INTRODUCTION
Some studies have shown that the tube shapes and their arrangement in heat
exchangers have positive influence on heat transfer [13]. The effect of flow past
bluff bodies, especially cylinders, has been a major attraction for fluid mechanics
investigators for a long time. Most of these studies were concerned with the flow over
a circular cylinder. Williamson [4] and Zdrakovich [5] wrote comprehensive reviews
Received 2 October 2005; accepted 7 December 2005.
Support for this research by King Fahd University of Petroleum and Minerals is gratefully
appreciated.
Address correspondence to N. K. Anand, Department of Mechanical Engineering, Texas A&M
University, College Station, TX 77843-3123, USA. E-mail: nkanand@tamu.edu
97
98
H. M. S. BAHAIDARAH ET AL.
NOMENCLATURE
A
cp
D
f
f0
H
k
l
L
_
m
Nu
Nu0
Nu
Nu
p
Re
T
Tm
Tw
un , ug
Uin
!
V
y
Dp
Dp
DTb
n
n, g
q
r
pressure
Reynolds number ( Uin H=n)
temperature
average of module inlet bulk
temperature and module outlet bulk
temperature
temperature of heat transfer surfaces
contravariant components of velocity
velocity at the channel inlet
velocity field
vertical distance from the channel
bottom
pressure drop across a module
normalized pressure drop ( Dp=qUin2 )
change in bulk temperature across a
module
kinematic viscosity
curvilinear coordinates
density
del operator
on this topic. Other researchers considered many different tube shapes, such as rectangular, circular, elliptical, diamond, and flat. Kundu et al. [6, 7] and Kundu [8]
studied heat transfer and fluid flow numerically and experimentally over a row of
in-line cylinders placed between two parallel plates. Incompressible two-dimensional
laminar flow was considered. In general, the pressure drop and heat transfer were
spatially periodic, indicating periodically fully developed characteristics. Numerical
solutions were obtained by Grannis and Sparrow [9] for the fluid flow in a heat
exchanger consisting of an array of diamond-shaped pin-fins. The solutions were
based on the periodically fully developed regime. The solution domain was discretized by subdividing it into two-dimensional, nine-noded quadrilateral elements to
implement the finite-element method. Chen et al. [10, 11] studied the effect of
Reynolds number on flow and conjugate heat transfer in a high-performance finned
oval tube heat exchanger element for a thermally and hydrodynamically developing
three-dimensional laminar flow. Computations were performed with a finite-volume
method based on the SIMLPEC algorithm for pressure correction. Breuer et al. [12]
investigated in detail the confined flow around a cylinder of square cross section
mounted inside a plane channel by two entirely different numerical techniques, lattice-Boltzmann automata and the finite-volume method. Both numerical methods
are second-order-accurate in space and time. Flat tube designs have been recently
introduced for use in modern heat exchanger applications such as automotive radiators [13]. Availability of heat transfer and pressure drop characteristics of flat tube
heat exchangers in open literature is rather limited. Bahaidarah [14] and Bahaidarah
et al. [15] have studied these characteristics numerically for various tube layouts.
Recently numerical analysis was conducted to study the laminar two-dimensional
steady forced convection of air crossflow over square and nonsquare in-line circular
tube arrangements [16]. A finite-volume method with a nonorthognal body-fitted
99
Figure 1. Flow domains studied in this work, showing five in-line tubes in a channel: (a) circular crosssection tubes; (b) flat-cross-section tubes; (c) oval-cross-section tubes; (d) diamond-cross-section tubes.
grid and collocated variables was used to solve the Navier-Stokes and energy conservation equations for a tube bundle with five longitudinal rows, including inlet and
outlet sections. Comparison of the results with well-established experimental data
and empirical correlations [17, 18] showed good overall agreement.
A single geometric tube shape has been studied for fluid flow and heat transfer
by many researchers in the past. In this article we study heat transfer over four different geometries for the purpose of comparison. Consideration is given to 2-D
steady laminar incompressible flow over noncircular (flat, oval, and diamond) and
circular in-line tubes confined in a channel (Figure 1). The four different geometries
studied are circular, flat, oval, and diamond, as shown in Figure 2.
MATHEMATICAL FORMULATION
The governing mass, momentum, and energy conservation equations for steady
incompressible flow of a Newtonian fluid with constant thermophysical properties,
no heat generation, and negligible viscous dissipation are
Mass conservation :
!
r V 0
100
H. M. S. BAHAIDARAH ET AL.
Figure 2. Geometric modules for various tube cross sections (shaded area in each module is the calculation
domain): (a) circular; (b) flat; (c) oval; (d) diamond.
Momentum conservation :
Energy conservation :
!
!
!
V rV rp nr2 V
!
qcp V rT kr2 T
101
secondary flux is zero. The secondary flux is treated explicitly to avoid nine-point
formulation. The power law is used to represent the solution as a one-dimensional
convection-diffusion equation. The contravariant components of velocity, un and
ug , are used as the dependent variables. In this formulation, the discretized momentum equations are expressed in terms of parallel velocities instead of actual velocities.
The velocity and pressure fields are linked by the Semi-Implicit Method for Pressure
Linked Equations (SIMPLE) algorithm. The resulting set of discretization equations
is solved iteratively using a line-by-line procedure which is a combination of the
tri-diagonal matrix algorithm (TDMA) and the Gauss-Seidel scheme.
Mathematical details of the finite-volume technique used in this study can be
found in Bahaidarah [14], Bahaidarah et al. [15], and Karki [20], and are not
repeated here.
GEOMETRIC CONFIGURATIONS
Four different geometric shapes of tube cross section were investigated
circular, flat, oval, and diamond (Figure 1). Calculations were made for five in-line
tubes in a channel. The longitudinal distance between two consecutive tubes was
kept constant. Tube modules were defined as shown in Figure 1. A tube module contained one tube and was bounded by two fictitious surfaces located midway between
centers of the two consecutive tubes. The ratio of area of cross section of the channel
to the blockage area of the tube (H=D) was kept constant for all geometric shapes of
tube cross section considered in this study. The case of circular-cross-section tubes
was taken as a reference. Geometric similarity was ensured for the other three cases
by taking constant values for the ratio of module length to tube height normal to the
flow, the ratio of channel height to tube height, and the ratio of tube width in the
flow direction to the tube height. Unobstructed length of the channel upstream of
the first module was equal to one module length. To ensure a fully developed flow
at the channel exit, an unobstructed channel length equal to three modules length
was included downstream of the last module. The tubes and the channel walls were
assumed to be of infinite extent in the direction perpendicular to the paper, so that
the flow could be considered as two-dimensional. Owing to the symmetric behavior
of the laminar flow in the Reynolds number range investigated, only the lower half
of the channel was used as the calculation domain, as shown in Figure 2.
GRID GENERATION
In each configuration, the domain was discretized into a structured grid. The
computational domain was divided into three individual regions. These regions were
the entry region, the tube modules, and the exit region (Figure 1). A uniform orthogonal grid was used for both entry and exit regions. However, to handle the arbitrarily
shaped domain of the tube modules, the Geometry and Mesh Building Intelligent
Toolkit (GAMBIT) was used. The irregular physical domain was discretized into
numerous quadrilateral grid elements and a body-fitted coordinate system was generated. Typical single-module grids for all four geometries considered in this study
are illustrated in Figure 3. The grid distribution for a single tube module can be
102
H. M. S. BAHAIDARAH ET AL.
Figure 3. Grids for the geometries studied: (a) circular-cross-section tube; (b) flat-cross-section tube;
(c) oval-cross-section tube; (d) diamond-cross-section tube.
repeated successively to generate the domain of any number of tube modules. In this
study, five consecutive tubes were included in the computational domain.
BOUNDARY CONDITIONS
In Figure 3, the surface BC is the bottom wall of the tube and the surface EF
represents the lower wall of the channel. A no-slip boundary condition was assigned
for these surfaces, where both velocity components were set to zero at that boundary
(u v 0). The tubes and the channel walls were considered to be at the same
constant temperature (T Tw, Dirichlet condition). AB and CD are the lines of
symmetry where no flow crosses these boundaries, and the normal component of
velocities and the normal gradient of the parallel component of velocity are set to
zero. Finally, the lines AF and DE are recognized as the module inlet and module
outlet, respectively. A uniform inlet velocity profile (u Uin) and a constant inlet
temperature (T Tin), different from the walls temperature, were assigned at the
channel inlet. At the exit, fully developed flow boundary condition was applied.
The streamwise gradients of all variables were set to zero at the outlet boundary.
CONVERGENCE CRITERIA
In the momentum and energy equations, the sum of relative residuals for all the
nodes was monitored for the variables un , ug , and T. For the pressure equation, it is
appropriate to check for mass imbalance in the continuity equation. Convergence
was declared when the normalized residuals of mass, momentum, and energy became
less than 10 6 for all cases considered in this study.
103
VALIDATION
The developed code was validated by reproducing solutions for some of the
benchmark problems.
The fluid flow and heat transfer in a parallel-plate channel subjected to constant wall temperature was predicted. The Nusselt number for the fully developed
region between two parallel plates subjected to constant wall temperature using
the developed code is 7.56, which agrees favorably with the Nusselt number value
of 7.54 reported by many authors such as Incropera and DeWitt [21].
Normalized pressure drop (Dp ) was computed for the third module of a
parallel-plate channel with circular tubes using the developed code and compared
with the numerical predictions of Kundu et al. [6] and the experimental data of
Kundu [8]. It is evident from Figure 4 that there is a very good agreement between
the results obtained by the developed code and results in the literature [6, 8]. The
average heat transfer coefficient and the corresponding Nusselt number across the
modules were also computed for the geometric parameter values L=D 3 and
H=D 2. Table 1 shows the average Nusselt number (Nu) at two different Reynolds
numbers. As shown in Table 1, the numerical predictions of heat transfer using the
104
H. M. S. BAHAIDARAH ET AL.
Table 1. Validation: module average Nusselt number for L=D 3 and H=D 2
Re 50
Kundu et al. [6]
Present work
Re 200
Kundu et al. [6]
Present work
Second module
Third module
Fourth module
9.4
9.23
9.4
9.23
9.8
9.23
12.5
12.44
12.6
12.43
12.8
12.42
developed code agree very well with the numerical predictions of Kundu et al. [6],
thus further validating our code.
GRID INDEPENDENCE
The grid independence test was performed for each configuration for several
grid sizes at the highest Reynolds number considered in this study (i.e., Re 350).
Table 2. Grid independence study for module 3 for Re 350, Pr 0.7
Grid
Circular tubes
35 57
40 125
50 160
59 188
70 200
80 212
Flat tubes
35 57
40 125
50 160
59 188
70 200
80 212
Oval tubes
35 57
40 125
50 160
59 188
70 200
80 212
Diamond tubes
35 58
40 126
50 160
59 188
70 200
80 212
Number
of
nodes
Normalized
pressure
drop (Dp )
Relative %
difference
Module average
Nusselt
number (Nu)
Relative %
difference
1,995
5,000
8,000
11,092
14,000
16,960
1.244
1.163
1.143
1.132
1.126
1.120
6.451
1.755
0.938
0.584
0.488
14.621
14.169
13.999
13.915
13.853
13.803
3.092
1.198
0.602
0.446
0.356
1,995
5,000
8,000
11,092
14,000
16,960
1.455
1.358
1.335
1.324
1.315
1.310
6.630
1.761
0.809
0.630
0.437
16.713
16.128
15.932
15.837
15.769
15.713
3.497
1.214
0.596
0.434
0.354
1,995
5,000
8,000
11,092
14,000
16,960
1.258
1.184
1.166
1.157
1.151
1.157
5.887
1.541
0.732
0.495
0.497
15.621
15.153
14.993
14.910
14.855
14.910
2.997
1.059
0.550
0.373
0.375
1,995
5,000
8,000
11,092
14,000
16,960
1.266
1.190
1.169
1.160
6.052
1.759
0.765
The solution did not
13.821
13.476
13.333
13.265
converge.
2.499
1.061
0.510
105
Figure 5. Grid independence studyvariation of relative percent difference in normalized pressure drop
and module average Nusselt number with grid density for module 3: (a) circular tubes; (b) flat tubes;
(c) oval tubes; (d) diamond tubes.
The grids used for grid independence study were 40 125, 50 160, 59 188,
70 200, and 80 212 with 5,000, 8,000, 11,092, 14,000, and 16,960 nodes, respectively. The grid was refined successively by adding about 3,000 nodes at each step.
Relative percentage changes in normalized pressure drop and module average
Nusselt number for module 3 were monitored as the grid was refined successively.
The results are presented in Table 2 and plotted in Figure 5. It is evident from
Table 2 and Figure 5 that the relative percentage change in normalized pressure drop
and module average Nusselt number between two successive grid sizes is insignificant
as the grid is refined beyond 59 188 (bolded in Table 2). Therefore, to optimize
CPU resources with an acceptable level of accuracy, all parametric runs were made
with the 59 188 grid (11,092 nodes).
RESULTS AND DISCUSSION
The code was run for different values of Re in the range 25350. The value of
Prandtl number in this study was fixed at 0.7. Only representative cases are chosen
for discussion in this article.
106
H. M. S. BAHAIDARAH ET AL.
Figure 6. Streamlines over circular-cross-section tubes: (a) Re 25; (b) Re 150; (c) Re 350.
Onset of Recirculation
For any given tube cross section, no flow separation was observed below a certain value of Re. As Re increases, the flow separation occurs downstream of the
tubes and a recirculation region develops behind every tube. Circular tubes displayed
recirculation at the lowest Re value (Re 32). Flat-cross-section tubes showed no
separation up to Re values just under 35. Flow across the oval tubes started separating downstream of the tubes at Re 45. Diamond tubes were the most gentle to the
flow and showed no separation up to Re values just under 55. As shown by the
stream-function contour plots in Figures 69, for all tube cross sections the size of
recirculation region increases along the flow direction with increase in Re.
Figure 7. Streamlines over flat-cross-section tubes: (a) Re 25; (b) Re 150; (c) Re 350.
107
Figure 8. Streamlines over oval-cross-section tubes: (a) Re 25; (b) Re 150; (c) Re 350.
recirculation region downstream of the last tube is considerably different from that
of the recirculation region downstream of the other tubes (Figures 69). These figures confirm the presence of strong elliptic effects.
Figure 9. Streamlines over diamond-cross-section tubes: (a) Re 25; (b) Re 150; (c) Re 350.
108
H. M. S. BAHAIDARAH ET AL.
drop and module average Nusselt number for each module (Table 3) confirm the
existence of periodically fully developed flow downstream of the second module.
109
greater than 1 for both the flat and oval tubes over the entire Reynolds number range
studied. Flat and oval tubes, therefore, offer more flow resistance when compared to
circular tubes. A heat exchanger with flat or oval tubes will cause more pressure
drop, hence require more pumping power, than one with circular tubes. If the use
of flat or oval tubes is dictated by some other considerations, then the heat exchanger should be designed for high fluid velocities, as the ratio Dp =Dp0 is low at high
Reynolds numbers. However, going beyond Re 200 does not help substantially, as
the Dp =Dp0 curves become nearly horizontal. For diamond tubes, the ratio Dp =Dp0
increases with Re, as opposed to flat and oval tubes. This result can be explained
with the help of Figure 13, which shows that Dp for diamond tube module decreases
with Re at a lower rate compared to that for circular tubes. For Re < 250 the ratio
Dp =Dp0 for diamond tubes is less than unity (Figure 14). Hence, if pumping power is
a concern, diamond tubes should be preferred over circular tubes only in the low
Reynolds number range. In the Re > 250 range, circular tubes are better than diamond tubes from the viewpoint of pressure drop.
110
H. M. S. BAHAIDARAH ET AL.
Module average Nusselt number curves tend to merge at very low Reynolds
numbers. Below Re 50, module average Nusselt number increases rapidly with
Re, with almost a linear trend. The curves are again almost linear for Re > 150,
but the slope is much less than that for Re < 50. Therefore, if heat transfer rate is
a major design consideration, fluid velocity should be maximized in the low-Re
range (i.e., Re < 50), because heat transfer rate is very sensitive to the fluid velocity
in this range regardless of the tube shape. However, in the high-Re range (i.e.,
Re > 150), a better tube cross section should be chosen instead of increasing the fluid
velocity, in order to maximize heat transfer rate. The rapid increase in module average Nusselt number for Re < 50 can be explained as follows. At low Re values, there
is no recirculation. Advection effects are significant, and heat transfer is very sensitive to the fluid velocity. Therefore, a small increase in Re value causes a large
increase in module average Nusselt number. When recirculation starts, the fluid in
the recirculation region keeps swirling in the form of vortices. Heat transfer occurs
from the tube surface adjacent to the recirculation region to the swirling fluid. However, these recirculation pockets do not transfer heat energy effectively to the main
stream of fluid. Thus they have little contribution to advective heat transfer. Flat
tubes have highest module average Nusselt number for all values of Re. Oval tubes
are less effective from a heat transfer point of view than flat tubes, but are better in
111
Re 25
1
2
3
4
5
Re 50
1
2
3
4
5
Re 100
1
2
3
4
5
Re 150
1
2
3
4
5
Module
6.578
6.579
6.579
6.578
6.586
9.198
9.220
9.220
9.220
9.226
10.996
10.880
10.881
10.881
10.859
12.018
11.732
11.729
11.728
11.686
4.742
4.727
4.726
4.725
4.709
3.284
2.786
2.792
2.791
2.710
2.965
2.086
2.084
2.083
1.994
Nu0
8.571
8.658
8.657
8.655
8.657
Dp 0
Circular
3.606
2.617
2.609
2.608
2.523
4.294
3.648
3.648
3.648
3.545
6.754
6.619
6.618
6.617
6.570
12.553
12.605
12.603
12.600
12.591
Dp
1.216
1.255
1.252
1.252
1.265
1.307
1.310
1.307
1.307
1.308
1.424
1.400
1.400
1.400
1.395
1.465
1.456
1.456
1.456
1.454
Dp
Dp0
14.803
13.903
13.885
13.884
13.576
13.317
12.971
12.971
12.971
12.900
10.791
10.731
10.731
10.731
10.729
7.103
7.094
7.093
7.092
7.095
Nu
Flat
1.232
1.185
1.184
1.184
1.162
1.211
1.192
1.192
1.192
1.188
1.173
1.164
1.164
1.164
1.163
1.080
1.078
1.078
1.078
1.077
Nu
1.154
1.099
1.098
1.098
1.074
1.108
1.090
1.090
1.090
1.086
1.043
1.040
1.040
1.040
1.041
0.951
0.951
0.951
0.951
0.951
Nu
3.072
2.239
2.228
2.227
2.139
3.604
3.087
3.086
3.086
2.997
5.631
5.550
5.549
5.548
5.519
10.499
10.559
10.558
10.556
10.553
Dp
1.036
1.073
1.069
1.069
1.073
1.098
1.108
1.106
1.106
1.106
1.187
1.174
1.174
1.174
1.172
1.225
1.220
1.220
1.220
1.219
Dp
Dp0
13.857
13.113
13.099
13.098
12.918
12.625
12.310
12.311
12.311
12.258
10.387
10.337
10.337
10.337
10.337
6.992
6.985
6.984
6.983
6.986
Nu
Oval
1.153
1.118
1.117
1.117
1.105
1.148
1.131
1.131
1.131
1.129
1.129
1.121
1.121
1.121
1.120
1.063
1.062
1.062
1.062
1.061
Nu
1.139
1.092
1.092
1.092
1.080
1.113
1.093
1.094
1.094
1.092
1.066
1.063
1.063
1.063
1.063
0.993
0.994
0.994
0.994
0.993
Nu
2.952
1.988
2.003
2.003
1.925
3.068
2.559
2.570
2.570
2.494
4.100
4.092
4.091
4.090
4.076
7.170
7.246
7.244
7.243
7.247
Dp
0.996
0.953
0.961
0.961
0.966
0.934
0.919
0.921
0.921
0.920
0.864
0.866
0.866
0.866
0.866
0.837
0.837
0.837
0.837
0.837
Dp
Dp0
12.021
11.272
11.290
11.289
11.122
10.823
10.547
10.548
10.548
10.526
9.098
9.080
9.080
9.080
9.084
6.549
6.547
6.547
6.547
6.552
Nu
Diamond
Table 3. Comparison of normalized pressure drop and module average Nusselt number data for the four geometric tube shapes
1.002
0.976
0.975
0.975
0.963
1.007
0.997
0.996
0.996
0.997
1.038
1.033
1.033
1.033
1.033
1.057
1.056
1.056
1.056
1.056
Nu
(Continued)
1.000
0.961
0.963
0.963
0.952
0.984
0.969
0.969
0.969
0.969
0.989
0.985
0.985
0.985
0.985
0.995
0.995
0.995
0.995
0.995
Nu
112
Re 200
1
2
3
4
5
Re 250
1
2
3
4
5
Re 300
1
2
3
4
5
Re 350
1
2
3
4
5
Module
13.459
12.456
12.436
12.431
11.821
14.960
13.025
13.012
13.005
11.720
16.361
13.481
13.492
13.484
11.600
17.689
13.885
13.915
13.907
11.465
2.703
1.471
1.437
1.434
1.393
2.626
1.306
1.261
1.255
1.228
2.568
1.184
1.132
1.123
1.102
Nu0
2.808
1.709
1.690
1.689
1.625
Dp 0
Circular
2.786
1.404
1.324
1.316
1.288
2.897
1.555
1.490
1.485
1.455
3.047
1.767
1.721
1.718
1.679
3.262
2.086
2.061
2.060
2.002
Dp
1.085
1.186
1.169
1.172
1.169
1.103
1.191
1.181
1.183
1.184
1.127
1.201
1.197
1.198
1.205
1.162
1.221
1.219
1.219
1.232
Dp
Dp0
20.402
15.940
15.837
15.822
13.434
19.070
15.535
15.448
15.435
13.589
17.712
15.091
15.019
15.010
13.692
16.292
14.570
14.518
14.514
13.718
Nu
Flat
1.153
1.148
1.138
1.138
1.172
1.166
1.152
1.145
1.145
1.171
1.184
1.159
1.154
1.154
1.168
1.210
1.170
1.167
1.168
1.161
Nu
1.123
1.085
1.080
1.079
1.112
1.128
1.087
1.083
1.082
1.107
1.138
1.090
1.087
1.087
1.098
1.151
1.094
1.093
1.093
1.083
Nu
2.460
1.243
1.157
1.147
1.120
2.542
1.368
1.296
1.290
1.257
2.654
1.543
1.489
1.486
1.441
2.814
1.805
1.773
1.771
1.706
Dp
0.958
1.050
1.022
1.021
1.017
0.968
1.048
1.028
1.028
1.024
0.982
1.049
1.036
1.036
1.034
1.002
1.056
1.049
1.049
1.050
Dp
Dp0
Table 3. Continued
19.182
15.019
14.910
14.896
12.806
17.891
14.638
14.547
14.535
12.961
16.560
14.227
14.145
14.136
13.064
15.189
13.730
13.675
13.670
13.099
Nu
Oval
1.084
1.082
1.072
1.071
1.117
1.094
1.086
1.078
1.078
1.117
1.107
1.092
1.087
1.087
1.115
1.128
1.102
1.100
1.100
1.108
Nu
1.100
1.064
1.064
1.064
1.111
1.105
1.069
1.068
1.068
1.109
1.114
1.075
1.074
1.074
1.102
1.128
1.082
1.082
1.082
1.090
Nu
2.875
1.133
1.160
1.153
1.148
2.884
1.261
1.281
1.278
1.263
2.899
1.429
1.444
1.443
1.416
2.920
1.656
1.670
1.670
1.620
Dp
1.120
0.956
1.024
1.024
1.042
1.098
0.966
1.016
1.018
1.028
1.073
0.972
1.005
1.006
1.016
1.040
0.969
0.988
0.988
0.997
Dp
Dp0
18.753
12.950
13.265
13.265
10.022
17.157
12.646
12.862
12.860
10.310
15.497
12.286
12.415
12.415
10.633
13.752
11.837
11.902
11.901
10.962
Nu
Diamond
1.060
0.933
0.953
0.953
0.874
1.049
0.938
0.953
0.954
0.889
1.036
0.943
0.954
0.955
0.907
1.022
0.950
0.957
0.957
0.927
Nu
1.021
0.947
0.946
0.945
0.862
1.016
0.949
0.948
0.948
0.881
1.012
0.952
0.953
0.953
0.902
1.008
0.960
0.961
0.961
0.928
Nu
113
Figure 13. Variation of normalized pressure drop (Dp ) with Re for module 3.
the same respect than circular tubes. Diamond tubes showed the lowest module
average Nusselt number over the entire Reynolds number range studied.
114
Dp 0
8.657
7.329
6.390
5.692
5.154
4.726
4.087
3.631
2.792
2.084
1.690
1.437
1.261
1.132
Re
25
30
35
40
45
50
60
70
100
150
200
250
300
350
6.579
7.361
7.984
8.483
8.887
9.220
9.733
10.115
10.881
11.729
12.436
13.012
13.492
13.915
Nu0
Circular
12.603
10.597
9.170
8.103
7.277
6.618
5.632
4.928
3.648
2.609
2.061
1.721
1.490
1.324
Dp
1.456
1.446
1.435
1.424
1.412
1.400
1.378
1.357
1.307
1.252
1.219
1.197
1.181
1.169
Dp
Dp0
7.093
8.117
8.970
9.673
10.252
10.731
11.465
11.994
12.971
13.885
14.518
15.019
15.448
15.837
Nu
Flat
1.078
1.103
1.123
1.140
1.154
1.164
1.178
1.186
1.192
1.184
1.167
1.154
1.145
1.138
Nu
0.951
0.975
0.996
1.014
1.028
1.040
1.059
1.071
1.090
1.098
1.093
1.087
1.083
1.080
Nu
10.558
8.875
7.680
6.788
6.098
5.549
4.729
4.145
3.086
2.228
1.773
1.489
1.296
1.157
Dp
1.220
1.211
1.202
1.193
1.183
1.174
1.157
1.142
1.106
1.069
1.049
1.036
1.028
1.022
Dp
Dp0
6.984
7.947
8.737
9.381
9.907
10.337
10.991
11.459
12.311
13.099
13.675
14.145
14.547
14.910
Nu
Oval
1.062
1.080
1.094
1.106
1.115
1.121
1.129
1.133
1.131
1.117
1.100
1.087
1.078
1.072
Nu
0.994
1.013
1.029
1.043
1.054
1.063
1.076
1.084
1.094
1.092
1.082
1.074
1.068
1.064
Nu
Table 4. Variation of heat transfer and pressure drop with Re for module 3
7.244
6.172
5.418
4.860
4.431
4.091
3.586
3.228
2.570
2.003
1.670
1.444
1.281
1.160
Dp
0.837
0.842
0.848
0.854
0.860
0.866
0.878
0.889
0.921
0.961
0.988
1.005
1.016
1.024
Dp
Dp0
6.547
7.312
7.914
8.391
8.772
9.080
9.547
9.886
10.548
11.290
11.902
12.415
12.862
13.265
Nu
Diamond
0.995
0.993
0.991
0.989
0.987
0.985
0.981
0.977
0.969
0.963
0.957
0.954
0.953
0.953
Nu
1.056
1.052
1.047
1.043
1.038
1.033
1.025
1.016
0.996
0.975
0.961
0.953
0.948
0.946
Nu
Figure 15. Variation of module average Nusselt number with Re for module 3.
115
116
H. M. S. BAHAIDARAH ET AL.
Figure 16. Variation of heat transfer enhancement ratio (Nu ) with Re for module 3.
Figure 17. Variation of heat transfer performance ratio (Nu ) with Re for module 3.
117
in the low-Re range (i.e., Re < 50), advantage in heat transfer over circular tubes can
be maximized by increasing the fluid velocity for any of the flat and oval tubes. However, diamond tubes are not a good replacement for circular tubes from the viewpoint of heat transfer, especially at high fluid velocities. At high Reynolds
numbers, change in the geometric shape of the tube cross section rather than increase
in fluid velocity should be considered in equipment design for better heat transfer rate.
Heat Transfer Performance Ratio
Variation of heat transfer performance ratio (Nu ) with Re for module 3 is presented in Figure 17 and Table 4. Nu signifies heat transfer enhancement per unit
increase in pumping power. This parameter is useful to compare various tube cross
sections if both the heat transfer rate and pumping power are of major concern in the
design of equipment. For Re < 40, diamond tubes are better than both the flat and
oval tubes. However, for Re > 50, flat and oval tubes have substantial advantage
over diamond tubes. Again, below Re 50 (low-Re range), increase in fluid velocity
helps substantially. However, at high Re values, geometric shape is more important
than fluid velocity.
CONCLUSIONS
Flat and oval tubes offer more flow resistance compared to circular tubes. A
heat exchanger with flat or oval tubes will cause more pressure drop, and hence
require more pumping power, than one with circular tubes. Oval tubes, however,
offer less flow resistance when compared to flat tubes. Diamond tubes cause less
pressure drop than circular tubes for Re < 250. Hence, if pumping power is a concern, diamond tubes should be preferred over circular tubes only in the low Reynolds
number range (Re < 250). For Re > 250, circular tubes are better than diamond
tubes from the viewpoint of pressure drop.
Oval tubes are less effective from a heat transfer point of view compared to flat
tubes, but are better in the same respect than circular tubes. Circular tubes are in
turn better than diamond tubes with respect to heat transfer rate. Rate of heat transfer is very sensitive to fluid velocity in the low-Re range (i.e., Re < 50). Therefore, if
heat transfer rate is the major design consideration, fluid velocity should be maximized in this Re range. However, in the high-Re range (i.e., Re > 150), fluid velocity
has less effect on heat transfer rate than the tube cross-section geometry does.
A better tube cross section should be chosen instead of increasing the fluid velocity,
in order to maximize heat transfer rate in the high Reynolds number range (i.e.,
Re > 150).
At very low fluid velocity, the geometric shape of the tube cross section does
not affect the heat transfer rate significantly.
Diamond tubes are better than both flat and oval tubes for Re < 40 from the
viewpoint of heat transfer enhancement per unit increase in pumping power (Nu ).
However, for Re > 50, flat and oval tubes perform better than diamond tubes. For
Re < 50, increase in fluid velocity causes Nu to increase steeply for oval and
flat tubes, and to decrease rapidly for diamond tubes. Oval tubes remain superior
to flat tubes for Re 100. For Re > 150, flat tubes are better than oval tubes,
118
H. M. S. BAHAIDARAH ET AL.
whereas oval tubes are superior to diamond tubes from the viewpoint of heat transfer performance ratio. If the design of a system is to be optimized for heat transfer
enhancement per unit increase in pumping power, both geometry and fluid velocity
must be carefully chosen for Re < 150. However, for Re > 150, fluid velocity is not
as important as the geometric shape of the tubes. For Re > 350, Nu is far more
sensitive to geometric shape than fluid velocity.
REFERENCES
1. T. Ota, H. Nishiyama, and Y. Taoka, Heat Transfer and Flow around an Elliptic Cylinder, Int. J. Heat Mass Transfer, vol. 27, no. 10, pp. 17711779, 1984.
2. T. Ota, H. Nishiyama, J. Kominami, and K. Sato, Heat Transfer from Two Elliptic Cylinders in Tandem Arrangement, J. Heat Transfer, vol. 108, pp. 525531, 1986.
3. T. Wung, J. Niethammer, and C. Chen, Measurements of Heat-Mass Transfer and Pressure Drop for Some Non-standard Arrays of Tubes in Cross Flow, in C. Tien, V. Carey,
and J. Ferrell (eds.), Heat Transfer, Proc. Eighth Int. Heat Transfer Conf., pp. 10411046,
Hemisphere, San Francisco, 1986.
4. C. H. K. Williamson, Vortex Dynamics in the Cylinder Wake, Annu. Rev. Fluid Mech,
vol. 28, pp. 477539, 1996.
5. M. M. Zdrakovich, Flow Around Circular Cylinders, 1: Fundamentals, Oxford University
Press, New York, 1997.
6. D. Kundu, A. Haji-Sheikh, and D. Y. S. Lou, Heat Transfer Predictions in Cross Flow
over Cylinders between Two Parallel Plates, Numer. Heat Transfer A, vol. 19, pp. 361
377, 1991.
7. D. Kundu, A. Haji-Sheikh, and D. Y. S. Lou, Pressure and Heat Transfer in Cross Flow
over Cylinders between Two Parallel Plates, Numer. Heat Transfer A, vol. 19, pp. 345
360, 1991.
8. D. Kundu, Computational and Experimental Studies of Flow Field and Heat Transfer
from a Row of In-line Cylinders Centered between Two Parallel Plates, Ph.D. thesis,
University of Texas at Arlington, Arlington, TX, 1989.
9. V. B. Grannis and E. M. Sparrow, Numerical Simulation of Fluid Flow through an Array
of Diamond-Shaped Pin Fins, Numer. Heat Transfer A, vol. 19, pp. 381403, 1991.
10. Y. Chen, M. Fiebig, and N. K. Mitra, Conjugate Heat Transfer of a Finned Oval Tube
Part A: Flow Patterns, Numer. Heat Transfer A, vol. 33, pp. 371385, 1998.
11. Y. Chen, M. Fiebig, and N. K. Mitra, Conjugate Heat Transfer of a Finned Oval Tube
Part B: Heat Transfer Behaviors, Numer. Heat Transfer A, vol. 33, pp. 387401, 1998.
12. M. Breuer, J. Bernsdorf, T. Zeiser, and F. Durst, Accurate Computations of the Laminar
Flow past a Square Cylinder based on Two Different Methods: Lattice-Boltzmann and
Finite-Volume, Int. J. Heat Fluid Flow, vol. 21, pp. 186196, 2000.
13. R. L. Webb, Principles of Enhanced Heat Transfer, Wiley, New York, 1993.
14. H. M. Bahaidarah, A Numerical Study of Heat and Momentum Transfer over a Bank of
Flat Tubes, Ph.D. thesis, Texas A&M University, College Station, TX, 2004.
15. H. M. Bahaidarah, N. K. Anand, and H. C. Chen, Numerical Study of Fluid Flow
and Heat Transfer over a Bank of Flat Tubes, Numer. Heat Transfer A, vol. 48,
pp. 359385, 2005.
16. A. M. F. El-Shaboury and S. J. Ormiston, Analysis of Laminar Forced Convection of Air
Crossflow in In-line Tube Banks with Non-Square Arrangements, Numer. Heat Transfer
A, vol. 48, pp. 99126, 2005.
17. M. Fujii, T. Fujii, and T. Nagata, A Numerical Analysis of Laminar Flow and Heat
Transfer of Air in In-line Tube Bank, Numer. Heat Transfer A, vol. 7, pp. 89102, 1984.
119