CORE Metadata, citation and similar papers at core.ac.

uk
Provided by Brunel University Research Archive

Modelling and Simulation Techniques for Forced Convection Heat

Transfer in Heat Sinks with Rectangular Fins

Hussam Jouhara *

School of Engineering and Design, Brunel University, Uxbridge, Middlesex, UB8 3PH, UK,
Tel: +44 (0) 1895 267656, Fax: +44 (0) 1895 256392, Email: hussam.jouhara@brunel.ac.uk

Brian P. Axcell
Institut Catholique d’Arts et Métiers, 6 rue Auber, 59046 Lille Cedex, France
Phone: +44 (0) 1925 267540, Fax: +44 (0) 1925 600139, E-mail: brian.axcell@gmail.com

*
Corresponding Author
Abstract

This paper provides a comprehensive description of the thermal conditions within a heat sink
with rectangular fins under conditions of cooling by laminar forced convection. The analysis,
in which increasing complexity is progressively introduced, uses both classical heat transfer
theory and a computational approach to model the increase in air temperature through the
channels formed by adjacent fins and the results agree well with published experimental data.
The calculations show how key heat transfer parameters vary with axial distance, in
particular the rapid changes in heat transfer coefficient and fin efficiency near the leading
edges of the cooling fins. Despite these rapid changes and the somewhat ill-defined flow
conditions which would exist in practice at the entry to the heat sink, the results clearly show
that, compared with the most complex case of a full numerical simulation, accurate
predictions of heat sink performance are attainable using analytical methods which
incorporate average values of heat transfer coefficient and fin efficiency. The mathematical
modelling and solution techniques for each method are described in detail.

Keywords: Fins, heat sink, laminar heat transfer, parallel flow.

1 Introduction
Electronic components invariably generate unwanted heat during operation and as
electronic systems have become both more compact and more powerful the problem of
efficient and reliable heat removal, which is needed for safe operation of the component, has
become onerous. Electronic components and assemblies tend to be of a small scale and they
are typically cooled by air flowing at moderate velocities. The combination of small
dimensions, the use of air as the cooling fluid and low velocities normally results in laminar
convection and hence correspondingly low values for heat transfer coefficients.
In general, when gas cooling needs to be enhanced, it is achieved by the use of cooling
fins, which increase the surface area available for heat transfer. Extended heat transfer
surfaces are often used with electronic systems with fins providing a heat sink for the thermal
loading. A typical arrangement for a heat sink with rectangular fins and the air flow direction
through the channels are shown in Figures 1 and 2. Heat generated by the electronic
component is conducted to a base plate of length L, to which cooling fins of height H and
width t are attached. Both sides of each fin participate in the cooling process, with convection
taking place in the channels of width b formed by adjacent fins.
Although various flow configurations and fin geometries are possible for the cooling
process, to date most investigations have considered rectangular fins. Teertstra et al. [1]
examined air flow parallel to the base plate; in this case the heat transfer may therefore be
represented by convection in a parallel plate channel. Duan and Muzychka [2] considered an
impinging flow on the central region of the fin assembly, which then turned through 90º and
exited outwards in both directions through the rectangular channels formed by the cooling
fins. Shin [3] also investigated rectangular fins and included cooling by natural convection.
Probably the most comprehensive study of the topic to date is provided by Bar-Cohen and
Iyengar [4]. The authors considered not only the thermal performance of a heat sink with
rectangular fins but also the pumping power required during operation and the material
needed for construction in order to provide a “least material” and “least energy” design.
None of the papers published to date, however, appears to have given a
comprehensive description of the heat transfer within the heat sink. The heat transfer
coefficient and fin efficiency will vary along the flow path because of entry effects for the
thermal and momentum boundary layers in the inlet region; in the entrance region heat
transfer coefficients derived for fully-developed flow are not applicable. In addition the flow
will be ill-defined at the entry because of disturbances caused by the bluff nature of the
leading edges of the fins. There will also be some axial heat conduction in the fins in the
upstream direction.
This paper aims to show how heat transfer principles can be applied to rectangular
heat sinks for electronic components. Initially the simplest scenario will be considered, in
which the array of cooling fins is isothermal. Then thermal conditions within the heat sink
will be modelled by a single, average fin efficiency and an average heat transfer coefficient.
Greater complexity will be added using numerical integration by allowing both heat transfer
coefficient and fin efficiency to vary along the flow path. Finally the effects of longitudinal
conduction in the fins and the effect of flow disturbance at the leading edges of the fins will
be modelled in a three-dimensional CFD analysis with a finite element treatment of the
cooling fins. It is shown however that the simplest engineering approach using analytical
expressions for uniform heat transfer coefficient and fin efficiency gives good accuracy for
engineering purposes.

2 Test cases for calculations

In order to demonstrate the efficiency of the different modelling and solution
techniques, the geometrical arrangement of Teertstra et al. [1] is considered: two identical
heat sinks of the form shown in Figure 1 were tested back-to-back. Each heat sink had a
length L = 115 mm and contained 15 fins. The height H of each fin was 49 mm, the thickness
t was 1.25 mm and the gap between fins b was 2.18 mm. The heat sinks were presumably
constructed from aluminium because the authors specified a thermal conductivity for the fins
ks of 200 W/m K.
The heat sink assembly was installed in a wind tunnel with a shroud preventing the air
flow from by-passing the assembly. The air approach velocity in the wind tunnel was varied
from 1 m/s to 8 m/s; the mean velocity in the flow passages between fins was 87.7% higher
because of a reduced flow area through the heat sinks.
Teertstra et al. [1] performed tests for power inputs from 100 W to 500 W. When
expressed in dimensionless form, their experimental results did not vary significantly with
power input. This is because heat sink temperatures are a linear function of power input and
experimental Nusselt numbers will not change with power level if the variations in transport
properties for air with temperature are small and the contribution of thermal radiation is
negligible. Consequently the calculations were restricted for the given range of velocities to a
single base plate temperature of 50ºC for an air inlet temperature of 20ºC. Air properties were
evaluated at 300 K (27ºC).

3 Theory
(i) Idealised case
Initially it is assumed that the fin surfaces are at a uniform temperature Ts equal to that
of the base plate. The influence of a reduction in fin temperature away from the base plate
will be accommodated later, first by the use of a traditional fin efficiency η and then in a finite
element study. In the analytical treatment of the heat sink, heat transfer correlations for
convection from surfaces of uniform temperature will be utilised because this boundary
condition represents most closely the case under examination.
Consider a flow in the channel between two fins having a mean velocity um and a local
bulk mean temperature Tb. In a length of the flow channel dx the bulk mean temperature
increases by an amount dTb given by:
 q P dx = ρ u m A c p dTb (1)

where P is the perimeter of the heated surface, A is the flow area, ρ and cp are respectively the
density and specific heat capacity of the air flow and the heat flux q is related to the local
temperature difference Ts - Tb by the local heat transfer coefficient h:
q = h (Ts − Tb ) (2)

Denoting the local temperature difference by the symbol θ; (θ =Ts - Tb), equation (1) can be
rewritten
dθ hP
=− dx (3)
θ ρ um A c p

Introducing a hydraulic diameter for the flow defined by

Dh = 4 A / P (4)

the differential takes the form:

dθ 4h
=− dx (5)
θ ρ u m c p Dh

It should be noted that for a flow channel in which H/b>>1, the hydraulic diameter Dh is
simply twice the gap 2b and this value will be used in all of the calculations. The heat
transfer from the base plate will be added to that from the side walls.

Finally we introduce a dimensionless length scale x* defined by

x
x* = (6)
Re Pr D h

where Re and Pr are the Reynolds number and Prandtl number, respectively.

h ⋅ Dh
Equation (5) can then be written in terms of a local Nusselt number, Nu = :
k

dθ
= −4 Nu dx * (7)
θ

Integrating from the entrance of the flow channel where x* = 0 to a particular value of x*:

x*
ln (θ / θ 0 ) = −4 ∫ Nu dx * (8)
0

or
θ x*
= exp{−4 ∫ Nu dx *} (9)
θ0 0

Since the mean value of the Nusselt number for a length x can be defined by
 1 x*
x * ∫0
Nu m = Nu dx * (10)

equation (9) can be written in the form:

θ
= exp{ − 4 Nu m x*} (11)
θ0

The rate of heat removal from all channels formed by adjacent fins can then be calculated
from the change in the temperature difference θ:

Q = N ρ u m A c p (θ 0 − θ L ) (12)

It is also possible to define the rate of heat removal from the heat sink to the temperature
difference at inlet:

Q = N ρ u m A c p θ 0 [1 − exp(− 4 Nu m L*)] (13)

in which L* corresponds to the value x* at the end of the flow channels.

This expression also provides a limit for the maximum rate of heat removal for a given flow,
since if the length is sufficiently large, the exponential term is negligible.
In order to use equation (13) one requires either a mean Nusselt number or an
expression from which the mean value can be evaluated. The development of the heat
transfer coefficient in such passages however is a complex process.
When the cooling flow enters a channel formed by two adjacent fins, thermal and
momentum boundary layers begin to grow on the heat transfer surfaces. Close to the leading
edge local heat transfer coefficients have high values because small thermal resistance to the
convective process is presented by the thin boundary layers. At the leading edge itself the
heat transfer coefficient is, in theory, infinite. As the boundary layers grow in size the heat
transfer coefficient decreases with distance; in a region in which the boundary layer
thicknesses are small compared with the gap between adjacent fins it may be determined from
[5]:

Nu x = h x / k = 0.332 Re x
1/ 2
Pr 1 / 3 (14)

in which the length scale in the Nusselt number and Reynolds number is x, the distance from
the leading edge, and the heat transfer coefficient is based on θ0, the difference between the
fin surface temperature and the air inlet temperature.
As the boundary layer thicknesses become significant in terms of the gap b, the flow
takes on more of the characteristics of an internal flow. Eventually the boundary layers meet
at the centre of the channel and as soon as the shape of velocity and temperature profiles cease
to change, the local heat transfer coefficient is given by a fully-developed value, defined in
dimensionless terms by the following simple equation [6]:

Nu = 7.54 (15)
It can be seen from equation (11) that, in the region of fully-developed heat transfer, when the
Nusselt number is a constant, the bulk mean temperature for the air, Tb, approaches Ts, the
value of the fin surface temperature, exponentially. Hence strictly the air temperature reaches
the surface temperature only in a flow channel of infinite length.
No analytical solutions for mean Nusselt number cover the full range of heat transfer
development from boundary layer flow at the leading edge to fully-developed internal heat
transfer, many hydraulic diameters downstream. However Hwang and Fan [7] have
performed a numerical analysis for laminar heat transfer in a parallel-sided duct for uniform
wall temperature with simultaneously developing velocity and temperature profiles. The
following expression fits their data for mean Nusselt number to an accuracy of 3% for
0.1<Pr<1000:
0.024 x * −1.14
Nu m = 7.55 + (16)
1 + 0.0358 Pr 0.17 x * − 0.64

This expression may be used to calculate the rate of heat removal from the heat sink for this
idealised case using equation (13) above.

(ii) Approximate analysis: heat transfer with non-uniform wall temperatures

Conduction heat transfer from the base plate into the cooling fins takes place only in
the presence of a temperature gradient. Hence fin temperatures decrease with increasing
distance from the base plate and the idealised thermal output provided above will
overestimate the thermal performance.
Heat transfer from a cooling fin to the surrounding air is often calculated using a fin
efficiency η, which relates the rate of heat transfer from the fin to that which would occur for
a fin at a uniform temperature equal to the base plate. The fin efficiency for rectangular fins
with zero heat transfer at the tips is given by the following expression [8]:

tanh (mH )
η= (17)
mH
where the parameter m is defined by
2h
m2 = (18)
ks t

It is possible to estimate the rate of heat transfer from the heat sink using a mean heat
transfer coefficient and a single fin efficiency based on this heat transfer coefficient, even
though it is noted that both the heat transfer coefficient and fin efficiency vary along the flow
passage.
For a small element of the flow passage between the two adjacent fins of length dx the
rate of heat transfer is given by

δQ1 = h θ ( b + 2 Hη ) dx (19)

Equating this to the enthalpy increase of the flow

h θ ( b + 2 H η ) dx = − ρ u m b H c p dθ (20)

Noting that h = Nu k / Dh, ρ um = Re μ / Dh and cP = Pr k / μ, and setting Dh = 2b, the

equation may be written in dimensionless form:
 dθ b Nu
= −4 (η + ) dx (21)
θ 2 H Re Pr D h

which may be integrated for a constant value of η to give the temperature difference θ at
outlet (x = L)
θL b
= exp {−4 (η + ) Nu m x * } (22)
θ0 2H

from which the heat removal from the heat sink may be calculated using equation (12).

(iii) Analysis using numerical integration along the flow passage

As noted above, both the heat transfer coefficient h and the fin efficiency η vary along
the flow. Consequently better accuracy should be achieved by calculating the increase in bulk
temperature using local values for the two parameters. Now, the mean Nusselt number Num
can be defined in terms of the local Nusselt Number Nu in equation (10). It follows that an
expression for the local Nusselt number can be obtained from the equation for the mean
Nusselt number by differentiation:

d d
Nu = [ x * Nu m ] = Nu m + x * [ Nu m ] (23)
dx * dx *

Shah and Bhatti [9] performed the differentiation using equation (16) and obtained the
following expression for local Nusselt number:

0.024 x *−1.14 [0.0179 Pr 0.17 x *−0.64 − 0.14]

Nu = 7.55 + (24)
[1 + 0.0358 Pr 0.17 x *−0.64 ]2

Local heat transfer coefficients and local fin efficiencies may be calculated from this
expression and again, after including the contribution of the base plate to convective heat
transfer, the increase in bulk air temperature can be evaluated in steps along the flow passage.
Finally the overall rate of heat removal from the flow passage can again be calculated using
equation (12). This essentially was the approach adopted by Bar-Cohen and Iyengar [4] for
the thermal analysis in their paper.

(iv) Finite element/ CFD study

The computational study was performed using Ansys CFX and Ansys Multiphysics software.
Symmetry considerations allowed a model to be constructed using a single cooling fin split
along the centre line. The boundary conditions on the surfaces, the geometry and the domains
can be seen Figure 3.
The computational model comprises two parts: the solid model representing half of a
fin (Figure 3), and the fluid model to represent the flow through the heat sink. The fluid
model consisted of 401420 hexahedral elements with 100 wedge-shaped elements and the
solid mesh consisted of 218270 hexahedral elements and 460 wedge elements. Grid
independence was achieved by increasing the number of the elements and plotting the
convergence of certain parameters of interest such as free surface velocity, and also tracking
global parameters such as total rate of heat transfer through the system. The code solved the
two dimensional equations for continuity, momentum and energy as follows:

∂u ∂v ∂w
+ + =0 (25)
∂x ∂y ∂z

 ∂u ∂u ∂u  ∂P
ρ  u + v + w  = − + µ∇ 2u (26)
 ∂x ∂y ∂z  ∂x

 ∂v ∂v ∂v  ∂P
ρ  u + v + w  = − + µ∇ 2v (27)
 ∂x ∂y ∂z  ∂y

 ∂w ∂w ∂w  ∂P
ρ  u +v + w  = − + µ∇ 2 w (28)
 ∂x ∂y ∂z  ∂z

∂T ∂T ∂T
u +v +w = α∇ 2T (29)
∂x ∂y ∂z

Calculations were made for approach velocities of 2 m/s and 4 m/s, which would
respectively generate Reynolds numbers in the flow passages of 1044 and 2088. These values
would give laminar flow within the heat sink (see section 5 below). It should be noted in any
case that cooling systems are normally designed to operate with Reynolds numbers no greater
than about 1000 because of the need to avoid excessive noise.
The calculations were made with a maximum residual of 10-5 and a maximum of 100
iterations. Sample results for the temperature and velocity profile can be seen in Figures 4
and 5. Figure 5 shows the edge effect on the approaching flow (the disturbance).

4 Results

Results were obtained for each of the approaches outlined in the theory section above.
First base-line calculations were performed assuming the cooling fins were at the base plate
temperature and hence fin efficiencies were 100% (Qideal). Then a uniform fin efficiency was
employed, based on the mean Nusselt number for the flow in the channel formed between
adjacent fins, giving Qapprox. Next both heat transfer coefficient and fin efficiency were
allowed to vary along the heated length and local conditions established from numerical
integration. The thermal output of the heat sink Qnum could be found by integrating the rate of
heat loss from each element or from an overall enthalpy balance. Step lengths of 1mm and
5mm were investigated, corresponding to axial distances of approximately 0.2 Dh and Dh.
There was less than a 1% difference in the cooling performance of the heat sink for the two
step lengths in every case and hence the use of a 1 mm step length was considered sufficiently
small. Finally a CFD study was conducted using commercial software for two of the flow
rates to give values for QCFD.
The predicted thermal performance of a single heat sink using the four methods of
calculation is shown in Table 1 for approach velocities from 1 to 8 m/s. The rate of heat
transfer Qnum calculated by numerical integration along the flow path can be seen to be lower
than Qapprox, the value found using average values of the Nusselt number and fin efficiency,
by an amount which varied from almost zero at the lowest flow rate to about 4% at the highest
flow rate. The results from the computational study are marginally higher than those
calculated using from the approximate calculations, the discrepancies being 3% for the lower
flow rate and 1% for the higher flow rate.
Other heat transfer parameters are also given in Table 1: the mean velocity in the flow
passage, the Reynolds number for the flow passage, the mean Nusselt number, the fin
efficiency based on the mean Nusselt number and the air outlet temperature found by
numerical integration. Figure 6 shows the output from the heat sink graphically as a function
of mean velocity in the flow passages between adjacent fins. Figure 7 shows how θ, the
temperature difference between wall and bulk calculated from Qnum, decreases along the flow
channel for various flow rates. Even at the lowest flow rate, the air temperature at outlet is
substantially lower than the base plate temperature.

5 Discussion
For each theoretical approach it was assumed that the flow channels formed by the
base plate and cooling fins were completed by a cover at the fin tips. In practice this is
necessary to prevent cooling flow leakage from the heat sink and this also represented the
system investigated experimentally by Teertstra et al. [1].
Provided that conduction in the fins was modelled, either using fin efficiencies or
finite elements, there was reasonable agreement between the predictions for the various
approaches. The increasing discrepancy with flow rate between Qapprox and Qnum can be
understood from the variations of Nusselt number and fin efficiency calculated using
equations (24) and (17), respectively, and shown in Figures 8 and 9, respectively. The higher
flow rates generate higher heat transfer coefficients near the leading edges of the fins, which
in turn produce significantly lower fin efficiencies.
Some comments must be made about the accuracy of the modelling for flows in real
heat sinks. Firstly, thermal radiation from the heat sink has been neglected. This is probably a
reasonable assumption; both emissivity and shape factor for the aluminium cooling fins would
be relatively low and Teertstra et al. [1] demonstrated experimentally that thermal radiation
contributed no more than 1% to overall heat transfer. Simple thermal radiation calculations
for a grey body with black surroundings using representative emissivities (assuming
isothermal boundaries for the heat sink at the base plate temperature) indicate a rate of
radiation heat transfer of less than 2% of the convective heat transfer.
Free convection effects have also been neglected. The reason for this is that free
convection heat transfer coefficients for surfaces within the flow channel would typically be
six or seven times lower than that for forced convection. Heat transfer coefficients in laminar
mixed forced-and-free convection are often combined [10,11] using an equation of the
following form:

hmixed convection = h forced convection + h free convection

n n n
(30)

where the index n typically has a value of between 3 and 7. For such a combination it is
easily shown that buoyancy effects could not possibly influence the rate of heat transfer in
this application. The influence of free convection can also be assessed from the Richardson
number:
Ri = Gr / Re 2 (31)
In conditions of mixed forced and free convection Ri ~ 1, see for example [12]. Even at inlet
for the lowest flow rate and hence the lowest Reynolds number combined with the maximum
value of the Grashof number, the Richardson number is only of the order of 10-3. Hence we
may safely neglect the free convection term in equation (30).
Other assumptions are less easy to justify. The heat transfer correlations used to
calculate local Nusselt numbers in the analytical treatment, equations (16) and (24), were
based on an analysis of simultaneously developing velocity and temperature in a flow with
uniform wall temperatures, which had uniform velocity and temperature at the inlet. The
velocity profile at inlet to the heat sink would almost certainly in reality be ill-defined; the
blunt leading edges of the cooling fins would cause flow disturbance and modify the rate of
heat transfer locally, although the effect would diminish downstream. Also, as demonstrated
clearly by values obtained for the fin efficiency near the leading edge, the wall temperatures
for the flow passage would not have been uniform. The finite element treatment of the cooling
fins allows for heat conduction in the axial upstream direction, which is ignored when using a
simple fin efficiency; rising wall temperatures in the direction of flow tend to increase heat
transfer coefficients [13,14]. These two effects probably explain why the values for QCFD are
marginally higher than those for Qapprox and Qnum.
In addition, all of the theoretical approaches have assumed laminar flow, although the
Reynolds number for the higher flow rates analysed would probably have been in the
transitional region rather than the laminar region. There is little information about the critical
Reynolds number for transition from laminar flow to turbulent flow in a parallel-sided duct; a
range of 2200 to 3400 has been suggested [15]. Half of the flows considered exceeded the
lower limit of this range and two flows exceeded the upper limit.
Nevertheless, the predicted results are in very good agreement with the experimental
data of Teertstra et al. [1]. Their experimental data were presented in graphical form but
using different dimensionless groups. Teertstra et al. [1] based their Nusselt number Nub on
the inlet temperature difference and took b not Dh to be the characteristic length. In their
paper the flow was characterised by Re*b, a group identical to the reciprocal of 4 L* Pr.
When our predictions are represented by the same dimensionless groups and a curve is drawn
through them, the line follows the variation of the experimental data very precisely, Figure
10.
Teertstra et al. [1] modelled the overall rate of heat transfer successfully by combining
two expressions for mean Nusselt number empirically. The first expression described heat
transfer at entry. The second (in the present authors’ opinion described erroneously as “fully-
developed”) was based on the maximum rate of heat removal for a given flow (i.e. equation
(13) with the exponential term negligible), when the air outlet temperature would be equal to
the temperature of the base plate.
Various researchers have used the limit of the maximum rate of heat removal from a
heat sink but it is not a particularly useful concept. As can be seen in Table 1 and Figure 7 the
air outlet temperature did not approach the base plate temperature for the heat sink considered
even for the lowest flow rate. Further, a heat sink with an air outlet temperature equal to the
surface temperature does not represent an optimum design because much of the cooling fin
would be operating with a small temperature difference and hence a low convective heat flux.
Hence the cooling performance of much of the fin material would be poor and much of the fin
material would therefore be wasted. It is also misleading to describe such a heat transfer
condition as fully developed, since this state (a constant relationship between heat flux and
temperature difference) is normally reached well before the temperature difference and heat
flux decay to zero.
The present study provides a much more rigorous analysis and makes an interesting
observation missed by previous researchers. Although increased complexity has
progressively been introduced into the study, it is apparent that classical analytical heat
transfer theory provides very acceptable engineering accuracy with good agreement with both
experimental data and a full numerical simulation. The approach employs average values of
heat transfer coefficient and fin efficiency, which are obtained from analytical expressions,
and as a consequence the thermal performance of the heat sink can be assessed very rapidly.
It is not necessary to resort to an empirical combination of limiting heat transfer cases,
namely, heat transfer in the entry region and heat transfer for a very long heat sink [1]. There
is also no need for a numerical integration along the flow passage using local values of heat
transfer coefficient and fin efficiency [4].

6 Conclusions
Heat transfer conditions in the flow channels formed by the rectangular fins of a heat
sink have been analysed using analytical, numerical and computational methods of varying
complexity.
Initially results are presented for idealised fins which are 100% efficient using an
expression for mean Nusselt number to describe heat transfer conditions from entry to exit.
Next a uniform fin efficiency based on this mean Nusselt number is utilised. More detail is
then introduced by allowing both Nusselt number and fin efficiency to vary along the flow
passage and by calculating air temperatures using numerical integration. Finally the effects of
flow disturbance at entry to the heat sink and axial conduction in the cooling fins are
modelled in a CFD study with a finite element treatment of the fins.
Good agreement has been obtained between all heat transfer predictions and
experimental data of Teertstra et al. [1] except, of course, for those made assuming the fins to
be at uniform temperature. Although the calculations show that heat transfer coefficients and
fin efficiencies vary substantially along the flow path, good engineering accuracy for heat
sink performance can be obtained for laminar flow using calculations in which mean values
for the Nusselt number and fin efficiency are calculated from analytical expressions.

Acknowledgement

This work was part of a project funded by Solas Technology Limited, Ireland.

Nomenclature

a width of base plate, m

A flow area of channel, m2
b distance between fins, m
cp specific heat capacity, J/kg K
Dh hydraulic diameter 4A/P, m
Gr Grashof number β (Ts – Tb) g Dh3/ν2, dimensionless
h heat transfer coefficient q/θ, W/m2K
hm mean heat transfer coefficient, W/m2K
H height of fins, m
k air thermal conductivity, W/m K
ks thermal conductivity of fin material, W/m K
L length of flow channel, m
L* L/(Re Pr Dh), dimensionless
m exponent in fin efficiency, m-1
N number of channels in heat sink
Nu Nusselt number h Dh/ k, dimensionless
Nub Nusselt number based on the distance between fins and the inlet temperature
difference Q1 b/(2 k L H θ0), dimensionless
Nux Nusselt number based on axial position h x/ k, dimensionless
Num mean Nusselt number hm Dh/ k, dimensionless
P perimeter of flow channel, m
Pr Prandtl number, dimensionless
Q rate of heat removal from heat sink, W
Q1 rate of heat transfer for a single channel, W
q heat flux, W/m2
Re Reynolds number um Dh/ν, dimensionless
Re*b Reynolds number as defined by Teerststra et al. [1] um b2/L ν, dimensionless
Rex Reynolds number based on axial position um x/ν, dimensionless
Ri Richardson number Gr/Re2, dimensionless
Tb bulk temperature, ºC
Ts surface temperature for base plate, ºC
t thickness of cooling fin, m
Ua approach velocity, m/s
um mean velocity in flow channel, m/s
x axial position in flow channel, m
x* x/(Re Pr Dh), dimensionless
β coefficient of expansion, 1/K
η fin efficiency, dimensionless
ν kinematic viscosity, m²/s
θ temperature difference Ts - Tb, K
ρ air density, kg/m3

Subscripts
approx calculations made using a fin efficiency based on Num
CFD calculations made using a commercial CFD package
ideal calculations made assuming a fin efficiency of 100%
L value at x = L
num numerical calculations with variable Nu and η
0 value at x = 0

References

1. Teertstra, P., Yovanovich, M.M., Culham, J.R., and Lemczyk, T., 1999, “Analytical
Forced Convection Modeling of Plate Fin Heat Sinks”, Proc. 15th IEEE Semi-Therm.
Symposium, pp 34-41.
2. Duan, Z., and Muzychka, Y.S., 2006, “Experimental Investigation of Heat Transfer in
Impingement Air Cooled Plate Fin Heat Sinks”, Trans. ASME J. Electronic Packaging,
128, pp412-418.
3. Shin, D., 1998, “Thermal Design and Evaluation Methods for Heat Sink”, Tech. Report.
7th International FLOTHERM User Conference Verona, Italy - T226
4. Bar-Cohen, A., and Iyengar, M., 2003, “Least-energy Optimization of Air-cooled Heat
Sinks for Sustainable Development”, IEEE Transactions on Components and Packaging
Technologies, 26, pp16-25.
5. Kreith, F., and Bohn, M.S., 1993, Principles of Heat Transfer, 5th edition, West Publishing
Company, p. 270.
6. Shah, R.K., and Bhatti, M.S., 1987, “Laminar Convective Heat Transfer in Ducts”, in
Kakaç, S., Shah, R.K., and Aung, W., Handbook of Single-Phase Convective Heat
Transfer”, John Wiley & Sons, page 3.31.
7. Hwang, C.L., and Fan, L.T., 1964, “Finite Difference Analysis of Forced Convection Heat
Transfer in Entrance Region of a Flat Rectangular Duct”, Appl. Sci. Res., A13, pp401-
422.
8. Chapman, A.J., 1987, “Fundamentals of Heat Transfer”, Macmillan Publishing Company,
p. 71.
9. Shah, R.K., and Bhatti, M.S., 1987, “Laminar Convective Heat Transfer in Ducts”, in
Kakaç, S., Shah, R.K., and Aung, W., Handbook of Single-Phase Convective Heat
Transfer”, John Wiley & Sons, p. 3.42
10. Chen, T.S., and Armal, B. F., 1987, “Mixed Convection in External Flows”, in Kakaç, S.,
Shah, R.K., and Aung, W., Handbook of Single-Phase Convective Heat Transfer”, John
Wiley & Sons, pp 14-1 - 14-35.
11. Aung, W., 1987, “Mixed Convection in Internal Flows”, in Kakaç, S., Shah, R.K., and
Aung, W., Handbook of Single-Phase Convective Heat Transfer”, John Wiley & Sons, pp
15-1 - 15-51.
12. Lienhard IV, J.H. and Lienhard V, J.H., 2001, A Heat Transfer Textbook 3rd edn.,
Phlogiston Press, Cambridge, Massachusetts, USA, p.408.
13. Walker, V., and Rishehri, S.A., 1972, “Forced Convection in Parallel and Tapered
Passages: the Importance of the Heat Flux Boundary Condition”, Int. J. Heat Mass
Transfer, 15, pp 781-788.
14. Axcell, B.P., Tonge, D.A., and Jemsson, T., 1988, “Turbulent Forced Convection in a
Pipe with an Exponential Wall Heat Flux”, in Shah, R.K., Ganić, E.N., and Yang, K.T.,
Experimental Heat Transfer, Fluid Mechanics, and Thermodynamics, Elsevier, pp808-
815.
15. Shah, R.K., and Bhatti, M.S., 1987, “Turbulent and Transition Flow Convective Heat
Transfer in Ducts”, in Kakaç, S., Shah, R.K., and Aung, W., Handbook of Single-Phase
Convective Heat Transfer”, John Wiley & Sons, p.4.59
List of Figures

Figure 1: Arrangement of the heat sink

Figure 2: Flow through the channel

Figure 3: The studied model with boundary conditions

Figure 4: Temperature and velocity profiles along a central plane through the channel between

two fins, Ua=2 m/s

Figure 5: Velocity profile and the flow disturbance at the entrance region of the channel

(Ua=4 m/s)

Figure 6: Variation of Qideal, Qapprox, Qnum and QCFD with mean velocity

Figure 7: Variation of θ/θo against x/L for various flow rates

Figure 8: Variation of Nu against x/L for various flow rates

Figure 9: Variation of η against x/L for various flow rates

Figure 10: Comparison between the approximate analysis and the data of Teertstra et al. [1]
List of Tables:

Table 1: Thermal Performance of Heat Sink

Figure 1: Arrangement of the heat sink
 4
3.5
3
2.5
2
Nub
1.5
1 Experimental Data from [1]
0.5 Approximate Analysis
0
0 5 10 15 20 25 30 35 40 45

Re*b

Figure 10: Comparison between the approximate analysis and the data of Teertstra et al. [1].
Figure 2: Flow through the channel
I: Air (Fluid) Domain of the CFX analysis onto which boundary II: Fin (Solid) Domain of the CFX analysis onto which
conditions were applied. boundary conditions were applied.

IV: Cross section from the

fin model (the solid) with
boundary conditions

III: Profile view of the

(i): Liquid/Solid interface solid and fluid
(a): Adiabatic domains interacting in
(s): Symmetry the CFX analytical
(h): Surface at base environment.
temperature (isothermal)

Figure 3: The studied model with boundary conditions

Figure 4: Temperature and velocity profiles along a central plane through the
channel between two fins, Ua=2 m/s
Figure 5: Velocity profile and the flow disturbance at the entrance region of
the channel (Ua=4 m/s)
 300.0

250.0

200.0
Q, W

150.0

100.0 Qapprox
Qideal
50.0 Qnum
Q_CFD
0.0
0 2 4 6 8 10 12 14 16
Mean Velocity, um, m/s

Figure 6: Variation of Qideal, Qapprox, Qnum and QCFD with mean velocity
 1.00

0.75
θ / θ0

0.50

Re=522
0.25
Re=1566
Re=2609
Re=4175
0.00
0.0 0.2 0.4 0.6 0.8 1.0
x/L

Figure 7: Variation of θ/θo against x/L for various flow rates

 60
Re=522
Re=1566
50 Re=2609
Re=4175
40
Nu

30

20

10

0
0.0 0.2 0.4 0.6 0.8 1.0
x/L
Figure 8: Variation of Nu against x/L for various flow rates
 0.80

0.75

0.70

0.65

0.60
η

0.55

0.50 Re=522
Re=1566
0.45 Re=2609
0.40
Re=4175

0.35

0.0 0.2 0.4 0.6 0.8 1.0

x/L
Figure 9: Variation of η against x/L for various flow rates
 Table 1: Thermal Performance of Heat Sink
Ua um (m/s) Re Num Qideal ηm Qapprox Qnum QCFD TL
(m/s) (W) (W) (W) (W) (ºC)
1 1.88 522 7.96 96.3 0.776 89 89 45.1
2 3.75 1044 8.38 150.7 0.676 131 129 135 38.2
3 5.63 1566 8.79 184.2 0.759 154 151 34.2
4 7.51 2088 9.18 208.7 0.751 170 166 172 31.7
5 9.39 2610 9.56 228.5 0.744 183 178 30.0
6 11.3 3132 9.93 245.5 0.737 193 187 28.8
7 13.1 3653 10.28 260.7 0.731 203 196 27.9
8 15.0 4175 10.62 274.6 0.725 211 203 27.1