The penetration of the reversed current became deeper as

the farthest wall spacing increased. At large (S/D)R , the reverse
current region extended to include the highest cylinder in the
array. This flow around the highest cylinder enhanced its heat
transfer capability.This explains why the Nusselt number of the
highest cylinder became closer to or higher than the Nusselt
number of the second cylinder of the array for CC = 1.5D and 2D.
 111A V V V V

\kltv Ay V V V V

39.26

4.4\11 v V. V V
4r" ''''s
9.20 1 : V V

39.26
h\v....4-NiNR
35.94

8.46

32.63
V

29.31
)1,4 4.49 A

ilkX% 4 410 &

A

t#44444 A

0 ( °C), 2
0
0.00 0.1 0.00 0.12

a) Temperature distribution. b} Velocity distribution.

Figure 6.6 Temperature, and Velocity fields for CC =l.5D array with
two walls at S/D=0.5 and (S/D)1Z =2.0 .
 tI
______ 29.11

III Pkk 4 kr r v v V

\\\\ 38.45
4 e.R\ \ R v V
41.56
I \ A V V

tA\ I' 4 ? A l V

35.34 11 Ilt IR *.lik, ik. .1k *. , c V

Illa 38.45
tt \\ ?. , 1

6.7) \\ / 46 A I,-

t N k t t %t t.
32.22 s

r pA IA\ N 4 r A

t t t 4 r
29.11
it it 4 4. A 444 t I r

A44444444 A

oc)
9 0
0 12 D.00 C1.12
COO

a) Temperature distribution. b) Velocity distribution. c) Flow visualization .

Figure 6.7 Temperature, Velocity, and Flow fields for CC=2D array
with two walls at S/D=0.5 and (S/D)R=2.0 .
 182

Figures 6.6 and 6.7 show that the high temperature plume that
was rising from the cylinders shifted toward the closest wall to
the array as the other wall spacing increased.
 183

CHAPTER 7

CONCLUSIONS AND RECOMMENDATIONS

7.1 Conclusions

This study presented the results of the experimental

investigation of heat transfer from an array of three horizontal
cylinders aligned vertically parallel to a single wall or parallel to
two walls. Three different cylinder center-to-center spacings for
the array cylinders were examined, CC = 1.5D, 2D and 4D. The
wall-array spacings were varied from 0.081D to infinity for single
wall cases. For two wall cases, the left wall spacing was kept at
S/D = 0.5, while the right wall spacing varied from (S/D)R = 0.5 to
3.5. All the cylinders were maintained at constant heat flux.
Modified Rayleigh number, based on the diameter of the
cylinders, ranged from 6.2 x 104 to 1.2 x 106. Empirical equations
were proposed to predict the effects of the experimental
parameters on the heat transfer as expressed by the average
Nusselt number of each cylinder or the average Nusselt number
of the whole array.
This experimental investigation has provided new
information concerning natural convection heat transfer from an
array of cylinders confined by a single vertical wall or two
vertical walls. The new information obtained about the variables
 184

in this experiment can be useful in optimizing the design of heat

exchangers and in the packaging design of electrical components.
Also, the quantitative results can be employed in applied
research concerning heat transfer.

7.1.1 Single wall cases

The experiments have shown that the lowest cylinder in the

array behaves as a free single cylinder for all cases in which S/D >
0.155. The heat transfer from the upper cylinders of the array
was affected by the wall-array spacing (S/D values) and the
cylinder center-to-center spacing (CC values). The presence of
the wall enhanced the heat transfer from the upper cylinders of
the array with CC = 1.5D and 2D relative to the no wall cases,
while there was degradation in the cylinders' heat transfer
capability with CC = 4D. At the two small CC values the maximum
enhancement for the free array occurred at S/D 0.5. The third
cylinder had the maximum enhancement, the second cylinder had
the next highest enhancement, and the first cylinder had the
lowest enhancement. The Nusselt number for each cylinder can
be predicted from equation 4.7 with less than 5% error for q >
49.338 w/m2 and with 10% or less error for q = 49.338 w/m2.
The temperature and velocity distribution from the
numerical analysis and the flow visualization showed that the
space between the array and the wall acts like a chimney and
 185

creates a left wall jet flow. Further, the closest spacings between
each cylinder and the wall acts like a flow reinforcing station.
The second cylinder behaved more like the third cylinder
than the first cylinder (i.e., [Nul - Nu2] > [Nu2 - nu3]). This was
the result of additional cooling of the third cylinder due to
increased ambient air circulation above this cylinder. This
circulation was induced by the left wall jet. The bottom and top
stagnation points on each cylinder were shifted from the vertical
plane due to the presence of the wall. In the presence of a single
left wall, the stagnation points were shifted counter clockwise.
This shifting was at its maximum when S/D 0.5 and decreased
as S/D either increased or decreased.

7.1.2 Two wall cases

The lowest cylinder had the highest heat transfer coefficient.

This heat transfer coefficient was at its maximum when the walls
were symmetrically placed to the array. As the right wall
spacing, (S/D)R, increased the Nusselt number of the lowest
cylinder decreased. At the largest center-to-center spacing, CC =
4D, the lowest cylinder had the highest enhancement compared to
a free single cylinder.
For the upper cylinders at small center-to-center spacings
(CC = 1.5D and 2D), the heat transfer coefficient showed an
enhancement at (S/D)R > 1.5 compared to a free single cylinder or
 186

the single wall cases. At the largest center-to-center spacings

(CC = 4D), the upper cylinders showed enhancement in the heat
tranfer coefficient with the presence of the right wall for all
(S/D)R values. These enhancements were the highest at the small
(S/D)R and decreased as (S/D)R increased. When compared to
free array cases (i.e., no wall), all the cylinders in all the cases
showed an enhancement in the heat transfer coefficient.
The average Nusselt numbers for the whole array were
more sensitive to the right wall spacing as the center-to-center
values increased. The experimental results showed that the
average Nusselt numbers for the whole array were enhanced
relative to the average Nusselt number of the free single cylinder
and relative to the average Nusselt number of the whole array
for single wall cases. These enhancements increased and were
more sensitive to (S/D)R as the cylinders' center-to-center
spacings increased. At low center-to-center spacing values (CC =
1.5D and 2D), the enhancement of the whole array was at a
minimum at 1 < (S/D)R < 1.5. This can be explained by the fact
that the degradation in the upper cylinders was higher than the
enhancement in the lowest cylinder in this (S/D)R range. Table
5.2 shows the empirical equation to calculate the average Nusselt
number for the whole array.
The flow visualization and the numerical results indicate
that the flow was in transition above the array and a reversed
current was established between the two walls. This current
 187

region extended from the top of the walls down to enclose the
highest cylinder in the array.

7.2 Recommendations

The present work could be extended by considering the

following recommendations:
1) An investigation to explore the effects of misalligning
one or more of the cylinders of the array. This
misallignment might conceivably change the flow.
pattern around the cylinders, which in turn might affect
the heat transfer characteristics.
2) Using different fluids as the test medium, such as water
or oil, would provide greater understanding concerning
the effects of the Prandtl number on the heat transfer
behavior of the array.
3) Changing the diameter of the cylinders and/or the
number of the cylinders would add to the understanding
of the effects of the array-wall spacings on heat transfer.
4) Developing the flow visualization system to use laser
sheets for three dimentional heat transfer flow. This
could be accomplished by using two laser sheets
perpendicular to each other instead of superimposing
one laser sheet on the other.
 188

BIBLIOGRAPHY

1. Kreith, F.; Principles of Heat Transfer, International Textbook,

Scranton, 1958.

2. Parsons, J.R.,Jr. and Arey, M.L., Jr.; "Development of

Convection Heat Transfer near a Suddenly Heated Horizontal
Wire in a Still Fluid", ASME paper no. 84, WA/HT-50, 1984.
3. Parsons, J.R., Jr. and Arey, M.L., Jr.; "Development of
Convective Heat Transfer near Suddenly Heated Vertically
Aligned Horizontal Wires",Transactions of the ASME, Vol. 109,
pp. 912-918, Nov. 1987.
4. Sloan, J.L.; Design and Packaging of Electronic Equipment, Van
Nostrand Reinhold Company, Inc., New York, 1985.
5. Sparrow, E.M. and Niethammer, J.E.; "Effect of Vertical
Separation Distance and Cylinder-to-Cylinder Temperature
Imbalance on Natural Convection for a Pair of Horizontal
Cylinders", Transactions of the ASME, Vol. 103, pp. 638-644,
Nov. 1981.

6. Bejan, A.; Convection Heat Transfer, John Wiley & Sons, Inc.,
New York, 1984.
7. Levy, S.; "Integral Methods in Natural Convection Flow",
Journal of Applied Mechanics. pp. 515-522, Dec.1955.
8. Holman, J.P.; Heat Transfer, 6th edition, McGraw-Hill Book Co.,
New York, 1986.
9. Morgan, V.T.; "The Overall Convective Heat Transfer from
Smooth Circular Cylinders", in Advances in Heat Transfer, Vol.
11, Academic Press, New York, pp. 199-264, 1975.
 189

10. Dyer, J.R.; "Laminar Natural Convection from a Horizontal

Cylinder with a Uniform Convective Heat Flux", Mechanical
and Chemical Engineering Transactions, Institution of
Engineers, Australia, Vol. MC1, pp. 125-128, May 1965.
11. Kim, C.B., Pontikes, T.J., and Wollersheim, D.E.; "Free
Convection from a Horizontal Cylinder with Isothermal and
Constant Heat Flux Surface Conditions", Transactions of the
ASME, Journal of Heat Transfer, pp. 129- 130, Feb. 1975.
12. Lieberman, J. and Gebhart, B.; "Interactions in Natural
Convection from an Array of Heated Elements, Experimental",
Int. J. of Heat Transfer, Vol. 12, pp. 1385-1396, 1969.
13. Marsters, G.F.; "Array of Heated Horizontal Cylinders in
Natural Convection", Int. J. of Heat Mass Transfer, Vol. 15, pp.
921-933, 1972.

14. Sparrow, E.M. and Boessneck, D.S.; "Effect of Transverse

Misalignment on Natural Convection from a pair of Parallel,
Vertically Stacked, Horizontal Cylinders", Journal of Heat
Transfer, Vol. 105, pp. 241-247, May 1983.
15. Marsters, G.F.; "Natural Convective Heat Transfer from a
Horizontal Cylinder in the presence of nearby Walls", The
Canadian Journal of Chemical Engineering, Vol. 53, pp. 144-
149, April 1975.
16. Sparrow, E.M. and Pfeil, D.R.; "Enhancement of Natural
Convection Heat Transfer from a Horizontal Cylinder due to
Vertical Shrouding Surfaces", Transactions of the ASME, Vol.
106, pp. 124-130, Feb. 1984.
17. Karim, F., Farouk, B., and Namer, I.; "Natural Convection Heat
Transfer from a Horizontal Cylinder between Vertical
Confining Adiabatic Walls", Journal of Heat Tranfer, Vol. 108,
pp. 291-298, May 1986.
18. Marsters, G.F. and Paulus, G.; "Effects of Confining Walls on
Heat Tranfser from a Vertical Array of Heated Horizontal
APPENDICES
 190

Cylinders", Transactions of CSME, Vol. 1, No. 4, pp. 219-222,

1972.
19. Tokura, I., Saito, H., Kishinami, K., and Muramoto, K.; "An
Experimental Study of Free Convection Heat Transfer from a
Horizontal Cylinder in a Vertical Array set in Free Space
between Parallel Walls", Transactions of the ASME, Vol. 105,
pp. 102-107, Feb. 1983.
20. Sparrow, E.M. and Chrysler, G.M.; "Natural Convection Heat
Transfer Coefficients for a short Horizontal Cylinder attached
to a Vertical Plate", Transactions of the ASME, vol. 103, pp.
630-637, Nov. 1981.
21. Sparrow, E.M., Cook, D.S., and Chrysler, G.M.; "Heat Transfer
by Natural Convection from an Array of Short, Wall-Attached
Horizontal Cylinders", Journal of Heat Transfer, Vol. 104, pp.
125-131, Feb. 1982.
22. Sparrow, E.M., Mendes, P.S., Ansari, M.A., and Parta, A.T.;
"Duct-Flow versus External-Flow Natural Convection at a
Short, wall-attached Horizontal Cylinder", Int. J. of Heat Mass
Transfer, Vol. 26, No. 6, pp. 881-888, 1983.
23. Sparrow, E.M. and Ansari, M.A.; "All-Modes Heat Tranfser
from a Horizontal Cylinder Situated Adjacent to Adiabatic,
partially enclosing Walls", Int. J. of Heat Mass Transfer, Vol.
27, No. 10, pp. 1855-1864, 1984.
24. McCoy, T.J.; "Natural Convection from a Horizontal Cylinder
Parallel to a Heated Vertical Wall", Master's thesis,
Mechanical Engineering Dept., Oregon State University, 1988.
25. Temperature Handbook, OMEGA Engineering, Inc., Stamford,
CT, 1987.

26. Hot Spot Application Manual, DCC Corporation, Pennsauken,

NJ, January 1986.
 191

27. User Manual for DT2801 Series, Document UM-0066-D-1855,

Data Translation, Inc, January 1981.
28. User Manual for PC LAB, Document UM-02899-A, Data
Translation, Inc, January 1982.
29. Vines, R.G.; "Measurement of Thermal Conductivities of Gases
at High Temperatures", Transactions of the ASME, Feb. 1960.
30. Touloukian, Y.S.; "Thermophysical Properties of Matter", The
TPRC Data Series: Specific Heat Nonmetallic Liquid and Gases,
Vol. 6, 1970.
31. Hilsenrath, J. and Touloukian, Y.S.; "Viscosity, Thermal
Conductivity and Prandtl Number for Air, 02, N2, No, H2, Co,
H2, He and A", Transactions of the ASME, pp. 967-985, August
1954.

32. Trent, D.S. and Eyler, L.L.; Tempest: A Three-Dimensional

Time-Dependent Computer Program for Hydrothermal
Analysis, Vol. 1., Pacific Northwest Laboratory, Richland,
Washington, January 1989.
 192

Appendix A

DATA AQUISITION PROGRAM

 193

10'
20 ' This data aquisition program is developed to read four interface cards

30 ' by using four 012805 boards and four 01707-1 screw terminal panels.

40 PRINT 'Totale input power, 0 =

50 INPUT Q 0=GUATT
60 PRINT 'CYLINDER -TO- CYLINDER SPACIN6 INPUT CTCS

70 PRINT "Cylinder-Right wall spacing, US/0 . '; :INPUT USUR

80 PRINT "Cylinder-Left wall spacing, US/0 . '; :INPUT USOL

90 ' This file may be MER6Ed with a user program to define

100 ' routine offsets and establish the PCLAB segment.

110 '

120 XAU=3 : XA0T.6 : XSA.9 : 0.12 : XBA0.15 : XTAD=18

130 RUAD=21 : XOU =24 : X001=27 : 050 =30 : XOS=33 : X800=36

140 XI00=39 : MU00=42 : XSC0=45 : XSSC :48 : 01=51 : XSCP=54

150 0E11=57 1 00=60 : XIOU.63 : X00U:66 : X1001.69 : X0001.72

160 XSECU=75 : X6EC=78 : 058 =81 : XSBA=84 : XSOC=87 : XSAR=90

170 XSAC :93 : XSOR=96 : 0511 :99 : X516=102 : X51.105 : 060(.108

180 XR0=111 : X605.114 : XCUFC.111 : xcuro.in : XCUFI.123 : 00=126

190 00=129 : XCA0=132 : 0(00 =135 : XMLI.138 : X11I.141 : 4MC.144

200 XRIU=147 : XUTO=150 : XOTU.153 : XDLY=156 : 058=159 : RU0.162

210 X6C.165 : XCEU :168 : XREU :171 : X61.174 : XSC =1 ?7 : XINIT=180

220 X1ER11 .183 : XIXR =186 : XIXU.189 : XFOL.192 : XESC=195 : XOSC=198

230 '

240 OEF SE64110 ' get the PCLR8 segment

250 PCLSE6 : PEEK ( WIFE ) + 256*PEEK ( 1114FF )

260 OEF SE6=PCLSE6 ' REM address the PCLA8 segment

270 ERROR.UALUEK = 0

Z80 CALL XSECU(ERROR.UALUEZ)

290 '

300 RINI : 8
310 TYPE! : 116

320 MANX = 0
330 OPTION BASE 1

340 DIM TEMP(4,7)

350 '

360 CLS

370 ON TIMER(20) 60SUB 410

380 TIMER ON

390 6010 390

400 '

910 LOCATE 2,1

420 '

430 '

440 FOR HARE 1 TO 4

450 CALL XSUBORROZ)

460 FOR CHRNX = 1 TO 7

470 CJTOIRL . 0

480 TOTAL . 0
 194

490 FOR I . 1 TO 16

500 CALL XAU(CICHANX, GRIN:, CJADATAX)

510 CJTOTAL . CANTU + CJTOTAL

520 CALL XIMCRANZ, MINX, MAD
530 TOTAL = ROATAI + TOTAL

540 NEXT I

550 CARTA' = CINT(CJTOTAL / 16)

560 ROATAI = CINT(101fli / 16)

570 CJTEMP = MCJADATAZ * .04) / 4096) - .02) * 2000

580 CALL XOTU(TYPEI, CJTEIIP, CJUOLTS)

590 UOLIS = (((808181 * .01) / 4096) - .02) + CJUOLTS

600 CALL XUTO(TYPEX, UOLTS, TEMPS)

610 TEMP(80ARDI,CHAND = TEMPS

620 NEXT CHM

630 NEXT MARDI
640 RUER1 = (TEMP(1,1)+TEMP(2,1)+TEMP(3,1)+TEMP(4,1))/4!

650 AUER2 = (TEMP(1,2)+TEMP(2,2)+TEMP(3,2).TEMP(1,2))/4!

660 AUER3 = (TEMP(1,3)+TE11P(2,3)+TEMP(3,3)+1EMP(4,3))/4!

670 CIS

680 PRINT 'Total input per = '; : PRINT USING 11.11*;0;

690 PRINT ' Uatt

700 PRINT ' CC e:CICS

710 PRINT 'Cylinder-Right wall spacing, US/0 . "; :PRINT USIN6 11.1111":USOR

720 PRINT "Cylinder-Left wall spacing, US/0 . ' ::PRINT USING 11.1111';USOL

730 PRINT

740 PRINT ' ( TEMPERATURES IN DEGREE CENTIGRADE)

750 PRINT 'Thermocouple 1 1 2 3 4 5 6"

760 PRINT USING 'Cylinder 3 111.11 111.11 111.11 111.11 111.11 111.11';

TEMP(1,3):TE11P(2,3);TEMP(3,3);TEMP(4,3);TEMP(1,5);TEMP(2,5)

770 PRINT USING 'Cylinder 2 111.11 1#1.11 Mil 111.11 111.11 111.11';

TE11P(1,2):TEMP(2,2);TEMP(3,2);TEMP(4,2);TEMP(3,4);TEMP(4,4)

780 PRINT USING 'Cylinder 1 111.11 111.11 111.11 111.11 111.11 111.11';

TEMP(1,1);TEMP(2,1);TENP(3,1);TEMP(4,1);TEMP(1,4):TEMP(2,1)

790 '

800 PRINT

810 PRINT luerage temperature for cylinder 1 3 = ";

820 PRINT USING '111.11 C';AUER3

830 PRINT "Ruerage temperature for cylinder 1 2 = ';

840 PRINT USING '111.11 C";RUER2

850 PRINT 'Average temperature for cylinder 1 1 .

860 PRINT USING "111.11 C";RUER1

870 PRINT

880 3TTEMP.1 EMP(1,6)-TEMP(1,7)

990 4:MAUR.,TEMP(1,6)+TEMP(1,7))/2!

900 PRINT "imoient temperature differance = '; : PRINT USIN6 111.51 C";STTEMP
910 PRINT 'Rmbient temperature . '; : PRINT USING "1111.111 C";RMAUR

920 TENDS.TEMP(3,5)-TEMP(3,6)

930 7EMOZ.TEMP(2,6)-1EMP(2,7)
 195

940 TEND1=TEMP(I,S)-TEMP(4,6)

950 PRINT 'End cap temperature difference for cylinder 13 ';

960 PRINT USING '111.11';TEN03

970 '

980 PRINT 'End cap temperature difference for cylinder 12 = ';

990 PRINT USING '111.11';TEN02

1000 '

1010 PRINT 'End cap temperature difference for cylinder 11 = ';

1020 PRINT USING '111.11';IEN01

1030 PRINT

1040 '

1060 ' CHECK FOR STEADY STATE CONDITION:

1060 '

1070 P1.4UER1-00L1)100!/0101
1080 P2 :(AUER2-0102)*100!/0012

1090 P3 =(ADER3-01.03)'100!/01.23

1100 '

1110 IF P1(.2 AND P2(.2 RHO P3(.2 RHO FLA61( >1 THEN FLR61=1 :CTIME.TIMER+600!

1120 0101=RUERI : 0102411ER : 0103=RUER3

1130 If fl061.1 RHO P1).2 OR P2).2 OR P3)1 THEN FLR61.0

1140 If FUTON RHO (CTIMD600)>TINER THEN 6010 1190

1150 PRINT ",*****44***

1160 PRINT '---- STEADY STATE CONDITION UAS REACHED IN THE LAST TEN MINETS
1170 PRINT ' RI TIME . ';CTIME600;' SEC
1180 '

1190 RETURN

1200 STOP

1210 END
 196

APPENDIX B

RADIATION CORRECTION

The data reduction procedure, shown in section 3.4.2,

illustrates that the accuracy of the natural convection heat
transfer coefficient of the cylinders was dependent on the
accuracy of the radiation correction. The cylinders were designed
to make 0-rad << Qconvection This was accomplished by polishing
the surfaces of the aluminum cylinders to a mirror-like finish.
The procedure to determine the radiation corrections for the
cylinder in a free single cylinder, an array of cylinders without a
wall, an array of cylinders with a single wall, and an array of
cylinders with two walls will be discussed in the following
sections.

B.1 Radiation correction from a free single cylinder and an array

of cylinders without a wall

The radiation correction for a free single cylinder was

calculated from the following equation:

Qr = e Sb A (Tw4 - Tinf4) (B.1)

where Sb is Stefan-Boltzmann constant and equal to 5.6696 x

10-8 (w/m2.K4) and e is the emissivity = 0.05. The radiation
 197

correction for multiple elements depends on the view factors

between the surfaces as well as on the surface heating/cooling
condition. The surface condition could be at a constant
temperature or constant heat flux. Therefore, equation B.1 is no
longer valid for a free array.
The view factor between the cylinders with the same
diameter was calculated as follows:

Fc_c = [ (X2- 1)112 + sin l (X J X (B.2)

Where X = 1 + (H/D) and H is the surface-to-surface distance

between the cylinders, Howell [BA]. Marsters [B.2] showed that
the radiation corrections for the cylinders of an array at known
surface temperature can be calculated from equation B.3 :

Qr = (1- K Fc_c) e Sb A (Tw4 Tinf.4) (B.3)

where K = 1 if the cylinder is at the edge of the array or K = 2 if

the cylinder is in the interior of the array. This procedure was
implemented in the data reduction program, Appendix C, to
calculate the radiation correction for the cylinders in the no wall
cases.
 198

B.2 Radiation correction for an array with a single wall

This case included three heated cylinders (the array) and an

adiabatic wall. The wall was made of a poor thermal conductivity
material and its back surface was insulated. Owing to the
radiation heat transfer from the cylinders and the high
temperature of the raised plume from the cylinders by natural
convection, the surface of the wall was at a higher temperature
than the ambient. Therefore, the wall cannot be considered as an
adiabatic wall for radiation heat transfer considerations. Since
the wall had poor thermal conductivity, the wall temperature was
not uniform and the temperature along the wall was dependent
on the elevation from each cylinder of the array and the spacing
between the array and the wall (S/D). The closer spacing (S/D)
resulted in higher non-uniform temperature along the wall. To
account for this non-uniform temperature distribution, the wall
was treated as five separate elements. The middle three
elements had a length equal to the cylinders' center-to-center
spacing, CC. These elements were numbered 5, 6, and 7 as shown
in figure B.1. Hereafter these elements will be referred to as a
"wall" with a specific number assigned to each. The temperature
of each wall was recorded during the experiments. Wall numbers
5, 6, and 7 had the highest average temperature. Wall #8 had the
next highest and wall #4 had the lowest temperature. The
surfaces of the walls were black and had an emissivity equal to
 8
10.5'

T
CC
7
4 CC

25' CC
6
$

cc
5
1

Figure B.1 The wall numbers for the view factors.

 200

unity (i.e ek=1.0 for k=4,5,. 6, 7, 8).

To further facilitate the analysis, it is convenient to envision
an enclosure with black interior surfaces by adding surface #9, as
shown in figure B.1. The view factors, which are defined as the
fraction of energy leaving one surface that reaches another
surface , Seigel and Howell [B.3], were calculated by using three
different approaches. Surface shape and position of the surfaces
determined which approach was used.
The first approach, equation B.2, determined the view factor
between the cylinders, Fc..c. The view factors between each
cylinder and the walls (walls #4 - 8) were calculated by using the
crossed-string method. This method was used when the
extended strings between a specific cylinder and a wall element
were bent around the neighboring cylinder. In other words, this
method was used when the viewing of the wall by the cylinder
was obstructed by the cylinder directly above or below the
specified cylinder. This condition is shown in figure B.2 and
equation B.4 illustrates how the view factor was calculated by
using the crossed-string method.

(fg+gc+cd+deh+ha) (ab+bgc+cd+de+ef)
F3_4 (B.4)
2 al D)

When the wall could be viewed by the specified cylinder

without any obstruction by the other cylinder(s), the view factor
was calculated from equation B.5.
 201

Figure B.2 Crossed-String method to calculate F3-4.

 202

1 b1 b2
F1_4 = [tan-1 B1 tan'l B 2 where B1= and. B2= (B .5)
2il 1
a a

The dimensions in this equation are shown in figure B.3.

Equation B.5 yielded the same results as the crossed-string
method and it was used for simplicity in the calculations. The
reciprocal relations for the view factors were used to determine
the view factors from Wall numbers 4, 5, 6, 7, and 8 to the
cylinders. This relationship is illustrated as follows:

A- Fi = A. Fj-j -i

The view factors from the walls numbers 1 through 8 to wall #9

were calculated as follows:

9
Fi_9 = 1 - E Fi_i
j=1

Table B.1 shows a sample of the view factors for the array
with CC = 2D at S/D = 0.5. Once the view factors were
determined, the radiosity values (the outgoing radiant energy per
unit area), qo , were determined from equation B.6, Siegel and
Howell [B.2], by considering the surfaces as a diffuse-gray
enclosure.

9
Z [8k. - (1-ek) Fk_j] qo = ek Sb Tk4 l< k< 9 (B.6)
j=i
 203

T
i
b2

bl

[
Figure B.3 The dimensions of equation B.5 .
 Table B.1 View factors for a single wall case at CC= 2D and S /D= 0.5.

1 4 3 4 5
1 0.00000E+00 8.13758E-02 0.00000E400 1.09888E-01 2.50000E-01
2 8.13758E-02 0.00000E+00 8.13758E-02 1.08139E-02 7.37918E-02
3 0.00000E+00 8.13758E-02 0.00000E+00 2.35292E-03 5.19180E-03
4 3.63393E-02 3.57609E-03 7.78096E-04 0.00000E400 0.00000E+00
5 3.92699E-01 1.15912E-01 13.15526E-03 0.00000E100 0.00000E+00
6 1.15912E-01 3.92699E-01 1.15912E-01 0.00000E400 0.00000E+00
7 8.15526E-03 1.15912E-01 3.92699E-01 0.00000E+00 0.00000E+00
8 7.78096E-04 3.57609E-03 3.63393E-02 0.00000E+00 0.00000E+00
6 7 8 9
1 7.37918E-02 5.19180E-03 2.35292E-03 4.77400E-01
2 2.50000E-01 7.37918E-02 1.08139E-02 4.18037E-01
3 7.37918E-02 2.50000E-01 1.09888E-01 4.77400E-01
4 0.00000E+00 0.00000E100 0.00000E100 9.59307E-01
5 0.00000E+00 0.00000E400 0.00000E+00 4.83234E-01
6 0.00000E400 0.00000Et00 0.00000E+00 3.75477E-01
7 0.00000E400 0.00000(i00 0.00000E400 4.83234E-01
8 0.000001:400 0.00000L)00 0.00000E100 9.59307E-01
 205

where Ski = 1 if k=j or Ski = 0 if k#j

Since surfaces #4 to 9 were black surfaces, the radiosity values of
these surfaces were known, owing to (ek=1 for k=4 to 9). Hence
equation B.6 can be written for l< k< 3 . From this, three
equations were obtained which were solved for q0,1, q0,2, q0,3.
Then, the radiation heat transfer values from the surfaces were
calculated from equation B.7.

9
Qrk = Ak (Clo,k Fk-j Clo,j ) l< k< 9 and l< j< 8 (B.7)

A computer program to calculate the heat transfer was

written. This program is listed at the end of this appendix. A
sample of the results of this program is shown in Table B.2.

B.3 Radiation heat transfer for two wall cases

Each of the side walls were considered as a single element.

This was based on the following:
a. Owing to a higher flow rate between the two walls than
the flow rate for a single wall, the temperature
differences between the measured temperatures along
the walls and the ambient were less in two wall cases
at (S/D)R = 0.5 than that for a single wall at (S/D) = 0.5.
b. The temperature gradients along the wall at S/D > 0.5
were less than the temperature gradient at smaller
 206

Table B.2 Radiation corrections for the array at CC= 2D and S/D= 0.5.

TOTAL RADIATION

--- 0= 1WATT 9/0= .5

1 4.84352E-02 WATT
2 5.19953E-02 WATT
3 5.34983E-02 WATT
0 SUM 123 = .1539288

Q= 3 WATT 9/0= .5
1.17373E-01 WATT
1.26267E-01 WATT
1.28733E-01 WATT
Q SUM 123 = .3723728

-- 0= 10 WATT S/0= .5
1 3.80710E-01 WATT
2 3.93905E-01 WATT
3 3.991E2E-01 WATT
0 SUM 123 = 1.173777

0= 20 WATT 9/0= .5
7.91351E-01 WATT
8.19280E-01 WATT
3 8.37882E-01 WATT
Q SUM 123 = 2.448513

0= 30 WATT S/0= .S
1 1.27777E+00 WATT
1.32216E+00 WATT
3 1.34775E*00 WATT
Q SUM 123 = 3.947678
 207

spacings.
c. The calculations of the radiation heat transfer for a
single wall at S/D = 0.5 (by considering a single side
wall element) were compared to the results of multiple
wall elements. This showed that there was less than a
0.5% difference in the radiation correction from the
multiple wall results.
Figure B.4 shows the numbers of the walls for two wall
cases. Wall #6 was the flow outlet side and its temperature was
considered to be the average of the transverse temperature
readings between the two walls. Wall #7 was the inlet flow side
and its temperature was considered to be the ambient
temperature. The average temperatures for walls # 4 and 5 were
calculated from the average temperature of the wall elements
shown in figure B.1. The temperature of each element was
weighted by its area as shown in the following equation:

8
Z (Ai Ti 8
i=4
Tay = where A=-- SAL and i = wall numbers in
1=4
figure B.1.

The view factors between the cylinders were calculated by

using equation B.1. The crossed-string method was used to
calculate the view factors from cylinders 1, 2, and 3 to the side
walls 4 and 5. F1.7 and F3_6 were determined from equation B.5.
 A' 6

B
L 010 p.

H
K
G

H. 7

Figure B.4 The numbers of the walls for two walls cases.
 209

The view factors between wall 4 and wall 5 were calculated as

follows:

F4_5 =.0 F4.5 + F4_5 + F4_5 + F4_5 (B.8)

through A-B through C-D through E-F through G-H

The spacings A-B, C-D, E-F, and G-H are shown in figure B.4.
The view factors on the right hand side of equation B.8 were
calculated by using the crossed-string method. For walls 6 and 7,
the view factors were calculated as follows:

F6_7 = F6_7 + F6_7 (B.9)

through L-M through N-0

The view factors from cylinder 2 to wall 6 and 7 were

determined as follows:
F2-6 = F2-6 + F22-6
-6
through L-M through N-0

F2_7 = F2_7 + F2_7

through H-I through J-K

By using the reciprocal relation and by considering the fact

that the fraction of energy leaving one surface and reaching the
surfaces of the enclosure must equal unity, all the view factors
were determined. Table B.3 shows the view factors for CC = 2D at
S/D = 0.5 and (S/D)R = 2.00.
Wall numbers 4 through 7 were considered as black surfaces
Table B.3 View factors for two walls case at CC= 2D with S/D= 0.5
and (S/D)R= 2.0 .

VIEW FACTORS
S/D= 0.5 (S/D)R= 2.00

WALL .... 1
--a
..a 4
1 0.00000000+00 8.13758000-02 0.00000000,00 4.43981020-01
2 8.13758000-02 0.00000000i00 8.1375800D-02 4.18020300-01
3 0.00000000+00 8.1375800D-02 0.00000000 +00 4.43981020-01
4 5.5792300D-02 5.2599806D-02 5.57923000-02 0.00000000+00
5 5.2612197D-02 5.14230690702 5.2612187D-02 7.1310886D-01
6 3.28098890-03 4.49531990-03 4.6956411D-02 4.38487610-01
7 4.6956411D-02 4.4953199D-03 3.2809880D-03 4.38487610-01

WALL .... S 6 7
1 4.18674480 -01 3.65529960-03 5.23134150-02
2 4.0921179D-01 5.0081660D-03 5.00816600-03
3 4.18674480 -01 5.23134150-02 3.65529900-03
4 7.13108860-01 6.1380281D-02 6.13882810-02
5 0.00000001)00 6.51218500-02 6.51218500-02
6 4.6515607D-01 0.00000000 i00 4.1623493D-02
7 4.6515607D-01 4.16234930-02 0.00000000+00
 211

(i.e., ek = 1 for k = 4, 5, 6, and 7). Therefore, the radiosity values

of these walls were known and equation B.6 was written for 1 < K
< 3 only. Once the radiosity, q0, of each surface was known, the
radiation heat transfer was calculated from equation B.7 for i< k<
7 and l< j< 7 These calculations were performed by modifying
the computer program for a single wall by changing the wall
numbers. The original single-wall computer program is shown at
the end of this appendix. Table B.4 shows the output results of
the modified program for CC = 2D at S/D = 0.5 and (S/D)R = 2.00.

B.1 Howell, John R., "A Catalog of Radiation Configuration Factors",

McGraw-Hill Book Company, New York.
B.2 Marsters,G.F.; "Arrays of Heated Horizontal Cylinders in
Natural Convection"; Int. J. Heat Mass Transfer, vol.15, 1972.
B.3 Siegel, Robert and Howell, John R.; "Thermal Radiation Heat
Transfer"; 2nd Edition, Mc Graw-Hill Book Company, New
York.
 212

Table B.4 Radiation corrections for the array at CC= 2D and two
walls at S/D= 0.5 and (S/D)R =2.0 .

TOTAL RADIATION

--- 0= 1 WATT S/0= 0.5 (S/O)R= 2.00

1 4.337980-02 WATT
2 5.211380-02 WATT
3 5.323930-02 WATT
4 2.458520-01 WATT
S -3.361760-01 WATT
6 6.209870-02 WATT
7 -7.633740-02 WATT
0 SUM 123 = .149382 0 SUM 4-7 . -.154563

0= 3 WATT S/0= 0.5 (S/O)R= 2.00

1 1.052450-01 WATT
1.245480-01 WATT
3 1.280790-01 WATT
4 3.440620-01 WATT
5 -7.100740-01 WATT
6 1.109930-01 WATT
7 -1.081100-01 WATT
0 SUM 123 = .357871 0 SUM 4-7 = -.363129

0= 10 WATT S/0= 0.5 (S/D)R= 2.00

1 3.214370-01 WATT
2 3.741630-01 WATT
3 3.306150-01 WATT
4 8.929230-01 WATT
S -1.554820+00 WATT
6 -2.259600-02 WATT
7 -3.972620-01 WATT
0 SUM 123 = 1.076215 Q SUM 4-7 = -1.081759

--- 0= 20 WATT S/0= 0.5 (S /0)R= 2.00

1 6.647890-01 WATT
2 8.051140-01 WATT
3 7.987950-01 WATT
4 1.189830+00 WATT
S -2.462100+00 WATT
6 -2.481900-01 WATT
7 -7.542420-01 WATT
0 SUM 123 = 2.268698 0 SUM 4-7 = -2.274709

--- 0= 30 WATT S/0= 0.5 (S/O)R= 2.00

1 1.387270+00
WATT
2 WATT
1.276830+00
3 1.292990+00
WATT
4 1.133670+00
WATT
5 -7.232500+00
WATT
5 -4.7021SO-01
WATT
S472.0000 wATT
:UM 122 - 3.637032 0 SUM 4-7 - -3.88.7S47
 213

Appendix C

DATA REDUCTION PROGRAM

 214

10 ' ********** Program (CALRONUP.TRA) to calculate the radiation and

20 ' ********** Nuselt number for the three cylinders and printout the results.

30 REM

40 CORRECT THE FILE NAMES IN LINE 390, 540 RHO LINE 940
50 OEFOOL A-Z

60 aErsms I,K,I

70 DIM TU(9,9,3),TA(9,9),TR(9,9,3),OR(9,9,3),Q0(9,9,3),00(9,9,3),0(9),US(9)

80 DIM TEN0(9,9,3),PR(9,3,3),6RM(9,9,3),R111(9,9,3),NUEX(9,9,3)

90 OIM TUAU(9,9,3),*(9,9,3)
100 REM

110 REM ****************** INPUT DATA ******************************A****

120 REM

130 PRINT "CYLINDER LENGTH, CL=';:INPUT CL

140 PRINT "CYLINDER DIAMETER, 0=';:INPUT 0

150 PRINT 'CYLINDER CENTER-TO-CENTER SPACING, tCS='; : INPUT CCS

160 REM

170 R=0/2! : P1=3.141592651 : REM R=CYLINDER RADIUS.

180 EH ************IHHI**********.INHI******IH111-1HEIHt***114iHHHHHHHHHHHHHI*******

190 REM

200 ' **** CALCULATE CYLINDER TO CYLINDER UIEU FACTOR, FCC, FOR NO WILL CASES:
210 REM

220 CCIAL/R : REM CEL= NORMALIZED CYLINDER LENGTH.

230 CR=CCS/R : REM CR. NORMALIZED CYLINDER CENTER TO CENTER SPACING.

240 SCR=1!/CR : SSCR=SCR/(1!-SCR"2)".5 :ASCR=ATN(SSCR)

250 CCX=((((CR"2-1!)".5 - (PI/2!))/ASCR)+1!)".5

260 CCY=CCL"2!-CCX"2!+1!

270 CCZ=CCL"2!+CCX"2!-1!

280 REM

290 SUX=1!/CCX : SSCCX=SCCX/(1!-SCW2)".5 : ASCEMIN(SSECX)

300 YOZ=CCY/CCZ : CYOZ=Y02/(1!-Y02-2)".5 : ACYOZ=(PI/2!)-ATN(CYOZ)

310 CR2=2!/CR : CCR2=CR2/(1!-CR2"2!)".5 : ACCIT=(P1/2!)-ATN(CCR2)

320 CYX2=CCY/(CCX*CCZ) : CCYX2=CYX2/(1!-CYX2"2!)".5 : ACYX2=(PI/2!)-RTN(CCYX2)

330 CCA.(1!/(2!*PI))*(tCR"2-4!)".5 - CR+ PI- 24ACCR)

340 REM

350 CCI=1!-(1!/PI)*( ACYOZ (1!/(2!*CCL)) *MCCY+2!*CCX"2)"21-(2!*CCX)^2)".5*

ACM + CCY*ASCCX (PI/2!)*CCZ) )

360 FCC=CCA*CCI : FCK=1!-FCC : FCK2=1!-(2!*FCC)

370 ' *-11********111HHf*****MHOHHHHHHHHHHHHHHHHHHHO*1141141.*******IHHHHH1*******

380 ' Read the data from the file, RUSCC1.011T, in drive C:

390 OPEN "RUSCCII.OAT" FOR INPUT AS 11

400 INPUT 111,INO,O,L,KUS

410 '

420 FOR K0=1 TO INO

430 INPUT 111,0(1(0)

440 FOR KS=1 TO KUS

450 INPUT 11,16(KS),TA(KILKS)

460 FOR KC=1 TO 3

470 INPUT 11,TU(KO,KS,KC),TUAU(KO,KS,KC),TENO(KO,KS,KC)

480 NEXT KC
 215

490 NEXT KS
SOD NEXT KO

510 CLOSER

520 '

530 ' Read the data from the file, ORUSCCIAT, in drive C:

540 OPEN "ORUSCCII.ORT* FOR INPUT RS 11

550 INPUT 11,INO;04,KUS

560 FOR KO=1 TO INO

570 INPUT 11,0(KO)

580 FOR KS=1 TO gUS-1

590 INPUT 11,US(KS),OR(KO,K6,1),OR(KO,K6,2),CIR(KO,KS,3)

600 NEXT KS

610 NEXT KO

620 CLOSEll

630

640 ' 14*********************4HHHI***********WCHMKOHHHIMI************WI

650 ' Calculate the Husselt :umber:

660 6=1270000001 : ' 6= gravity in m/hr2

670 8= PI*0 *1 R=cylinder surface area in sq. meter.

680 RS=(0/2!)"2*PI : ' RS= Cylinder cross section area in 112 .

690 FOR K0=1 TO INO

700 FOR KS=1 TO (US

710 FOR KC=1 TO 3

720 '

730 If KSOKUS THEN 6010 760

740 IF KC=2 THEN FK=FCK2 ELSE FK=FCK

750 OR(10,(6,(C)=7.900001E-02*.000000056991*FPRWTU(KO,KS,(C)+273.15)"4-(INKO

,KS)4273.15)"4)

760 ' OR() = Radiation from the cylinder.

??0 OCO(K4,K5,(C)=RS*.000551*418.68*(TENO(MKS,KC)/.007366)*21

780 ' 000= Condection heat loss in Uatt.

790 4CU(KO,KS,KC)=0(K0)-0R(K0,16,KC)-0C0(1(0,6,KC)

800 OPR=OCU(KU,KS,KC)/R

810 TT=(TU(KO,KS,KC)+TR(KO,KS))/2!

820 TR(KO,KS,KC)=TT

830 TRX=TR(KOJS)+273.15
840 ' Go to the subroutine to calculate the air properties at temp.= IRK

850 GOSH 1160

860 NUEX(KO,KS,C)=OPR*0/(CX*(TU(KO,KS,KC)-TR(KO,KS)))

270 6R11(KOJS,KC). R0"244*OPR*0"4/(CK*11"2)

880 PR(KO,KS,KC)=CP*M/CK

890 RRM(KO,KS,KC)=601(KO,KS,KC)*PR(KO,KS,KC)

900 '

910 NEXT KC

920 NEXT KS

930 NEXT KO

940 OPEN "RUSCC11.CRL" FOR OUTPUT RS 11

950 URITE 11,CCS,INO,O,L,KUS

 216

960 FOR K0.1 TO INQ

970 URITE 11,0(KO)

980 rap (5.1 TO KUS

990 URITE 1145(6),TA(K0,0)

1000 FOR KC :1 TO 3

1010 URITE 11,TU(KO,KS,KC),NOEX(KO,K5,KC),R8M(KO,KS,KC),6RM(KO,KS,KC),PR(KO,KS,K

C)

1020 NEXT KC

1030 NEXT KS

1040 NEXT KO

1050 CLOSEll

1060 PRINT : PRINT .**** Uhat do you want to print:*

1070 PRINT ' 1- detailed output. "
1080 PRINT 2- End the program. "

1090 PRINT Put your 0015t 1'; :INPUT PCHO

1100 If PCH0.1 6010 605U8 1440 : 6010 1060

1110 IF PCHOO1 RHO PCH002 6010 1060

1120 '

1130 END

1140 '

1150 '

1160 ROl ""114444*** RIRPROP.TRA "**************/******4".**********

1170 ' This subroutine calculates the air properties at different
1180 temperatures, TT .

1190 '

1200 '

1210 ' Oefine function to calculate the thermal conductivity, FNIK(U,P1):

1220 OEF fliffi(XK,P1) .((.0000063251+4".5)/(1!+(245!/XIO*10"P1))*(360!/.860421)

1230 ' P1.-12.0.1K ;XK= temp. in K ; FNFK(XR,P1): URTT/METER.K

1240 '

1250 ' Oefine function to calculate the specific heat, FNFC(XK):

1260 OEF FN1C(2) . (.249679 .00007551791*U + .0000001691941*XK"2 .00000000

006461261*Xr3)*1.1631

1270 ' FNFC(XK)= specific heat in (U-Nr/Kg.K) for temperature 7 260 to 610 k.
1280 '

1290 ' Oefine function to calculate the viscosity, FNFM(XK):

1300 OEF DIr11(a).I45.8 .n.l.5)mxim10.4).100000001».3601

1310 ' FNFM(XK). giscosity In KgiM.hr ; XK. Temp. in K

1320 '

1330 'Oefine function to calculate the density , INFRO(XK):

1310 OEF FNFRO(XK). .46458*760!AK :' FNFRO(XK).0ensity in (Kg /n3) XK.Temp.K
1350 '

1360 TK-TT273.1S :
'TT.Temperature in degree centigrade; XK. Temp. in Kelvin.
1370 P1=-12±/11 : CK=FMRCIK,P1) : ' CPthermal conductivity (U /n.K)
1380 CP.FNE(TIO : ' CP.specific heat (U-hr/Kg.C)
 217

1390 ?1=FNOI(TK) : ' q :viscosity (Kg/m.hr)

1400 RD.FNFRO(TK) : ' RD:density (Kg/T13)

1410 8:11ifAK : ' 8 :thermal expansion (1/K)

1420 '

1430 RETURN

1440 .***44144 To print out the results in denies : 11************************

1450 LPRINT ALL TEMPERATURE, 1U, (C) CC/0 = ';CU:

1460 LPRINT "

1470 FOR KO:1 TO INO

1480 LPRINT ..."*" 0 .*:0(K0):'UATI'

1490 LPRINT US/0 CYLINDER fl CYLINDER 12 CYLINDER 13 "
1500 FOR K5.1 TO KUS

1510 IF K5.415 THEN 6010 1540

1520 LPRINT USN ' 11.111 1111111.1111 1111111.1111 1111111.1111'

:i4(16):111(10.KS,1):IU(KILKS.2):TU(10,KS,3)

1530 6010 1550

1540 MINT 0S1N6 NO UAL 1111111.1111 1:11141.11111 11411111.1111*

;TU(KO,K5,1):TU(10,K5,2):TU(KO,KS,3)

15E0 NEXT KS

1560 LPRINT

1570 NEXT KO

1580 LPRINT CHRS(12)

1550 LPRINT REFERENCE TEMPERATURE, TR, (C) CC/0 = ';CCS;'

1600

1610 FOR <0:1 10 INO

1620 LPRINT 4.*:+**4 0 .":Q(KO)CUATT'

:630 LPRINT ' USA CYLINDER 11 CYLINDER :2 CYLINDER 13-'

1640 FOR KS:1 TO 016

16E0 IF KS.(US THEN 6010 1680

1660 LPRINT USING ' 11.111 1111111.1111 1111111.1111 1111111.1111'

:4(K5):Ti(KOJS.1);TR(KO,K5,2);IR(KO,KS,3)

1670 0010 :690

1680 LPRINT USING NO UALL 1111111.1111 1111111.1111 1111111.1111'

TR(KO,KS,D; T9(10,KE,D
1690 NEXT KS

1700 !PRINT "

1710 NEXT KO

1720 LPRINT CHPS(12)

1730 LPRINT PMDIENT TEMPERATURE, TA, (C) CC/0 = ';CM"

1740 LPRINT

1750 FOR K0.1 TO INO

1760 LPPI41 ' "41'4*** .":0(K0):"411*

:770 LPRINT QS/0 CYLINOER 11 CYLINDER :2 CYLINDER 13

1780 FOR K5.1 TO KUS

:790 y KS:KU:3 THEN 3010 1820

1800 LPRINT 0SING muallot Immull

:USJS':TP.:KCAS):78(KO,KS):1A(K0,(5';

1510 6010 1630

 218

1820 LPRINT USING NO WALL 1111111.1111 4111411.1111 4111111.1111"

:IR(KO,KS);TA(KO,KS):18(KO4S)
1830 NEXT KS

1840 LPRINT "

1860 NEXT KO

1860 LPRINT CHR$(12)

1870 LPRINT ------ RADIATION HEAT LOSS, OR, (WATT) CC/0 = ';CCS:.
---------

1880 LPRINT "

1890 FOR K0.1 TO INO

1900 !PRINT ******* Q e:0(KO) ;'UATT'

1910 ANT ' US/0 CYLINDER 11 CYLINDER 12 CYLINDER 13
1920 FOR KS=1 TO (WS

1930 IF KS=KUS THEN 6010 1960

1940 LPRINT USING ' 44.x44 1111111.1111 1141114.1411 1141414.1411"

;U5(KS);OR(KO,KS,1);OR(KO,KS,2);OR(KO,KS,3)
1950 60T0 1970

1960 !PR1NT USING NO WALL 1114434.1111 112814.2141 141483.1111"

:OR(KO,KS,1):0R(KO,KS,2);OR(KO,KS,3)
1970 NEXT KS

1980 LPRINT "

1990 NEXT KQ

2000 LPRINT CHR$(12)

2010 MINT CONVECTION HEAT LOSS, OCO, WATT) CC/0 . ';CCS;'

1Q20 LPPIKT

2030 FOR KO=1 10 INO

:040 LPRINT 4******* 0 =';2(K0);"WAIT'

2050 LPRINT US/0 CYLINDER 11 CYLINDER 12 CYLINDER 13
2060 FOR KS =1 10 KUS

2070 IF KS.KUS THEN 6070 2100

2080 LPRINT USING 11.313 1111111.1111 4144141.1111 3111141.1141'

:WS(KS);OCD(KO,KS,1):00(KO,KS.2);OCD(KO,KS,3)
2090 6010 2110

2100 LPRINT USING NO WALL 411:111.1111 1111111.1111 111811.1111"

;OCO(K0.1( 1 .QCD(KO,KS,2):0CD(KO,KS,3)
2110 NEXT KS

2120 LPRINT "

2130 NEXT KO

2140 LPRINT CHR$(12)

2150 LPRINT NET CONVECTION HEAT TRANSFER, ICU, (WAIT) CC/0 . *;CC
S;'

2160 LPRINT ""

2170 FOR K0=1 TO INO

2180 LPRINT 01.****** 0 =';NOWURIT'

:190 LPRINT US/0 CYLINDER 11 CYLINDER 12 CYLINDER 13
1:00 FIR 6.1 10 KUS

2210 IF KS=KUS THEN 6010 2240

1220 !PRINT USING 11.111 1211411 4111141.1111 111184.4111"

;US(KS::OCU(KO,K5,1) ;ICU(KO,KS.2);OCU(KUS,3)
2230 GOO 2250
2240 LPRINT USING ' NO WALL 1111113.811 344«117.1111 111388.3311"
;OCLI(K0AS,1):OCV(KO.KS,2';ICI(KO.KS,3)
 219

2250 NEXT KS

2260 LPRINT

2270 NEXT KO

2280 LPRINT OHRS(12)

2290 LPRINT RAWL NUMBER, PR CC/0 = ";CCS;"

2300 LPRINT "

2310 FOR <0.1 TO INO

2320 LPRINT " ******** Q ";Q(KO);'URIT.

2330 LPRINT US/0 CYLINDER 11 CYLINDER 12 CYLINDER 13

2340 FOR KS.1 TO KUS

2350 1r KS.KUS THEN 6010 2380

2360 LPRINT USING t1.111 1111111.1111 1111114.1111 1111111.1111'

;US(KS)0( K0,KS,1);PR(KO,KS,2);PR(KO,KS,3)

2370 6010 2390

2380 LPRINT USING HO UAL 11:7111.1111 4111141.4811 1111111.1111'

;PR(KO,KS,1);PEKO,KE,2)0(KO,KE,3)

2390 NEXT KS

2400 LPRINT "

2410 NEXT KO

2420 LPRINT CHRS(12)

2430 LPRINT flODEFIED GRRSNOF NUMBER, GRM CC/0 . ';CCS;'

2440 LPRINT

2450 FOR K0.1 TO IHO

2460 LPRINT 4***"44 0 .';0(KO) ;"URIT'

2470 LPRINT US/0 CYLINDER 11 CYLINDER 12 CYLINDER 13

2480 FOR KS.1 TO KUS

2490 If KS.KUS THEN 6010 2520

2500 LPRINT USING 71.111 1111110.1111 11:101:1111 24:4411.1:11*

WS(KS:6871(KO.KS,1):6RM(KO,KS,2);6R1(KO,KE,3)

2510 60T0 2E30

2E20 LPRINT USING NO URLL 1111111.1111 141:111111# 1111:111.11:1'

6811(KO.KS.1);61M(KO,KS,2):601(KO,KS,3)

2530 NEXT KS

2540 LPRINT "

2550 NEXT KO

2560 LPRINT CHR$(12)

2570 LP;INT MODEFIED RAYLEIGH NUMBER, RAM CC/0 = ";CCS;"

2580 LPRINT ""

2530 FOR K0=1 TO IMO

2600 LPRINT ******** Q =';0(KO)CUATT"

2610 LPRINT US/0 CYLINDER 11 CYLINDER 02 CYLINDER 13

2620 FOR KE.1 TO KUS

2630 IF KS=KUS THEN 6010 2660

2640 LPRINT USING 11.111 1111111.1111 1114111.1111 1141111.1111"

;USCKS);RAM(KO,KS,1);RACK04.2);RAM(KO.K8,3)
2630 60TO 2670

2660 OW USING NO ALL 11:1311.41:4 4411404.1141 :144144.4:Ir

:RAM(KO.KE.1:,:iRM(K0.KS,2):RAM(KO.KS,3)

2670 ND! KS
 220

LPRINT "
2690 NEXT KO

2700 LPRINT CHRS(12)

2710 LPRINT ' --------- CALCULATED NUSSELT NUMBER, NUEX=0.0/(A.K.OT), CC/0 =

';CCS;"

2720 LPRINT "

2730 FOR 0=1 10 IND

2740 LPRINT ******* Q =';0(KO);"URIT'

2750 LPRINT ' US/0 CYLINDER 11 CYLINDER 12 CYLINDER 13
2760 FOR KS=1 TO KUS

2770 IF KS=KUS THEN 60T0 2800

2780 LPRINT USING ' 11.111 1111111.1111 1111114.1111 1111111.1111'

;US(KS);NDEX(KO,KS,1):NUEX(KO,KS,2);NNEX(KO,KS,3)
2790 60T0 2810

2800 LPRINT USIN6 ' NO al 1111111.1111 1111111.1111 1111111.1111"

;NUEX(KO,KS,1):NUEX(KO,KS,2);NUEX(K0,0,3)
2810 NEXT KS

2820 LPRINT "

2830 NEXT KO

2840 LPRINT CHR$(12)

289 RETURN
 221

APPENDIX D

UNCERTAINTY ANALYSIS

This section will address the uncertainty in the experimental

values of the average Nusselt number for the cylinders. The
uncertainty in an experimental result, 91, where 91 = f(X1,X2, . .

,Xn), was evaluated by using the following equation, Holman [D.1]:

2 21 1/2
891
°31 T72 c°2] ± "4 83Cn (1111]
(D.1)

where X1,X2, . ,Xn are the independent variables of the

experimental result 91 , and (01, (02, . ,con are their uncertainty
values, respectively.
To calculate the uncertainty in the average Nusselt number,
Nu, of the cylinders, the definition of Nu , as shown in equation
3.2, should be considered.

QD
Nu =
AK -Tind
V2
Where A= II D L and Q=IV= R1

By assuming AT= Tw - Tinf , The average Nusselt number

becomes
 222

V2
Nu f(V,R 1 ,L , AT ,K) (D.2)
(R1 L II AT K)

Therfore, the uncertainty in the average Nusselt number,

c°Nu , can be calculated from equation (D.1) as follows

8Nu 8Nu 2 8u -2
+
°Nu F
L 8y _I L 8R1 8L
1/2
[ 8Nu ]2 8N11 12 (D.3)
WK
SK

The partial derivative terms in this equation were evaluated

from equation (D.2) as shown below :
SNu 2V
8V R1 L II AT K
8Nu V2
SRI R12L II AT K
8Nu V2
81- R1 L2 II AT K
SNu V2
SOT R1 L II AT2 K
8Nu V2
81( R1 L AT K2

Substituting these terms in equation (D.3), the uncertainty in Nu

is
 223

v2 2 2
1 2
1
R1 L lI AT k CV °)v [R1 c°R1 L (13/- +

1 -12 coKi2
L AT .1 L k j
or
2 2
C° Nu [7-1
CON/ [ 11 coR1 ÷
Nu R 1'
1/2
[ IT
MTA
2

[ k coK (D.4)

Considering the following conditions :

V =23.0 Volts ± 0.01%

R1 = 1.0 Ohm + 5.00%
AT = 100.0 °C + 0.5%
K = 0.029181 W/m.k ± 0.5%
L =0.254 m + 0,01%

the uncertainties in these independent variables were,

cov = 0.023 Volt
coRi= 0.05 Ohm
coA = 0.5 °C
z-xT

coK= 0.0001459 W/m.K

coL= 0.000025
 224

Substituting for V, R1, L, and K and their uncertainty values in

equation (D.4), the uncertainty percentage was

coNu
Nu = 5.0 %

D.1 Holman, J. P., " Experimental Methods for Engineers", 4th

Edition, McGraw-Hill Book Company, New York.