Applied Thermal Engineering 50 (2013) 352e360

Contents lists available at SciVerse ScienceDirect

Applied Thermal Engineering

journal homepage: www.elsevier.com/locate/apthermeng

Investigation of heat transfer characteristics in plate-fin heat sink

Han-Taw Chen*, Shih-Ting Lai, Li-Ying Haung
Department of Mechanical Engineering, National Cheng Kung University, 1 University Road, Tainan , Taiwan ROC

h i g h l i g h t s

< A few researchers used the present methods to perform this study.
< The present results agree with those obtained from the existing correlations.
< Some numerical results may not satisfy physical phenomena.
< Three-dimensional fluid flow characteristics can be obtained.

a r t i c l e i n f o a b s t r a c t

Article history: The present study applies the inverse method in conjunction with the experimental temperature data to
Received 16 February 2012 investigate the accuracy of the heat transfer coefficient on the fin in the plate-fin heat sink for various fin
Accepted 25 August 2012 spacings. The commercial software is applied to solve the governing differential equations with the RNG
Available online 31 August 2012
ke3 model in order to obtain the heat transfer and fluid flow characteristics. Under the assumption of the
non-uniform heat transfer coefficient, the entire fin is divided into several sub-fin regions before per-
Keywords:
forming the inverse scheme. The average heat transfer coefficient in each sub-fin region is assumed to be
Inverse and numerical methods
unknown. Later, the present inverse scheme in conjunction with the experimental temperature data is
Heat transfer characteristics
Plate-fin heat sink
applied to determine the heat transfer coefficient and fin efficiency. In order to determine a more reliable
heat transfer coefficient, a comparison between the present inverse and numerical results and those
obtained from the existing correlations will be made. The numerical fin temperatures will also be
compared with the experimental data.
Ó 2012 Elsevier Ltd. All rights reserved.

1. Introduction pressure drop, and a simpler structure should be considered. The

fins in the heat sink are always applied to increase the heat flow per
Due to the rapid development of electronic technology, elec- unit of basic surface. However, it is necessary to note the interac-
tronic devices are often found in our daily life. However, the size tions between the local heat transfer and flow distribution between
and weight of electronic components must be reduced, and the two fins in designing the heat sink. It was known that most of
heat flux per unit area increased dramatically. Under the circum- previous works about the effect of the fin spacing in the heat sink
stance, the working temperature of electronic components may were limited to the experiment or the numerical analysis. More-
exceed the desired temperature level. Thus promoting the heat over, the measurement of the local heat transfer coefficient on the
transfer rate and maintaining the die at the desired operating plate fin under steady-state heat transfer conditions may be very
temperature have played an important role in insuring a reliable difficult to perform because the local fin temperature and local heat
operation of electronic components. This implies that the proper flux must be required [1]. Reliability is also an important concept in
design of the heat sinks has gradually become attractive for these engineering design, and the use of reliable components enables the
applications because it can provide a more economical solution to designers to utilize more sophisticated techniques to improve the
the problem. Conventional electronic cooling normally used the performance of the heat sinks [2]. Thus the estimation of a more
heat sink with the cooling fan to show the superiority in terms of accurate heat transfer coefficient on the fin may be an important
unit price, weight and reliability. In order to design a practical heat task for the device of the high-performance heat sinks. It can be
sink, some criterions such as a large heat transfer rate, a low found from Ref. [3] that the heat transfer coefficient encountered in
forced convection is typically much higher than those encountered
in natural convection because of the higher fluid velocities associ-
* Corresponding author. ated with forced convection. As a result, most of researchers tend to
E-mail address: [email protected] (H.-T. Chen). ignore natural convection in the forced-convection heat transfer

1359-4311/$ e see front matter Ó 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.applthermaleng.2012.08.040
 H.-T. Chen et al. / Applied Thermal Engineering 50 (2013) 352e360 353

Nomenclature Q total heat transfer rate dissipated from the fin (W)

qj heat transfer rate dissipated from the jth sub-fin region
Af lateral surface area of the fin (m2) (W)
Aj area of the jth sub-fin region (m2) Re Reynolds number, Vairt/n
[A] global conduction matrix Red hydraulic diameter Reynolds number, VairDh/n
C tip-to-shroud clearance S fin spacing (m)
cp specific heat of the air T fin temperature (K)
[K] global conduction matrix Ta air temperature (K)
[F] force matrix Tj measured fin temperature at the jth measurement
H fin height (m) location (K)
H heat transfer coefficient (W/m2 K) cal
Tj;inv calculated fin temperatures obtained from Eq. (8) (K)
h average heat transfer coefficient (W/m2 K) T0 fin base temperature (K)
hb heat transfer coefficient based on the fin base TN ambient air temperature (K)
temperature (W/m2 K) [T] global temperature matrix
hj average heat transfer coefficient in the jth sub-fin t fin thickness (m)
region (W/m2 K) ui index notation of air velocity components (m/s)
ka thermal conductivity of the air (W/m2 K) Vair frontal air velocity (m/s)
kf thermal conductivity of the fin (W/m2 K) X, Y spatial coordinates (m)
L fin length (m)
l x distance between two neighboring nodes in the X- Greek symbols
directions, L/(Nx
1) m total dynamic viscosity of the air (kg/s m)
l y distance between two neighboring nodes in the Y- mt eddy viscosity of the air (kg/s m)
directions, H/(Ny
1) n kinematic viscosity of the air (m2/s)
N number of sub-fin regions r density of the air (kg/m3)
Nt total number of grid points
Nu Nusselt number, ht=ka Subscripts
Nud hydraulic diameter Nusselt number, hb Dh =ka i i ¼ 1, 2, 3, Cartesian coordinate
Nx number of nodes in the X-direction i, j, k tensor notations
Ny number of nodes in the Y-direction (),j derivative with respect to the j direction
p pressure

analysis, although natural convection always accompanies forced et al. [8] proposed the relation between the average Nusselt
convection. However, this error involved in ignoring natural number and Reynolds number. Saini and Webb [9] applied the
convection is negligible at high velocities but may be considerable analytically based models to predict the optimum geometry of the
at low velocities associated with forced convection. plate-fin and plate-pin-fin heat sinks. The results showed that the
The experimental and numerical studies for the heat transfer optimum geometric parameters of the fin thickness and fin space
from an array of parallel rectangular finned surfaces on a horizontal respectively are 0.8 mm and 2.0 mm in the PFHS. Culham and
surface have been studied for a long time. Such problems may Muzychka [10] applied the entropy generation minimization to
exhibit the complex three-dimensional flow and thermal fields for obtain the optimization of the PFHS. Iyengar and Bar-Cohen [11]
a wider range of fin geometry. Although most existing studies presented a coefficient of performance analysis for the PFHS in
provided valuable results for the present problem, results were still forced convection using the least-energy optimization with the
inconclusive, especially for the comparative experimental results. entropy minimization methodology.
In addition, these available experimental data can remain very It was known that the heat transfer within the fins was closely
limited. Thus a more accurate predictive scheme based on the related to the airflow pattern. Velayati and Yaghoubi [12] applied
experimental data can still be needed in order to obtain the new the finite volume method in conjunction with the SIMPLE pressure-
heat-transfer characteristics in the plate-fin heat sink (PFHS). velocity coupling algorithm to solve the three-dimensional turbu-
Due to easy to manufacture, simple structure and low cost of the lent flow and heat transfer characteristics of parallel heated rect-
PFHS, they were widely used in cooling electronic equipment. angular plates mounted over an insulated base plate. It can be
Various forms of the PFHS have been manufactured and supplied to found from Ref. [12] that there exhibited very complex three-
markets in large quantity [4]. Numerical studies and some experi- dimensional flow characteristics within parallel bluff plates. This
ments for thermal performances of the plate-fin and plate-pin-fin complex flow pattern was accompanied by fluid separation and
heat sinks have been investigated by Yu et al. [4], Kim et al. [5], Li reattachment. They [12] also proposed a correlation of the average
and Chao [6] and Yang and Peng [7]. However, their thermal Nusselt number, Reynolds number and blockage ratio for an
resistance of the heat sink was obtained from the ratio of the average overall heat transfer coefficient on the bluff plates. Sparrow
temperature difference between the highest temperature at the fin et al. [13] used experimental investigation to study heat transfer
base and the ambient air temperature to the heat dissipation power and pressure drop for airflow in arrays of heat generating rectan-
applied on the fin base. El-Sayed et al. [8] varied the fin height, fin gular modules deployed along one wall of a flat rectangular duct.
width, fin spacing, number of fins and the distance from the fin tip Under the assumptions of a uniform air velocity and one-
to the shroud to study the performance of a PFHS. El-Sayed et al. [8] dimensional heat flow, Elshafei [14] applied theoretical and
concluded that the pressure drop decreased with increasing the fin experimental studies to investigate the effects of the duct velocity,
space. The mean Nusselt number increased with the fin spacing. fin density and tip-to-shroud clearance on the flow bypass, the
Under the assumptions of the one-dimensional heat conduction pressure drop across a longitudinal aluminum fin array and its
model and the constant mean heat transfer coefficient, El-Sayed thermal performance.
354 H.-T. Chen et al. / Applied Thermal Engineering 50 (2013) 352e360

Chen et al. [15] applied the finite difference method in

conjunction with experimental measured temperatures and least-
squares method to predict the natural-convection heat transfer
coefficient and fin efficiency from three fins mounted on a hori-
zontal plate for various fin spacings and fin heights. Their estimated
result of the heat transfer coefficient based on the fin base
temperature was in good agreement with those obtained from the
correlations recommended by the current heat transfer handbook
[16]. This implied that their experimental and inverse methods [15]
had good accuracy and reliability. Thus the present study further
applies the similar inverse scheme [15] and the commercial soft-
ware of FLUENT [17] in conjunction with the experimental
temperature data to determine the heat-transfer characteristics on
the fin in the plate-fin heat sink for various fin spacings. In order to
determine a more reliable heat transfer coefficient, a comparison
Fig. 2. Physical geometry of the present problem with measurement locations and
between the present inverse and numerical results and those ob-
sub-fin regions.
tained from the existing correlations will be made. In addition, the
relationship between the total number of grid points and the fin
spacing will be discussed. over a horizontal plate. The unknown heat transfer coefficient is
assumed to be the non-uniform distribution in the present study.
2. Mathematical formulation of inverse scheme Thus the entire fin is divided into several sub-fin regions before
performing the inverse scheme. The heat transfer coefficient in
A two-dimensional inverse heat conduction problem is first each sub-fin region is assumed to be an unknown constant. The
introduced to estimate the unknown heat transfer coefficient and physical geometry with measurement locations and sub-fin
fin efficiency in the plate-fin heat sink for various air velocities. For regions is shown in Fig. 2. Under the assumption of the thin fin,
the convenience of performing the inverse scheme, three vertical the temperature gradient in the Z-direction (the fin thickness) can
fins attached over a horizontal plate can be considered as a plate- be negligible and the fin temperature varies only in the X- and Y-
fin heat sink in the present study. The experimental fin tempera- directions. In addition, the edge surface area of the fin is small
ture data, ambient air temperature and air velocity are measured enough when comparing with the total fin surface area. This
from the experimental apparatus conducted in a small wind implies that the actual heat transfer rate dissipated through the fin
tunnel, as shown in Fig. 1. The inverse scheme of the finite differ- tip is rather smaller than the total heat transfer rate drawn from
ence method in conjunction with the experimental temperature the fin base. Thus the boundary conditions at the edge surface of
data and least-squares method is applied to predict the heat the fin may be assumed to be insulated [18,19]. Under the
transfer coefficient and fin efficiency for three vertical fins attached assumptions of the steady-state and constant thermal properties,

Fig. 1. Experimental apparatus configuration of the present study conducted in a small wind tunnel.
 H.-T. Chen et al. / Applied Thermal Engineering 50 (2013) 352e360 355

the two-dimensional heat conduction equation for the thin fin can coefficient based on the fin base temperature and fin efficiency can
be expressed as be determined.
The heat transfer rate dissipated from the jth sub-fin region qj is
v2 T v2 T 2hðX; YÞ Z
þ ¼ ðT
TN Þ for 0 < X < L; 0 < Y < H (1)
vX 2 vY 2 kf t qj ¼ 2hj ðT
TN ÞdA for j ¼ 1; 2; .; N (9)
Aj
Its corresponding boundary conditions are

vT The average heat transfer coefficient on the fin h may be defined as

¼ 0 at X ¼ 0 and X ¼ L (2) Z
vX
ðT
TN Þhðx; yÞdA
T ¼ T0 at Y ¼ 0 (3) Af
h ¼ Z (10)
and ðT
TN ÞdA
Af
vT
¼ 0 at Y ¼ H (4) However, it might be difficult to measure the temperature
vY
distribution on the middle fin of the present problem using the
where X and Y are the spatial coordinates. L, H and t denote the infrared thermography. Excessive thermocouples in the fin may
length, height and thickness of the rectangular fins, respectively. significantly affect the flow and heat transfer within the fins.
h(X, Y) is the unknown heat transfer coefficient on the middle fin. kf Moreover, Hu and Jacobi [20] also pointed out that the fin
is the thermal conductivity of the fin. T0 and TN, respectively, temperature measurements were inherently difficult in order not
denote the fin base temperature and ambient air temperature. The to interrupt the heat transfer behavior. This implies that Eq. (10)
fin temperature is higher than the ambient air temperature in the may not be appropriate for a practical problem. For these reasons,
present study. the thermocouple may be applied to measure the fin temperatures
The unknown heat transfer coefficient in each sub-fin region is at selected measurement locations. Under the circumstance, the
assumed to a constant. Thus the finite difference form of Eq. (1) in definition of the average heat transfer coefficient h for the inverse
the kth sub-fin region can be expressed as scheme and commercial software may be introduced as [17,18]
Z X
N
Tiþ1;j
2Ti;j þ Ti
1;j Ti;jþ1
2Ti;j þ Ti;j
1 2hk h ¼ hdA=Af z hj Aj =Af for the inverse scheme (11)
þ ¼ T
l 2
x l 2
y
kf t i;j (5) Af
j¼1

for i ¼ 1; 2; .; Nx ; j ¼ 1; 2; .; Ny and k ¼ 1; 2; .; N
and
where Nx and Ny are, respectively, the number of nodes in the X- X
Nt
and Y-directions. l x and l y are defined as l x ¼ L=ðNx
1Þ and h ¼ hi =Nt for the commercial software (12)
l y ¼ H=ðNy
1Þ. i¼1
The finite difference forms of the boundary conditions (2)e(4)
can be written as where Af is the lateral surface area of the fin. Nt denotes the total
number of grid points.
T2;j ¼ T0;j and TNx
1;j ¼ TNx þ1;j for j ¼ 1; 2; .; Ny (6) The actual total heat transfer rate dissipated from the fin to the
ambient Q can be expressed as
and
X
N
Q ¼ qj (13)
Ti;1 ¼ T0 and Ti;Ny
1 ¼ Ti;Ny þ1 for i ¼ 1; 2; .; Nx (7) j¼1

The difference equations for the nodes in the interface between The heat transfer coefficient based on the fin base temperature hb
two neighboring sub-fin regions and the intersection of four can be defined as
neighboring sub-fin regions are similar to those shown in Ref. [15].
In order to avoid repetition, they are not shown in this manuscript. Q
hb ¼ (14)
Rearrangement of the above difference equations can produce 2Af ðT0
TN Þ
the matrix equation as The fin efficiency hf can be defined as the ratio of the actual total
heat transfer rate from the fin to the dissipated heat from the fin
½K½T ¼ ½F (8)
maintained at the fin base temperature To and can be expressed as
where [K] is the global conduction matrix. [T] is the matrix repre- Q hb
senting the nodal temperatures. [F] is the force matrix. The fin hf ¼ ¼ (15)
2Af ðT0
TN Þh h
temperatures at selected measurement locations can be obtained
from Eq. (8) using the Gauss elimination algorithm. The least-squares minimization technique can be applied to
In order to estimate the unknown heat transfer coefficient hj in minimize the sum of the squares of the deviations between the
the jth sub-fin region, additional information of the measured calculated and measured fin temperatures at selected measure-
temperatures at N measurement locations in the fin is required. The ment locations. The error in the estimates Eðh1 ; h2 ; .; hN Þ is mini-
more the number of the analysis sub-fin region are, the more mized and is defined as
accurate the estimate of the unknown heat transfer coefficient may
be. However, more computational time may be required. Once   N h
X i2
cal
the unknown heat transfer coefficient in each sub-fin region is E h1 ; h2 ; .; hN ¼ Tj;inv
Tjmea (16)
obtained, the average heat transfer coefficient, heat transfer j¼1
356 H.-T. Chen et al. / Applied Thermal Engineering 50 (2013) 352e360

where Tjmea and Tj;invcal , respectively, denote the experimental and

measured and calculated fin temperatures at the jth measurement  

cal is determined from Eq. (8).

location. Tj;inv mt 3 cm h3 ð1
h=h0 Þ 3 2 3
2
ruj 3 ;j ¼ mþ 3 ;j þ c1 G

c2 r
The estimated values of hj for j ¼ 1, 2,., N can be determined s3 ;j k 1 þ bh 3 k k
provided that the value of Eðh1 ; h2 ; .; hN Þ is minimum. The details (23)
of estimating the unknown value hj can be found in Ref. [15]. In
order to avoid repetition, they are not shown in this manuscript. where mt and G denote the eddy viscosity and shear generation,
The computational procedures of the present study are repeated respectively. Their definitions are mt ¼ cmrk2/3 and G ¼ 2mtSijSij. cm, c1
 
 cal =T mea  for j ¼ 1, 2,., N are all less than and c2 are coefficients in the turbulence model. The terms c1G(3 /k)
until the values of 1
Tj;inv 
j and c2r(3 2/k) represented, respectively, the shear generation and
10
5. Once the hj values for j ¼ 1, 2,., N are determined, the values viscous dissipation of 3 . The parameter h can be written as:
of h, hb and hf can be obtained from Eqs. (11), (14) and (15). qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
h¼ c
1
m ðG=r3 Þ (24)
3. Numerical method with RNG ke3 model
A half section of a fin is selected for the computation of the
commercial software. Under assumptions of the steady-state and
Understanding the details of the local heat transfer and flow
constant thermal properties, the three-dimensional heat conduc-
distribution between two fins can be very important in designing
tion equation for the fin is assumed as
the heat sink. At the same time, in order to validate the accuracy of
the heat transfer coefficient obtained from the inverse scheme
v2 T v2 T v2 T
further, the computational fluid dynamics (CFD) commercial soft- þ þ ¼ 0 (25)
vX 2 vY 2 vZ 2
ware of FLUENT [17] may be introduced to solve the present
problem. The ambient air with constant properties can be assumed In addition to Eqs. (2) and (3), the boundary conditions at Z ¼ 0
to be incompressible. The flow between the fins can be assumed to and Z ¼ t/2 can be given as
be three-dimensional, symmetric, turbulent, steady and no viscous
dissipation. The continuity, momentum and energy equations in vT
¼ 0 at Z ¼ 0 (26)
the ambient air region may be expressed in tensor form as [12]: vZ

ui;i ¼ 0 (17) and

 
   
r uj ui;j ¼
p;i þ m ui;j þ uj;i ;j
r u0i u0i (18) vT
;j
k ¼ hðT
TN Þ at Z ¼ t=2 (27)
vZ
and 3.1. Boundary conditions
 
 
r uj Tai;j ¼ ka Ta;j þ rcp u0j Ta0 (19) At the inlet boundary, the ambient air is assumed to have the
;j uniform velocity Vair and constant temperature TN. Other velocity
components in the Y- and Z-directions are set to be zero. The gage
where ui, p and Ta are air velocity components, pressure and air pressure at this boundary is assumed to be zero. In other words,
temperature. r, m, cp and ka, respectively, are the density, total uin ¼ UN, vin ¼ 0 and win ¼ 0, Tin ¼ TN and kin ¼ 1:5  10
6 UN 2 are
dynamic viscosity, specific heat and thermal conductivity of the air. given in the present study. The inlet dissipation rate can be given as
They are all assumed to be constant. 3 in ¼ k
1:5 =ð0:446tÞ. At the solid surfaces, no-slip conditions are
in
The Reynolds stress tensor and turbulent heat fluxes in specified. The fin temperatures at various measurement locations,
conjunction with the Boussinesq approximation are given as Vair, TN and T0 are obtained from the present experiment. The
downstream outlet boundary is assumed to be located at seven the
2 fin length away from the test fins. Thus, stream-wise gradients for
ru0i u0j ¼ 2mi Sij
rkdij (20)
3 all dependent variables can be set zero.
The matching condition of the temperature and heat flux at the
and
finefluid interface can be written as

rcp u0i Ta0 ¼ ka Ta;j (21) vT vTa

T ¼ Ta and k ¼ ka (28)
vz vz
where Sij denotes the mean strain rate tensor and is defined as
Sij ¼ (ui,j þ uj,i)/2. 4. Experimental apparatus
The turbulent field of the k e 3 model was characterized in terms
of turbulent kinetic energy (k) and viscous dissipation rate of This experiment is conducted in a small wind tunnel, as shown
turbulent kinetic energy (3 ). Yakhot and Orszag [21] proposed in Fig. 1. This wind tunnel with 2.26 m in length, 0.22 m in width
a variant of the ke3 model such that its performance characteristics and 0.22 m in height is made of acrylic-plastic sheets. The sche-
were improved relative to the standard model. The RNG ke3 model matic diagram of the present experimental apparatus is shown in
was more responsive to the effects of rapid strain and streamlines Fig. 3. The ambient airflow is driven by the 115V-AC rotary fan with
curvature, flow separation, reattachment and recirculation than the a 200 W dimmer switch in order to provide the inlet air velocity.
standard ke3 model. The forms of the k and 3 equations in the RNG The airflow is straightened by the flow straightener installed in
ke3 model without buoyancy effect can be expressed as: the air inlet of the small wind tunnel during the present experi-
 
ment. The flow straightener is constructed by many pipettes
mt
ruj k;j ¼ mþ k þ G
r3 (22) (diameter ¼ 0.006 m, length ¼ 0.22 m). An anemometer installed at
sk ;j ;j 0.2 m in front of the airflow entering the test fins is used to measure
 H.-T. Chen et al. / Applied Thermal Engineering 50 (2013) 352e360 357

the frontal air velocity. The horizontal plate with 0.1 m in length,
0.1 m in width and 0.006 m and the test fins with 0.1 m in length,
0.1 m in width and 0.001 m in thickness are made of AISI 304
stainless material. The ambient air temperature and fin tempera-
ture are measured using T-type thermocouples. The limit of error of
the T-type thermocouple is 0.4% for 0 C  T  350  C. Fig. 4 shows
the schematic diagram of three parallel rectangular fins mounted
on the top surface of a horizontal plate with the frontal air velocity.
In order to heat three parallel fins, a square heater with 0.08 m in
length is fixed on the bottom of this plate using the adhesive tapes
(Nitto Denko Co., Ltd). The test fins and horizontal plate enclosed
the insulated material are placed in a small wind tunnel and then
are heated about 7600 s using the 40 W heater. The diameter of the
thermocouple is about 0.13 mm. In order to reduce the effect of the
thermal contact resistance between the fins and the horizontal
plate on the present results, their gap is filled with the cyanoac-
rylate (Satlon, D-3). Four thermocouples are also placed into their Fig. 4. Schematic diagram of three parallel rectangular fins vertically mounted on the
gap and are fixed at (L/5, 0), (2L/5, 0), (3L/5, 0) and (4L/5, 0). The top surface of a horizontal tested plate.
average of these four measured temperatures is taken as the fin
base temperature T0. Thus the thermal contact resistance between
respectively. It can be found from Table 1 that the Tjmea and Tjnum
the fins and the horizontal plate may be neglected in the present
values at X ¼ L/4 are lower than those at X ¼ 3L/4 for various S
study. In order to measure the ambient air temperature TN, a ther-
values. The fin has a higher temperature drop in the neighborhood
mocouple is positioned at 0.3 m in front of the airflow entering the
of the fin base from the fin base (Y ¼ 0) to Y ¼ 3H/8. On the other
test fins, as shown in Fig. 3.
hand, the Tjmea and Tjnum values in the downstream sub-fin regions
are higher than those in the upstream sub-fin regions. This
5. Results and discussion
phenomenon may result from a low-velocity region that may occur

All the physical properties are evaluated at the film temperature

or the average of the fin base temperature and frontal air temper- Table 1
ature. All the computations are performed with the fin height Temperature measurements and the fin efficiency for various S values.
H ¼ 0.06 m, front air velocity Vair ¼ 1 m/s and kf ¼ 14.9 W/(m K). S ¼ 0.005 m S ¼ 0.010 m S ¼ 0.015 m S ¼ 0.02 m
Nx ¼ 21 and Ny ¼ 17 are used to perform the present inverse
To ¼ 323.48 K To ¼ 321.66 K To ¼ 324.54 K To ¼ 326.02 K
method. The commercial software of FLUENT [17] is applied to
TN ¼ 296.72 K TN ¼ 297.76 K TN ¼ 299.19 K TN ¼ 299.28 K
determine the fluid flow and heat transfer of the present problem.
An unstructured mesh system containing non-uniformly distrib- Tjmea ðKÞ T1mea ¼ 312:54 T1mea ¼ 311:58 T1mea ¼ 311:03 T1mea ¼ 311:74
T2mea ¼ 316:07 T2mea ¼ 314:10 T2mea ¼ 314:13 T2mea ¼ 314:81
uted grid points is used to determine all the numerical results. The T3mea ¼ 302:59 T3mea ¼ 303:20 T3mea ¼ 303:09 T3mea ¼ 303:05
unknown heat transfer coefficients hj for j ¼ 1, 2,., N used to begin T4mea ¼ 306:55 T4mea ¼ 305:74 T4mea ¼ 305:78 T4mea ¼ 305:96
the iterations are taken as unity. The rectangular fin is divided into T5mea ¼ 298:10 T5mea ¼ 299:37 T5mea ¼ 300:92 T5mea ¼ 300:73
eight regions, i.e., N ¼ 8. These eight thermocouples are fixed at T6mea ¼ 301:27 T6mea ¼ 300:53 T6mea ¼ 302:67 T6mea ¼ 302:48
T7mea ¼ 297:04 T7mea ¼ 298:27 T7mea ¼ 299:88 T7mea ¼ 300:37
(L/4, H/8), (3L/4, H/8), (L/4, 3H/8), (3L/4, 3H/8), (L/4, 5H/8), (3L/4, 5H/
T8mea ¼ 299:13 T8mea ¼ 298:95 T8mea ¼ 301:41 T8mea ¼ 301:73
8), (L/4, 7H/8) and (3L/4, 7H/8), as shown in Fig. 2. Tjnum denotes the
calculated fin temperature at the jth measurement location ob- Tjnum ðKÞ T1num ¼ 314:13 T1num ¼ 312:45 T1num ¼ 313:19 T1num ¼ 312:95
T2num ¼ 315:21 T2num ¼ 313:22 T2num ¼ 314:76 T2num ¼ 314:88
tained from the commercial software of FLUENT [17]. The fin base
T3num ¼ 303:44 T3num ¼ 303:34 T3num ¼ 302:56 T3num ¼ 303:92
temperature, ambient air temperature, experimental measured fin T4num ¼ 304:91 T4num ¼ 303:38 T4num ¼ 305:03 T4num ¼ 304:72
temperature Tjmea and Tjnum are shown in Table 1. The approximate T5num ¼ 299:32 T5num ¼ 300:05 T5num ¼ 299:83 T5num ¼ 300:88
grid points may be chosen in order to reduce the difference of the T6num ¼ 300:47 T6num ¼ 300:39 T6num ¼ 301:49 T6num ¼ 301:91
T7num ¼ 297:98 T7num ¼ 298:87 T7num ¼ 298:88 T7num ¼ 300:07
average heat transfer coefficient between the inverse results and
T8num ¼ 298:91 T8num ¼ 299:27 T8num ¼ 299:80 T8num ¼ 300:72
the numerical results. The selection of grid points may be depen-
dent on the fin spacing. The grid points are 156,852 for S ¼ 5 mm, hf 0.16 0.21 0.32 0.33
399,759 for S ¼ 10 mm and 557,362 for S ¼ 15 mm and 20 mm,

Fig. 3. Schematic diagram of experimental apparatus.

358 H.-T. Chen et al. / Applied Thermal Engineering 50 (2013) 352e360

Table 2 Under the assumption of the one-dimensional heat flow,

Comparison of h and hb for various S values. Elshafei [14] proposed a correlation of the hydraulic diameter
S 0.005 m 0.01 m 0.015 m 0.02 m Nusselt number Nud and hydraulic diameter Reynolds number Red
hj Inverse h1 ¼ 39:90 h1 ¼ 45:49 h1 ¼ 86:20 h1 ¼ 39:99 for L ¼ 225 mm, H ¼ 51 mm, t ¼ 2 mm, S ¼ 15 mm, 0  C/H  0.89,
h2 ¼ 12:72 h2 ¼ 21:36 h2 ¼ 43:00 h2 ¼ 12:70 0  C/S  3 and 3000  Red  38,000 as
h3 ¼ 24:40 h3 ¼ 20:04 h3 ¼ 38:01 h3 ¼ 24:22
h4 ¼ 11:10 h4 ¼ 5:76 h3 h3 7:552Re0:182
d
h5 h5 h5 h5 Nud ¼ (30)
h6 h6 h6 h6 ½ð1 þ C=HÞð1 þ C=SÞ0:1096
h7 h7 h7 h7
h8 h8 h8 h8 where Nud and Red are defined as Nud ¼ hb Dh =ka and Red ¼ VairDh/
h Inverse 41.10 38.16 36.50 35.34 n. Dh is the hydraulic diameter of the fin array. Its definition can
FLUENT 38.91 37.22 34.74 35.73 refer to Ref. [14]. C denotes the tip-to-shroud clearance.
Eq. (24) 34.02 34.04 33.93 33.89 Table 2 shows the comparison of the average heat transfer
hb Inverse 6.77 8.12 11.60 11.72 coefficient in the whole fin h and heat transfer coefficient based on
FLUENT 6.66 7.17 13.52 14.53 the fin base temperature hb for various fin spacings. It can be found
Eq. (25) 6.68 7.18 7.51 7.74 from Table 2 that the estimated results of the average heat transfer
coefficient in the downstream sub-fin regions obtained from the
in the downstream sub-fin regions. Thus the difference between present inverse method are lower than those in the upstream sub-
the fin temperature and the ambient air temperature in the fin regions. Thus, in order to enhance the overall heat transfer of the
upstream region is lower than that in the downstream region. This present problem, it may be worth to find a way to increase heat
implies that the average heat transfer coefficients in the down- transfer in the downstream sub-fin regions. It can also be observed
stream sub-fin regions are lower than those in the upstream sub-fin from Table 2 that the h value decreases with increasing the fin
regions. An interesting finding is that the Tjnum values are in good spacing. However, the hb value increases with the fin spacing. An
agreement with the experimental temperature data Tjmea . important finding is that the present inverse results of h are in good
Various heat transfer correlations for the present problem have agreement with those obtained from the commercial software of
been proposed [8,12e14]. Among these heat transfer correlations, FLUENT [17] and the correlation proposed by Sparrow et al. [13] for
the correlations proposed by Sparrow et al. [13] and Elshafei [14] 0.005 m  S  0.02 m. The present inverse results of hb also agree
are selected to compare with the present inverse and numerical with the present numerical results obtained from the commercial
results of the heat transfer coefficient. Sparrow and Samie [22] software of FLUENT for 0.005 m  S  0.02 m and the correlation
pointed out that the heat transfer coefficient was non-uniform proposed by Elshafei [14] for S ¼ 0.005 m and 0.01 m. These
along the fin surface in practice, so that the fin efficiency ob- differences may result from the definition of the average heat
tained from the conventional fin model was uncertain accuracy. transfer coefficient, the assumptions of the one-dimensional heat
Sparrow et al. [13] used experimental investigation to study flow, constraints of the empirical formula, some parameter values
heat transfer and pressure drop for airflow in arrays of heat in the RNG ke3 model, total number of grid points and experi-
generating rectangular modules deployed along one wall of a flat mental errors, etc. It can be found from Table 1 that the fin effi-
rectangular duct. Sparrow et al. [13] did not investigate the effect of ciency hf increases with the fin spacing. The hb and hf values may
the fin spacing on the h value. It can be found from Refs. [12,13] that gradually approach an asymptotic value with increasing the fin
they proposed a correlation of the Nusselt number and Reynolds spacing. Thus the proper choice of the S/L value may be required in
number for t ¼ 26.67 mm, t/L ¼ 1, S/L ¼ 1/4, 3800  Re  13,300 and order to increase the hb value and fin efficiency.
height of the flow passage ¼ 5t/8 as The flow structure for the airflow coming into the region
between two parallel fins may contribute to the boundary layer
Nu ¼ 0:0935ð5Re=8Þ0:72 (29) development and may be affected by the presence of the fin base
and the free-stream above the fins. Their interaction may generate
where Nu and Re are defined as Nu ¼ ht=ka and Re¼Vairt/n. n is the the complex flow within the fins. The air velocity distributions
kinematic viscosity of the air. within the fins are shown in Fig. 5 for S ¼ 0.015 m and various

Fig. 5. Air velocity distributions for S ¼ 0.015 m and various heights: (a) Y ¼ H/4, (b) Y ¼ H/2, (c) Y ¼ 3H/4.
 H.-T. Chen et al. / Applied Thermal Engineering 50 (2013) 352e360 359

Fig. 6. Temperature distribution on the fin for various fin spacings: (a) S ¼ 0.005 m; (b) S ¼ 0.01 m; (c) S ¼ 0.015 m; (d) S ¼ 0.02 m.

elevations from the fin base. It can be found that the airflow at University to give us the computational fluid dynamics commercial
Y ¼ H/4 can form a low-velocity region in the middle region of the software of FLUENT.
fin. This low-velocity region may occur in the downstream region at
Y ¼ H/2 and Y ¼ 3H/4. On the other hand, it may change with the References
elevation from the fin base. This implies that the flow field of the
present problem is strongly three-dimensional. [1] R.L. Webb, Principles of Enhanced Heat Transfer, Wiley, New York, 1994, pp.
The fin temperature distribution is shown in Fig. 6 for various fin 125e153.
[2] Ş. Yildiz, H. Yuncu, An experimental investigation on performance of annular
spacings. It can be observed that the fin temperature distribution is
fins on a horizontal cylinder in free convection heat transfer, Heat Mass
non-uniform and asymmetric with respect to X ¼ L/2 for Transfer 40 (2004) 239e251.
S ¼ 0.005 m. However, the fin temperature distribution is [3] Y.A. Çengel, Heat Transfer e A Practical Approach, second ed., McGraw-Hill,
New York, 2004, pp. 486e488.
approximately symmetric with respect to X ¼ L/2 for
[4] X.L. Yu, J.M. Feng, Q.K. Feng, Q.W. Wang, Development of a plate-pin fin heat
0.01 m  S  0.02 m. The fins have a higher temperature and sink and its performance comparisons with a plate fin heat sink, Appl. Therm.
temperature gradient close to the fin base, as shown in Table 2. The Eng. 25 (2005) 173e182.
temperature variation near the fin tip seems to be enormous. [5] S.J. Kim, D.-K. Kim, H.H. Oh, Comparison of fluid flow and thermal charac-
teristics of plate-fin and pin-fin heat sinks subject to a parallel flow, Heat
Transfer Eng. 29 (2008) 169e177.
6. Conclusions [6] H.Y. Li, S.M. Chao, Measurement of performance of plate-fin heat sinks with
cross flow cooling, Int. J. Heat Mass. Transfer 52 (2009) 2949e2955.
[7] Y.T. Yang, H.S. Peng, Investigation of planted pin fins for heat transfer
The present study proposes the inverse method and the enhancement in plate fin heat sink, Microelectron. Reliab. 49 (2009)
commercial software of FLUENT in conjunction with the experi- 163e169.
mental temperature data to determine the average heat transfer [8] S.A. El-Sayed, S.M. Mohamed, A.M. Abdel-latif, A.E. Abouda, Investigation of
turbulent heat transfer and fluid flow in longitudinal rectangular-fin arrays of
coefficient h, heat transfer coefficient based on the fin base different geometries and shrouded fin array, Exp. Therm. Fluid Sci. 26 (2002)
temperature hb and fin efficiency for various fin spacings. An 879e900.
interesting finding is that the calculated fin temperatures obtained [9] M. Saini, R.L. Webb, Heat rejection limits of air cooled plane fin heat sinks for
computer cooling, IEEE Trans. Compon. Packag. Technol. 26 (2003) 71e79.
from the commercial software are in good agreement with the [10] J.R. Culham, Y.S. Muzychka, Optimization of plate fin heat sinks using entropy
experimental temperature data at various measurement locations. generation minimization, IEEE Trans. Compon. Packag. Technol. 24 (2001)
The inverse results of h and hb also agree with those obtained from 159e165.
[11] M. Iyengar, A. Bar-Cohen, Least-energy optimization of forced convection
the commercial software or the correlation. This implies that the plate-fin heat sinks, IEEE Trans. Compon. Packag. Technol. 26 (1) (2003)
present results have good accuracy. It is worth mentioning is that 62e70.
the selection of approximate grid points may not be negligible in [12] E. Velayati, M. Yaghoubi, Numerical study of convective heat transfer from an
array of parallel bluff plates, Int. J. Heat Fluid Flow 26 (2005) 80e91.
order to determine a more accurate numerical result. The total
[13] E.M. Sparrow, J.E. Niethammer, A. Chaboki, Heat transfer and pressure drop
number of grid points may increase with the fin spacing. Thus the characteristics of arrays of rectangular modules encountered in electronic
commercial software in conjunction with the inverse method and equipment, Int. J. Heat Mass. Transfer 25 (1982) 961e973.
experimental temperature data may help to the future develop- [14] E.A.M. Elshafei, Effect of flow bypass on the performance of a shrouded
longitudinal fin array, Appl. Thermal Eng. 27 (2009) 2233e2242.
ment of the heat sink. [15] H.T. Chen, L.S. Liu, S.K. Lee, Estimation of heat-transfer characteristics from
fins mounted on a horizontal plate in natural convection, CMES: Comput.
Model. Eng. Sci. 65 (2010) 155e178.
Acknowledgements
[16] G.D. Raithby, K.G.T. Hollands, Natural convection, in: W.M. Rohsenow,
J.P. Hartnett, E.N. Ganic (Eds.), Handbook of Heat Transfer Fundamentals,
The authors gratefully acknowledge the financial support second ed., McGraw-Hill, New York, 1985.
provided by the National Science Council of the Republic of China [17] FLUENT, Dynamics Software, FLUENT, Lehanon, NH, 2010.
[18] H. Ay, J.Y. Jang, J.N. Yeh, Local heat transfer measurements of plate finned-tube
under Grant No. NSC 98-2221-E-006-177-MY3. We would also like heat exchangers by infrared thermography, Int. J. Heat Mass. Transfer 45
to thank Professor Chin-Hsiang Cheng at National Cheng Kung (2002) 4069e4078.
360 H.-T. Chen et al. / Applied Thermal Engineering 50 (2013) 352e360

[19] C.H. Huang, I.C. Yuan, H. Ay, A three-dimensional inverse problem in imaging [21] V. Yakhot, S.A. Orszag, Renormalization group analysis of turbulence, J. Sci.
the local heat transfer coefficients for plate finned-tube heat exchangers, Int. J. Comput. 1 (1986) 3e51.
Heat Mass. Transfer 46 (2003) 3629e3638. [22] E.M. Sparrow, F. Samie, Heat transfer and pressure drop results for one- and
[20] X. Hu, A.M. Jacobi, Local heat transfer behavior and its impact on a single-row, two-row arrays of finned tubes, Int. J. Heat Mass. Transfer 28 (1985) 2247e
annularly finned tube heat exchanger, ASME J. Heat Transfer 115 (1993) 66e74. 2259.