2004 Report to Supercomputer Institute

Proceedings of HT2005
2005 ASME Summer Heat Transfer Conference
July 17-22, 2005, San Francisco, California, USA

NUMERICAL SIMULATION OF FLOW FIELD AND HEAT TRANSFER OF

STREAMLINED CYLINDERS IN CROSSFLOW

Zhihua Li, Jane Davidson and Susan Mantell

Department of Mechanical Engineering
University of Minnesota
111 Church St., SE
Minneapolis, MN 55455
jhd@me.umn.edu

ABSTRACT
The use of streamlined tubes to reduce pressure drop across polymer tube bundles is
considered because of the relative ease of fabrication. The drag and convective heat transfer
coefficients along the outer surface of lenticular and elliptical tubes with minor-to-major axis
ratios of 0.3, 0.5, and 0.8 are determined numerically for cross-flow Reynolds numbers from 500
to 10,000. An isothermal surface is assumed. The two-dimensional, unsteady Navier-Stokes
equations and energy equation are solved using the finite volume method. Laminar flow is
assumed from the front stagnation point up to the point of separation. Turbulent flow in the wake
is resolved using the shear stress transport k-omega model. Local heat transfer, pressure and
friction coefficients as well as a total drag coefficient and average Nusselt number are presented.
The results for streamlined tubes are compared to published data for circular and elliptical
cylinders. Drag of the elliptical and lenticular cylinders is similar and lower than a circular
cylinder. Reductions in drag may be increased by making the streamlined cylinders more
slender. Over the range of Reynolds number considered, an elliptical cylinder with an axis ratio
equal to 0.5 reduces pressure drop by 30 to 40 percent compared to that of a circular cylinder.
The lenticular and elliptical geometries have nearly identical average of Nusselt number. The
average Nusselt number of an elliptical or lenticular cylinder with axis ratio of 0.5 and 0.3 is 15
to 35% lower than that of a circular cylinder. A case study for an automotive radiator is
presented to illustrate comparison of shaped and circular tubes in terms of both heat transfer and
pressure drop.

Keywords: Heat transfer, enhancement, pressure drop, lenticular, elliptical, cylinder, cross flow

1
 2004 Report to Supercomputer Institute

INTRODUCTION
Polymer heat exchangers are commonly used with corrosive fluids [e.g. 1-4] and may
compete with metal heat exchangers in diverse applications where weight is a concern or
innovative shapes are desirable. Because of the relative ease of extruding polymer tubes and into
a variety of shapes, it is feasible to use streamlined rather than circular tubes in polymer tube
bundles. The present study focuses on elliptical and lenticular outer tube profiles, which have
the potential to reduce form drag for flow across small diameter tubes. For polymer tubes, there
is a tradeoff between thermal and mechanical performance and thus, optimization of the tube
geometry for a specific application must consider the inside geometry of the flow passage as well
as the outer profile. A procedure for selecting the inside geometry for a specified material, outer
profile, and heat exchanger application has been developed [5]. For many applications, a tube
with non-uniform wall thickness emerges as the optimum solution. For example, the “shaped”
tube in Fig. 1 has a circular inner flow passage and an elliptical outer profile. In this case, the
temperature distribution along the outer surface of the tube is non-uniform and determination of
the overall thermal resistance of the tubes requires special treatment. A relatively simple
approach is to treat the shaped tube as a base circular tube to which longitudinal fin(s) are added
to form the outer profile. A shaped tube efficiency can then be used in the same manner as fin
efficiency to determine the outside convective resistance once the average film coefficient for a
constant temperature surface is known [6].
The objective of the present numerical study is to determine the drag and average heat
transfer coefficient of various elliptical and lenticular tubes (Fig. 2) in cross flow. In the analysis
the tubes are treated as solid bodies and hence are referred to as cylinders. The flow range of
interest is 500≤ReD≤104 based on air speeds less than 30 m/s and small diameter cylinders (1.5~
4 mm). Prior data in this range are limited to the ellipse and do not adequately quantify the
effect of geometry, i.e. ratio of the minor to major axis. To facilitate a straightforward
comparison of shapes, the length scale used for Re and Nu is the minor axis, D = 2ro. The flow
is assumed to be in the direction of the major axis, i.e. with zero angle of attack, and a constant
temperature boundary is assumed at the surface of the cylinder. The numerical data are
presented in the form of correlations of drag coefficient and average Nusselt number with respect
to ReD and a length ratio, λo, equal to the ratio of the minor to major axis. A value of λo = 1
represents a circular cylinder. The streamlined cylinders become more slender as λo is
decreased. The results for the streamlined cylinders are compared with measured data for
circular cylinders in cross flow. A case study for an automotive radiator is presented to illustrate
under what circumstances a tube with a streamlined outer profile is beneficial.

PRIOR WORK
Studies of the flow and temperature distribution of circular cylinders in cross flow
abound in the literature; reference [7] provides an excellent summary. Measurement of drag and
heat transfer of individual noncircular cylinders in cross flow is limited to elliptical cylinders [7-
10] primarily for ReD ≥ 104 (see Fig. 2). Drag decreases as the ellipse is made more slender, i.e.,
λo → 0. At ReD ~ 104, flow separation occurs at α = 110 to 140 degrees, with α decreasing as λo
is decreased [7]. For ReD ≥ 104, the drag coefficient is estimated by [9]

2
 2004 Report to Supercomputer Institute

1
C D = 0.015(1 + ) + 1.1λ o . (1)
λo
This correlation indicates the drag of an ellipse with λo = 0.5 is 60% of that of a circular cylinder.
Drag reductions may be less significant at lower Re because the boundary layer remains laminar
prior to separation.
Additional reductions in drag are possible for arrays of streamlined cylinders [11, 12].
Measured drag of an array of lenticular cylinders with λo = 0.25 arranged with a transverse pitch-
to-diameter ratio of 2 at 103 < Res,max < 5×104 is 30% of that for a similar array of circular
cylinders [13]. Two-dimensional numerical results also show reduced drag for lenticular
cylindrical arrays as compared to arrays of circular cylinders [14, 15]. Tauscher [16] compared
performance of arrays of elliptical, rhomboid, lanceolate and circular cylinders at ReDh,max <
2500. Using the ratio of Nusselt number to drag coefficient as a performance criterion, the
arrays of lanceolate and elliptical cylinders are 80% and 70% superior to cylindrical tubes for
fixed volumetric flow rate and overall heat exchanger volume.
Heat transfer correlations developed for flow across an elliptical cylinder are presented in
Table 1 in the general form,
Nu L = C Re L , (2)
m

with some variation in the characteristic length scale L. Based on use of a common length scale
equal to twice the minor axis (D = 2ro), the Nu of an elliptical cylinder with λo = 0.5 is 12%
lower than that of a circular cylinder at ReD = 5000.

NUMERICAL METHOD
The flow and temperature distributions around individual streamlined cylinders are
modeled with FLUENT® assuming an unsteady two-dimensional flow. The boundary layer over
the cylinder is modeled by assuming a laminar zone from the front stagnation point at α = 0 to α
< 90°, and a turbulent zone from 90° ≤ α ≤ 180°, using the shear stress transport SST k-ω model
[17]. Increasing the laminar zone to α = 110°, which is closer to the predicted angle of
separation, had no effect on the calculated drag coefficient. This approach was taken after
unsuccessful trials of fully laminar and turbulent models. The laminar models overpredicted the
size of the bubble in the wake. The turbulent models (the standard k-ε model, the standard k-ω
model, the SST k-ω model and the v2F model) underpredicted drag, especially for slender
ellipses, by as much as 30%. The combined laminar/turbulent approach produces local and
average results that compare favorably to measured data.
The computational domain and boundary conditions are shown in Fig. 3; an elliptical
cylinder is shown for illustration. The size of the domain ensures the boundaries do not affect
vortex shedding or drag force. It has been suggested that the lateral boundary should be at least
eight diameters from the center of a circular cylinder at ReD = 100 [21]. Because the current
simulations involve higher Reynolds numbers, the lateral boundary is located sixteen diameters
from the center of the cylinder. The upstream and downstream boundaries are set to 16l and
40l , respectively based on prior simulation of a circular cylinder [22].

3
 2004 Report to Supercomputer Institute

Table 1. Heat transfer correlations for cross flow over individual elliptical cylinders.
Correlation Length Scale Source

Nu L = C Re L
m

and λo = 0.5 ,
10 3 < Re L < 10 5 L = 2l [18]
C = 0.344, m = 0.573 at φ = 0°.
C = 0.470, m = 0.537 at φ = 15°.
Nu L = C Re L ,
m

L = 2l [19]
10 3 < Re L < 10 5 and λ o = 0.3
C = 0.546, m = 0.539 at φ = 0°
C = 0.839, m = 0.492 at φ = 15°
0.25
⎛ Pr ⎞
Nu L = 0.27 Re L Pr ⎜⎜ ⎟⎟
0.6 0.37
L = 2l [7]
Pr
⎝ w⎠
10 3 ≤ Re L ≤ 10 4 , λ o = 0.5 , φ = 0
Nu L = 0.224 Re L ,
0.612

10 3 < Re L < 10 4 , φ=0 L = ro + l [20]

At the inlet, uniform air velocity and temperature boundary conditions are imposed.
p = p∞ , u x = u∞ , u y = 0 . (3)
T∞ = 300 K . (4)
The initial velocity and temperature fields are set equal to the uniform inlet condition. The inlet
freestream turbulence intensity is set to 1%. A fully developed boundary condition is applied at
the outlet.
∂ξ
= 0 , ξ = ux , u y ,T, p . (5)
∂x
Symmetry boundary conditions are applied at the lateral boundaries.
∂ξ
= 0 , ξ = u x ,T , p , (6)
∂y
uy = 0 . (7)

The no-slip boundary condition is applied at the isothermal cylinder wall.

u x,w = u y,w = 0 . (8)
Tw = 360 K . (9)
The second-order upwind method is used for spatial discretization. Time derivatives are
discretized using the second-order implicit method. The segregated solver is applied to solve the
momentum equations and the energy equation. At each step, the momentum equations are first
solved to obtain the velocity field. The pressure field is updated by solving the Poisson equation.
The pressure field and the velocity field are coupled by the SIMPLEC method. After solving the
continuity equation and momentum equations, the energy equation is solved to achieve a steady
periodic solution. The integrated pressure force and heat transfer over the cylinder surface are

4
 2004 Report to Supercomputer Institute

monitored as a function of time at each Re. The steady-periodic state is achieved when the
variation in the oscillating amplitudes of the pressure force and heat transfer rate is less than 1%.
The computational domain is meshed using GAMBIT [23] with unstructured triangular
meshes using a pave method except inside the boundary layer where a quadratic mesh was used
(Fig. 4). The boundary layer thickness before separation is estimated by a theoretical
formulation [7]. The inlet and outlet boundaries are meshed into 32 grids. The upper and lower
boundaries are meshed into 150 grids. The cylinder wall is meshed into 60 grids. Thirty layers
of quadratic meshes are attached inside the boundary layer. The radial thickness of the first
mesh layer is 5×10-5ro. The mesh is increased by a factor of 1.3 in the radial direction. The total
thickness of the attached quadratic meshes is 0.45ro. The mesh is refined until the variations of
time averaged CD and NuD are less than 0.5% and 2.5%, respectively
Large time steps are applied initially to reach a steady-periodic solution efficiently. At
that point, the time step is selected based on the expected vortex shedding frequency in the wake
area [24-27]. For each shape, 100 time steps are performed per period. The time step was
selected to assure convergence of the total drag over the range of Reynolds number considered.

ANALYSIS
The primary results of interest are the steady periodic overall drag coefficient and
average Nusselt number. Local values of friction coefficient, pressure coefficient and Nusselt
number are compared to prior experimental data to validate the computational approach. The
time averaged value of variable f(t) at the steady periodic state is calculated by
t1

∫ f (t )dt
t =t0
f = , (10)
t1 − t 0
where t1-t0 is ten times of the vortex shedding period.

Flow Field
Local pressure and friction coefficients as well as total drag are computed. The local
pressure coefficient is
pα − p0
C p ,α = 1 + , (11)
1
ρu ∞
2

2
where pα is the time-averaged local pressure on the cylinder surface with an angular coordinate
of α measured from the front stagnation point and p 0 is the stagnation pressure. The local
friction coefficient is defined as
τw
C f ,α = . (12)
1
ρu ∞
2

2
The wall shear stress is evaluated from the normal velocity gradient at the cylinder wall,
∂u (t )
τ w (t ) = µ . (13)
∂n w

5
 2004 Report to Supercomputer Institute

The drag coefficient, CD, is the sum of the time-averaged pressure coefficient and time-averaged
friction coefficients, i.e.,
CD = C p + C f , (14)

where
Fp
Cp = , (15)
ρu ∞
2

A⊥
2
and
Ff
Cf = . (16)
ρu ∞
2

A⊥
2
The projected area of the cylinder is perpendicular to the direction of flow; A⊥ = 2ro , assuming
cylinders of unit length. Total pressure force and wall shear stress are determined by integration
over the surface of the cylinder:
2π

F p (t ) = ∫p α (t )rα cos αdα , (17)

0
2π

F f (t ) = ∫ τ w (t )rα cos α dα . (18)

0

Heat Transfer
Local and average Nusselt numbers are determined. The local Nusselt number is
determined by
⎛ ∂T (t ) ⎞
t 1

− ∫ 2ro ⎜ ⎟ dt
⎝ ∂n ⎠ w
Nu D ,α =
t 0
. (19)
t1 − t 0
The average heat transfer coefficient and Nusselt number over the surface are calculated by
Q
h= , (20)
AHT (Tw − T∞ )
and
h ⋅ 2ro
Nu D = , (21)
k∞
where heat transfer rate Q is reported in FLUENT as a function of time. The heat transfer
surface areas, AHT, for elliptical and lenticular cylinders of unit length are given by Eqs. (22) and
(23), respectively:
AHT ,e = π (ro + l) (22)
⎛l⎞
AHT ,l = 4 R arcsin⎜ ⎟ . (23)
⎝R⎠

6
 2004 Report to Supercomputer Institute

RESULTS
The presentation which follows is structured to first validate the numerical approach and
second to highlight the effect of shape, both geometry and length ratio, λo, on drag and heat
transfer.

Validation of Numerical Approach

The numerical approach is assessed by comparing local and average values of drag
coefficient and Nusselt number to prior data for an ellipse with λo = 0.5. Local friction and
pressure coefficients are presented in Figs. 5 and 6, respectively. The overall drag coefficient is
plotted as a function of ReD in Fig. 7. Local and average Nusselt numbers are shown in Figs. 8
and 9.
Figure 5 compares numerical and measured [7] values of Cf,α at ReD = 3200. The
measured data were obtained for a freestream turbulence intensity of 0.3%. The error bars in the
graph are based on estimates in [7]. The overall trends in the predicted and measured data are
similar. The predicted separation point at α = 112 degrees (the angle at which Cf,α = 0) agrees
well with reported values of 110 to 140 degrees [7]. Skin drag for the Reynolds number range of
interest is much less than form drag. Predicted local pressure coefficients are within ±15% of the
data (Fig. 6). The predicted drag coefficient for 500 ≤ ReD ≤ 104 agrees with the measured data
[7, 8], which is somewhat scattered (Fig. 7). The data and numerical model indicate the drag
coefficient is relatively insensitive to ReD.
Local Nusselt numbers are plotted in Fig. 8 for ReD = 3074. The numerical Nusselt
number is compared with an empirical correlation and data obtained with a constant heat flux
boundary condition [18]. The predicted values are close to the measured data, which are about
20% lower than the correlation. Figure 9 shows the average Nusselt number for 500 ≤ ReD ≤
104. The numerical Nusselt number agrees well with empirical correlations [7, 18, 28].

Parametric study
Attention is first turned to the effect of streamlined cylinder shape, both geometry and
length ratio, λo, on drag. Numerical results for elliptical and lenticular cylinders with λo = 0.3,
0.5 and 0.8 at 500 ≤ ReD ≤ 104 are compared to experimental data for circular cylinders (λo = 1).
The results are interpreted to determine the conditions under which the shaped tube is feasible
from the perspective of reducing drag in laminar cross flow.
Figure 10 shows the friction, pressure and total drag coefficients versus Re for an
elliptical cylinder with λo = 0.3. The plot illustrates the importance of form drag as opposed to
friction drag even for the streamlined cylinders. The trends shown in this figure are similar for
the other length ratios and the lenticular cylinder.
Figure 11 shows the effect of geometry and λo on the overall drag coefficient of the
streamlined shapes. The drag coefficient can be best correlated at 5000 ≤ ReD ≤ 104 by
C D ,e = (0.112 ± 0.009)λo ln Re D ,e + (0.41 ± 0.04)
2.6 ± 0.5
(24)
for elliptical tubes with R2=0.9919 and
C D ,l = (0.120 ± 0.006)λo ln Re D ,l + (0.46 ± 0.05)
2.2 ± 0.4
(25)
2
for lenticular cylinders with R =0.9945. The drag coefficients of the elliptical and lenticular
cylinders are similar. The impact of Re within the range considered is insignificant for both
geometries. For both streamlined shapes, CD decreases as the cylinders are slenderized, i.e. λo is

7
 2004 Report to Supercomputer Institute

decreased. For example, for an elliptical cylinder, CD at λo = 0.3 is 25% lower than that at λo =
0.5 and 60% lower than the circular cylinder [29]. The decrease in drag coefficient is attributed
to a delay of separation as λo is decreased (Table 2). Significant reductions in drag compared to
that of the circular cylinder are possible for λo ≤ 0.5. To illustrate this point, consider elliptical
and lenticular cylinders with λo = 0.5 at ReD = 5000; drag is reduced by 37%, and 31%,
respectively, compared to the circular cylinder.

Table 2. Angle of separation for circular, elliptical and lenticular cylinders with λo = 0.3, 0.5 and
0.8 at 500 ≤ ReD ≤ 104.
Shape λo Angle of Separation
(degree)
Circular 1.0 80 (measured by [7] )
Elliptical 0.3 155-1351
0.5 130-100
0.8 110-90
Lenticular 0.3 145-110
0.5 130-95
0.8 110-90
1
The angle of separation increases slightly as Re is increased. For example, for an ellipse with λo = 0.3, separation
occurs at 155° at ReD = 500 and 135° at Re = 104.

Next, we consider the effect of shape on heat transfer. Although the streamlined
cylinders provide a reduction in drag, average Nusselt number is expected to be lower than that
for a circular cylinder. The potential benefit of added surface area depends on the material and
geometry. For cylinders with non-uniform wall thickness, heat transfer is expressed in terms of a
shaped tube efficiency χ, which depends solely on Biot number, based on the length of material
added to a base circular tube, and a dimensionless length, λo = ro / l [6]. The shaped tube
efficiency is used in the same manner as a fin efficiency to determine the outside convective
resistance.
1
Rth , o = . (26)
χ ho Ao
The average heat transfer coefficient for a constant temperature surface can be obtained from the
present simulations. The predicted average Nusselt number for elliptical and lenticular cylinders
is plotted along with that for a circular cylinder at 500 ≤ ReD ≤ 104 in Fig. 12. The numerical
results are best correlated for the elliptical cylinder by
Nu D , e = (0.37 ± 0.03)λ o , e
0 .4 ± 0 .1 0.554 ± 0.008
Re D , e , (27)
2
with a coefficient of determination R = 0.9992, and for the lenticular cylinder by
Nu D ,l = (0.3 ± 0.1)λ o ,l
0.46 ± 0.06 0.60 ± 0.05
o
Re D ,l . (28)
with R2 = 0.9817. The elliptical cylinder correlation is nearly identical to that reported in
literature [7, 18, 19]. NuD increases with increasing λo. An elliptical cylinder with λo = 0.5 has a
16% higher heat transfer coefficient than that one with λo = 0.3. At λo = 0.8, the predicted

8
 2004 Report to Supercomputer Institute

Nusselt number reaches that of a circular cylinder whose diameter equals the length of the minor
axis (D = 2ro).

Case Study: Comparison of Pressure Drop and Convective Heat Transfer of a

Circular Tube and Streamlined Tube for an Automotive Radiator
To illustrate determination of the overall heat transfer of a shaped tube and to compare its
performance in a heat exchange application to that of a circular tube, we consider the use of an
elliptical tube made of either nylon or aluminum for an automotive radiator. Based on the
procedure presented in [5], the optimal geometry for a tube with an elliptical outer profile is one
with a circular inner flow passage as depicted in Fig. 1. Here we assume the tube has an inside
circular inner flow passage of radius of ri = 2 mm. The length ratio, λo, is allowed to vary.
Nominal operating conditions for an automotive radiator are provided in Table 3. Table 4
t
provides the required wall thickness to outer semi-minor axis ratio, λt = , for a 5% strain limit.
ro
This strain limit will ensure that the streamlined outer profile is maintained. In this example, the
modulus for nylon 6,6 is 43 MPa (assuming 10 years of continuous use in a water antifreeze
mixture [30]) and the thermal conductivity is 0.24 W/m⋅K; the modulus for aluminum is 71 GPa
and the thermal conductivity is 237 W/m⋅K. Because Young’s modulus for aluminum is much
higher than that of nylon, the tube wall thicknesses for the aluminum tubes are much smaller
than that of the nylon tubes, i.e. λt , alu min um < λt , nylon .

Table 3. Nominal operating conditions of an automotive radiator.

.
Tinlet V U P
Fluid (K) (m3/s) (m/s) (MPa)
50% ethylene
Inside glycol/water 360 0.00121 - 0.2
Outsi
de air 300 - 2-35 0.1
1
The velocity in tubes is estimated assuming there are 600 tubes and the nominal total volumetric flow rate is 0.0012
m3/s. The velocity over the tubes is based on a vehicle speed of 7.2-126 km/hr.

9
 2004 Report to Supercomputer Institute

Table 4. Radiator nylon tube geometry based on 5% strain limit.

Required
Specified geometric
thickness-to- Calculated dimensions
parameters
radius ratio
NYLON 6,6
ri (mm) λo λt ro (mm) l o (mm) AHT (mm2)
2.0 0.3 0.15 2.30 7.67 31.32
2.0 0.5 0.15 2.30 4.60 21.68
2.0 0.8 0.14 2.28 2.85 16.12
2.0 1.0 0.10 2.20 2.20 13.82
ALUMINUM
ri (mm) λo λt ro (mm) l o (mm) AHT (mm2)
2.0 0.3 0.0002 2.0004 6.668 27.23
2.0 0.5 0.00015 2.0003 4.0006 18.85
2.0 0.8 0.00015 2.0003 2.5004 14.14
2.0 1.0 0.00002 2.00004 2.00004 12.57

The ratio of heat transfer of an elliptical shaped tube to that of a circular tube is termed
the shaped tube effectiveness,
Q
ε≡ e. (29)
Qc
The elliptical tube will enhance heat transfer if ε > 1. Assuming the circular tube and the
elliptical tubes have the same inner wall surface temperature (Ti) and unit length, the heat
transfer rate is given by
Ti − T∞
Q= . (30)
1 ro 1
ln +
2πk w ri χhAHT
The shaped tube efficiency χ expresses the heat transfer penalty of the material added to a
circular tube to achieve a streamlined outer profile. Values of χ can be found in [6] for elliptical
tubes with λo = 0.3, 0.5 and 0.8. For a circular tube, and for any tube made of aluminum, χ
equals 1. For polymer tubes, χ is less than 1. The heat transfer coefficient is determined using
Eq. (27) for the elliptical shaped tubes, and
0.25
0.37 ⎛ Pr f ⎞
Nu D ,c = 0.52 Re D ,c Pr f ⎜⎜ ⎟⎟ 40 < Re D ,c < 10 3
0.5
(31)
⎝ Prw ⎠
0.25
0.37 ⎛ Pr f ⎞
Nu D ,c = 0.26 Re D ,c Pr f ⎜⎜ ⎟⎟ 10 3 < Re D ,c < 10 5
0.6
(32)
⎝ Prw ⎠
for the circular tube [7]. Substituting expressions for Q , ri, AHT of the circular tube and elliptical
tube into Eq. (27) and expressing the result in terms of dimensionless variables yields

10
 2004 Report to Supercomputer Institute

⎛ ⎛ 1 ⎞ 1 ⎞
⎜ ln⎜ ⎟ ⎟
⎜ ⎜ 1 − λ ⎟ + Bi ⎟
⎝ ⎝ ⎠ ⎠c
ε=
t
. (33)
⎛ ⎛ 1 ⎞ 2 ⎞
⎜ ln⎜ ⎟ ⎟
⎜ ⎜ 1 − λ ⎟ + χBi(1 + 1 / λ ) ⎟
⎝ ⎝ t ⎠ o ⎠e
For a specified inside radius and material, the tube effectiveness depends on only λo and ReD.
The behavior of ε is displayed in Fig. 13 as a function of ReD for elliptical shaped tubes
with λo = 0.3, 0.5 and 0.8 at 500 ≤ ReD ≤ 104. Using a polymeric material, the elliptical tube
does not enhance heat transfer except at the very low end of Reynolds numbers considered, i.e.
Re < 1000. The enhancement is restricted to the less streamlined shapes with λo ≥ 0.5. The
penalty of added material to a circular tube out weighs the benefit of increased heat transfer
surface area. Thus, for polymers, the primary advantage of using elliptical or lenticular tubes as
compared to circular tubes is their ability to reduce pressure drop. On the other hand, the
aluminum streamlined tubes enhance heat transfer by 18 to 88% at 500 ≤ ReD ≤ 104, compared
with a circular tube. The enhancement in heat transfer increases as the ellipse is made more
slender, analogous to increasing fin length. For example, the heat transfer rate of elliptical tubes
is 61 to 88%, 33 to 56% and 18 to 38% higher than that of a circular tube for tubes with λo = 0.3,
0.5 and 0.8.
Performance evaluations of heat exchangers normally consider both heat transfer and fan
pumping power. The ratio of heat transfer rate to fan pumping power of an individual elliptical
to that of a circular tube is defined as the performance effectiveness,

ψ=
(Q/P e) . (34)
(Q/P c)
The fan pumping power is
1
P = C D ⋅ ρu ∞ ⋅ A⊥ ⋅ u ∞ .
2
(35)
2
If ψ > 1, the performance of a tube is enhanced by the use of an elliptical tube.
The performance effectiveness is plotted as a function of ReD in Fig. 14 for nylon and
aluminum tubes. The performance effectiveness of an aluminum elliptical tube with λo = 0.3 is
about four. The enhancement is due primarily to the reduction of pressure drop. For elliptical
tubes made of nylon, the enhancement is restricted to ellipses with λo ≤ 0.5. The difference
between the performance of nlyon and aluminum tubes is attributed to the low thermal
conductivity and thicker wall of the nylon tubes.

CONCLUSION
The numerical simulations of the drag and heat transfer around elliptical and lenticular
cylinders in cross flow for 500 ≤ ReD ≤ 104 demonstrate the effects of using streamlined
cylinders rather than circular cylinders in heat exchange applications. The simulations provide
generalized expressions for drag coefficient and average Nusselt number for both elliptical and
lenticular cylinders for a range of profiles. Minor-to-major axis ratios (λo) equal to 0.3, 0.5 and
0.8, are considered for each geometry.
The streamlined tubes result in a delay of separation and thus increasing reductions in
form drag as the cylinders are made more slender, i.e. λo is decreased. There is no substantial
difference in the drag over elliptical or lenticular cylinders with the same value of λo. Compared

11
 2004 Report to Supercomputer Institute

with a circular tube, the drag coefficient is reduced by 30 to 40% by the use of an elliptical tube
or lenticular tube with λo = 0.5. The numerical results are well correlated in the forms given in
Eqs. (24) and (25).
General Nusselt number correlations are developed for constant surface temperature
cylinders. As with drag coefficient, the average convective heat transfer coefficient is similar for
elliptical and lenticular cylinders for a fixed value of λo. This similarity is based on using a
common length scale based on the diameter of the cylinder perpendicular to the flow, i.e. the
minor axis. The Nusselt number increases with increasing λo and is lower than measured values
for a circular cylinder. The numerical results are well correlated in the forms given in Eqs. (31)
and (32).
Using the numerical drag and heat transfer results, a design case study for an automotive
radiator made of hundreds of tubes (as has been proposed for polymer heat exchangers) is
presented to illustrate when the streamlined tubes might be of benefit. The challenge to the use
of polymer tubes is low material strength and thermal conductivity. The tube geometry which
minimizes thermal resistance is one in which the wall thickness is non-uniform, specifically a
tube with a circular inner flow passage and a streamlined elliptical outer profile. For this
particular application, the polymer elliptical tube does not enhance heat transfer. However based
on a performance criterion of ratio of heat transfer rate to fan pumping power, a polymer
elliptical tube with λo ≤ 0.5 is superior to a circular tube for Reynolds numbers less than
approximately 7000. Although not necessarily intended for this application, elliptical tubes
made of aluminum are superior to circular tubes for all shapes considered. The assumption in
this analysis is that tube bundles behave as a collection of individual tubes. Future simulations
will consider tube bundles in which the ability of the streamlined tubes to reduce pressure drop
may be further improved.

ACKNOWLEDGEMENTS
The authors are indebted to Professor Krishnan Mahesh who helped guide the selection
of numerical method. The financial support of the U. S. National Science Foundation through
Grant DMII-0084765, the University of Minnesota Graduate School Dissertation Fellowship and
the University of Minnesota Supercomputing Institute are acknowledged.

NOMENCLATURE

AHT Heat transfer surface area of cylinders, per unit length, m2

A⊥ Projected are of the cylinder in the direction perpendicular to flow, m2
Bi hro
Biot number, Bi =
kw
CD Drag coefficient
Cf Friction coefficient
Cp Pressure coefficient
D Length of the minor axis, D = 2ro, m
Fp Pressure force, N
Ff Friction force, N
h Convective heat transfer coefficient, W/m2⋅K

12
 2004 Report to Supercomputer Institute

k Thermal conductivity, W/m⋅K

l Length of the semi major axis, m
Nu Nusselt number
P Fan pumping power, W
Pr Prandtl number of the fluid, evaluated at the fluid bulk temperature.
Prw Prandtl number of the fluid, evaluated at the tube wall surface temperature.
p Pressure, Pa
Q Heat transfer rate, W
R Arc radius of the lenticular cylinder, m
Re Reynolds number
r Radial coordinate, m
ri Radius of inside circular channel, m
ro Length of outside semi-minor axis, m
T Temperature, K
t Time, s
u Velocity, m/s
.
Volumetric flow rate, m3/s
V
Greek
α Angle measured from the front stagnation point, degree
χ Shaped tube efficiency
ε Tube effectiveness, Eq. (29)
φ Angle of attack, degree
λo Outside tube length ratio of the minor-to-major axis, λ o = ro / l
λt Length ratio of the wall thickness at α=90° to the outside minor axis, λ t = t / ro
τw Wall shear stress, N/m2
ψ Performance effectiveness, defined as the ratio of heat transfer rate to fan
pumping power of an individual elliptical to that of a circular tube, Eq. (34)
Superscript
Overbar indicates time-averaged quantity
Subscript
c Refers to a circular cylinder
e Refers to an elliptical cylinder
L, D Refers to characteristic length
l Refers to a lenticular cylinder
o Refers to value at cylinder outer surface
w Refers to value at the cylinder wall
x, y Refer to values at x and y direction
α Refers to values at angle α
∞ Refers to freestream

13
 2004 Report to Supercomputer Institute

REFERENCES

1. Bigg, D. M., Stickford, G H., Talbert, S G., 1989, "Applications of Polymeric Materials for
Condensing Heat Exchangers," Polymer Engineering & Science, 29(16), pp. 1111-1116.
2. Davidson, J. H., Oberreit, D., Liu, W., and Mantell, S. C., April, 1999, "Are Plastic Heat
Exchangers Feasible for Solar Water Heaters? Part I: A Review of the Technology, Codes
and Standards, and Commercial Products," ASME/KSME/JSME/ASHRAE/JSES
International Renewable and Advanced Energy Systems for the 21st Century, CD-ROM,
RAES99-7683, Maui, Hawaii.
3. El-Dessouky, H. T. and Ettouney, H. M., 1999, "Plastic/Compact Heat Exchangers for
Single-Effect Desalination Systems," Desalination, 122(2-3), pp. 271-289.
4. Jachuck, R. J. J., Ramshaw, C., 1994, "Process Intensification: Polymer Film Compact Heat
Exchanger (PFCHE)," Chemical Engineering Research & Design, 72(A2), pp. 255-262.
5. Li, Z., Mantell, S. C., and Davidson, J. H., 2004, "Mechanical Analysis of Streamlined Tubes
with Nonuniform Wall Thickness for Heat Exchangers," Journal of Strain Analysis for
Engineering Design (Accepted).
6. Li, Z., Davidson, J. H., and Mantell, S. M., 2004, "Heat Transfer Enhancement Using Shaped
Polymer Tubes: Fin Analysis," Journal of Heat Transfer, 126(2), pp. 211-218.
7. Zukauskas, A. and Ziugzda, J., 1985, Heat Transfer of a Cylinder in Crossflow, Hemisphere
Publishing Corporation, pp. 150.
8. Delany, N. and Sorensen, N., 1953, "Low Speed Drag of Cylinders of Various Shapes,"
NACA Technical Note, (Report No. 3038).
9. Hoerner, D.-I. S. F., 1965, Fluid-Dynamic Drag, Sighard F. Hoerner, Midland Park, NJ.
10. Modi, V. J., Wiland, E., Dikshit, A. K., and Yokomizo, T., 1992, "On the Fluid Dynamics of
Elliptic Cylinders," International Journal of Offshore and Polar Engineering, 2(4), pp. 267-
280.
11. Jang, J., and Li, B., 1996, "A Numerical Analysis of Two-Dimensional Laminar Flow Over
an Elliptic Tube Bank," Proceedings of the 4th International Symposium on Heat Transfer,
Beijing, China, pp. 547-552.
12. Merker, G. P., Hanke, H., and Baehr, M, 1987, "Analogy between Momentum and Heat
Transport in Cross-Flow Tube Banks with Oval-Shaped Tubes," Warme und
Stoffubertragung (Thermo and Fluid Dynamics), 21(2-3), pp. 95-102.
13. Ruth, E. K., 1983, "Experiments on a Crossflow Heat Exchanger with Tubes of Lenticular
Shape," Journal of Heat Transfer, 105, pp. 571-575.
14. Rühlich, I., and Quack, H., 1998, "New Regenerator Design for Cryocoolers," 17th
International Cryogenic Engineering Conference, Bournemouth, England, pp. 291-294.
15. Rühlich, I., and Quack, H., 1999, "Investigations on Regenerative Heat Exchangers," 10th
International Cryocooler Conference, Monterey, California, pp. 265-274.
16. Tauscher, R., 2000, "Wärmeübergang Mit Turbulenzanregung Bei Niedrigen
Reynoldszahlen", Vollständiger Abdruck der von der Fakultät für Maschinenwesen,
Technischen Universität München.
17. Menter, F. R., 1994, "Two-Equation Eddy-Viscosity Turbulence Models for Engineering
Applications," AIAA Journal, 32(8), pp. 1598-1605.
18. Ota, T., Aiba, S., Tsuruta, T., and Kaga, M., 1983, "Forced Convection Heat Transfer from
an Elliptic Cylinder of Axis Ratio 1:2," Bulletin of Japan Society of Mechanical Engineers,
26(212), pp. 262-267.

14
 2004 Report to Supercomputer Institute

19. Ota, T. and Nishiyama, H., 1984, "Heat Transfer and Flow around an Elliptic Cylinder,"
International Journal of Heat and Mass Transfer, 27(10), pp. 1771-1779.
20. Kondjoyan, A. and Daudin, J. D., 1995, "Effects of Free Stream Turbulence Intensity on
Heat and Mass Transfers at the Surface of a Circular Cylinder and an Elliptical Cylinder,
Axis Ratio 4," International Journal of Heat and Mass Transfer, 38(10), pp. 1735-1749.
21. Behr, M., Hastreiter, D., Mittal, S., and Tezduyar, T. E., 1995, "Incompressible flow past a
circular cylinder: Dependence of the computed flow field on the location of the lateral
boundaries," Computer Methods in Applied Mechanics and Engineering, 123(1-4), pp. 309-
316.
22. Reichel, C. and Strohmeier, K., 2003, "Circular Cylinder Exposed to Cross Flow Fluid
Forces-Parameters of Influence-Limits of Numerical Models," ASME Pressure Vessels and
Piping Conference, American Society of Mechanical Engineers, Cleveland, OH, United
States, pp. 35-44.
23. Abdulhadi, M., 1989, "Boundary Layer Calculations on Cylinders of Rankine-oval Sections,"
Transactions of the Canadian Society for Mechanical Engineering, 13(3), pp. 65-68.
24. Knauss, D. T., John, J. E. A., and Marks, C. H., 1976, "Vortex Frequencies of Bluff
Cylinders at Low Reynolds Numbers," Journal of Hydronautics, 10(4), pp. 121-126.
25. Ratliff, C. L., 1992, "Vortex Shedding Frequencies of Elliptical Cylinders in the Irregular
Reynolds Number Region," Winter Annual Meeting of the American Society of Mechanical
Engineers, Nov 8-13 1992, Publ by ASME, New York, NY, USA, Anaheim, CA, USA, pp.
53-60.
26. Modi, V. J. and Dikshit, A. K., 1975, "Near Wakes of Elliptic Cylinders in Subcritical Flow,"
AIAA (American Institute of Aeronautics and Astronautics) Journal, 13(4), pp. 490-497.
27. Modi, V. J. and Wiland, E., 1970, "Unsteady Aerodynamics of Stationary Elliptic Cylinders
in Subcritical Flow," AIAA (American Institute of Aeronautics and Astronautics) Journal,
8(10), pp. 1814-1821.
28. Knoblauch, O. and Reiher, H., 1925, Handbuch d. Experimentalphysik,
Waermeuebertragung, Leipzig, pp. 189.
29. Wieselsberger, E., 1921, "Neuere Feststellungen über die Gesetze des Flüssigkeits- und
Luftwiderstandes," Physikalische Zeitschrift, 22, pp. 321-328.
30. Wu, C., Mantell, S. C., and Davidson, J. H., 2004, "Polymers for Solar Domestic Hot Water:
Long-Term Performance of PB and Nylon6,6 Tubing in Hot Water," Journal of Solar Energy
Engineering, 126, pp. 581-586.

15
 2004 Report to Supercomputer Institute

Fig. 1. Example of a “shaped” polymer tube. The outer surface of the tube is an ellipse. The
inner flow passage is circular. The non uniform wall is selected to minimize the conductive
resistance across the wall and to prevent deformation and strain failure [5].

(a)

(b)

Fig. 2. Streamlined outer profile of (a) elliptical, and (b) lenticular tubes. The outer surface of a
lenticular cylinder is formed by two symmetrically intersected arcs, each given by
(r cos α )2 + (r sin α + R − ro ) 2 = R 2 , where and λo is the length ratio of the minor to major axis of
the outer surface. A value of λo = 1 represents a circular cylinder. The cylinders become more
slender as λo is decreased.

16
 2004 Report to Supercomputer Institute

Fig. 3. Computational domain and boundary conditions of flow and heat transfer.

Fig. 4. Triangular and quad meshes for flow over an individual elliptical cylinder.

17
 2004 Report to Supercomputer Institute

0.08
Zukauskas and Ziugzda, 1985 [7]
Numerical

0.04
Cf,α

0
0 45 90 135 180

α(degree)

-0.04

Fig. 5. Comparison of predicted local friction coefficient with experiment [7] for an elliptical
cylinder with λo = 0.5 at ReD = 3200.

Zukauskas and Ziugzda, 1985 [7]

0.5
Numerical

0
Cp,α

0 45 90 135 180
α(degree)
-0.5

-1

-1.5

Fig. 6. Comparison of predicted local pressure coefficient with experiment [7] for an elliptical
cylinder with λo = 0.5 at ReD = 3200.

18
 2004 Report to Supercomputer Institute

0.8

0.6
CD

0.4
Delany and Sorensen, 1953 [8]
Zukauskas & Ziugzda, 1985 [7]
0.2 Numerical

0
0 4000 8000 12000
ReD

Fig. 7. Comparison of numerical drag coefficient with experiment [7,8] for an elliptical cylinder
with λo = 0.5.

100

Ota, et al., 1983 [17]

75 Zukauskas and Ziugzda, 1985 [7]
Numerical
NuD,α

50

25

0
0 45 90 135 180

α (degree)

Fig. 8. Comparison of predicted local Nusselt number with experiment [17] and an empirical
correlation [7] for an elliptical cylinder with λo = 0.5 at ReD = 3074.

19
 2004 Report to Supercomputer Institute

60
Knoblauch and Reiher, 1925 [28]
Ota et al., 1983 [17]

45 Zukauskas and Ziugzda, 1985 [7]

Numerical
NuD

30

15

0
0 2000 4000 6000 8000 10000
ReD

Fig. 9. Comparison of predicted average Nusselt number with empirical correlations [7,17,28]
for an elliptical cylinder with λo = 0.5.

0.7

0.6

0.5
CD
Cf, Cp, CD

0.4 Cp

0.3

0.2

0.1
Cf
0
0 2000 4000 6000 8000 10000
ReD

Fig. 10. Friction, pressure and drag coefficient of an elliptical cylinder with λo = 0.3.

20
 2004 Report to Supercomputer Institute

1.2
λo = 0.8
Circular, λo = 1.0 [29]
λo = 0.8
0.9

λo = 0.5
CD

0.6
λo = 0.3

0.3
Wieselsberger, 1921 [29]
Numerical, elliptical cylinder
Numerical, lenticular cylinder
0
5000 6000 7000 8000 9000 10000
ReD

Fig. 11. Comparison of drag coefficients of elliptical, lenticular and circular cylinders. The drag
coefficient of a circular cylinder is calculated from correlation of [29].

21
 2004 Report to Supercomputer Institute

70

60
Circular, λo = 1.0 [7]

50

40 λo = 0.8
NuD

λo = 0.5
30
λo = 0.3

20

Zukauskas and Ziugzda, 1985 [7]

10

0
0 2000 4000 6000 8000 10000
ReD

(a)
70
Zukauskas and Ziugzda, 1985 [7]
60

Circular, λo = 1.0 [7]

50
λo = 0.8
40 λo = 0.5
NuD

λo = 0.3
30

20

10

0
0 2000 4000 6000 8000 10000
ReD

(b)

Fig. 12. Average Nusselt number of streamlined cylinders with λo = 0.3, 0.5 and 0.8 at 500 ≤
ReD ≤ 104, (a) elliptical, (b) lenticular. The Nusselt number of the circular cylinder is calculated
from correlations provided by [7].

22
 2004 Report to Supercomputer Institute

2
λo = 0.3

1.5 λo = 0.5
λo = 0.8

1 Circular
ε

λo = 0.8
λo = 0.5
λo = 0.3
0.5

0
0 2000 4000 6000 8000 10000
ReD

Fig. 13. Predicted shaped tube effectiveness of elliptical cylinders with λo = 0.3, 0.5, and 0.8 for
cross flow at 500 ≤ ReD ≤ 104. Solid lines and dashed lines represent nylon tubes and aluminum
tubes, respectively.

λo = 0.3
4

3
λo = 0.5
ψ

2
λo = 0.3 λo = 0.8
λo = 0.5
1
λo = 0.8

0
0 2000 4000 6000 8000 10000
ReD

Fig. 14. Performance of elliptical cylinders with λo = 0.3, 0.5, and 0.8 for cross flow at 500 ≤
ReD ≤ 104. Solid lines and dashed line represent nylon tubes and aluminum tubes, respectively.
The horizontal constant line represents performance of a circular tube.

23