You are on page 1of 9

Applied Energy 116 (2014) 260–268

Contents lists available at ScienceDirect

Applied Energy
journal homepage: www.elsevier.com/locate/apenergy

Optimum design of a radial heat sink with a fin-height profile for


high-power LED lighting applications
Daeseok Jang, Se-Jin Yook, Kwan-Soo Lee ⇑
School of Mechanical Engineering, Hanyang University, 222 Wangsimni-ro, Seongdong-gu, Seoul 133-791, Republic of Korea

h i g h l i g h t s

 A radial heat sink was designed for high-power LED lighting applications.
 Fin-height profiles reflecting the chimney-flow characteristics of a radial heat sink were proposed.
 Multi-disciplinary optimization was carried out to simultaneously minimize the thermal resistance and mass.
 The cooling performance of the optimized design showed improvement without additional mass increment.

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

Article history: Light-emitting diode (LED) lighting offers greater energy efficiency than conventional lighting. However,
Received 26 April 2013 if the heat from the LEDs is not properly dissipated, the lifespan and luminous efficiency are diminished.
Received in revised form 11 November 2013 In the present study, a heat sink of LED lighting was optimized with respect to its fin-height profile to
Accepted 23 November 2013
obtain reliable cooling performance for high-power LED lighting applications. Natural convection and
Available online 20 December 2013
radiation heat transfer were taken into consideration and an experiment was conducted to validate
the numerical model. Fin-height profiles reflecting a three-dimensional chimney-flow pattern were pro-
Keywords:
posed. The outermost fin height, the difference between fin heights, and the number of fin arrays were
LED lighting
Electronic cooling
adopted as design variables via sensitivity analysis, and the heat sink configuration was optimized in
Natural convection three dimensions. Optimization was conducted to simultaneously minimize the thermal resistance and
Heat sink mass. The result was compared with the Pareto fronts of a plate-fin heat sink examined in a previous
Optimization study. The cooling performance of the optimized design showed an improvement of more than 45% while
Fin-height profile preserving a mass similar to that of the plate-fin heat sink.
Ó 2013 Elsevier Ltd. All rights reserved.

1. Introduction its cooling performance via shape alteration. Optimization of the


radial heat sinks used in LED lighting applications is necessary to
Light-emitting diode (LED) lighting offers the advantages of a obtain reliable cooling performance in high-power LED products.
longer lifespan and greater energy efficiency than conventional Numerous researchers have analyzed natural convection heat
lighting. LED lighting uses 75% less energy than incandescent or sinks with diverse fin configurations [1–8]. Elshafei [7] examined
fluorescent lighting. Due to its greater energy efficiency, the mar- the heat transfer characteristics of hollow/perforated round pin–
ket share of LED lighting is growing rapidly, and the market do- fin heat sink, and compared the results with those obtained for
main now includes high-power LED products. However, if high solid round pin–fin heat sink. Sertkaya et al. [8] investigated the
power is applied to LEDs to produce more light output, the amount orientation effect of a pin–fin heat sink under natural convection,
of heat generated by the LEDs is greatly increased, which reduces and reported that the cooling performance with an upward-facing
both the lifespan and the luminous efficiency. Therefore, a technol- orientation was outstanding. However, most of these researches
ogy for properly dissipating the heat from inside the LED package considered rectangular heat sink, which is ineffective for cooling
to the surroundings is as important as the electronic and optical in round LED lighting applications. Accordingly, diverse types of ra-
characteristics. The LED lighting system shown Fig. 1 is composed dial heat sinks have recently been proposed and optimized [9–12].
of an LED chip, epoxy, a slug, a printed circuit board, and a heat Multidisciplinary optimization is the primary method for improv-
sink. This study is confined to the heat sink, which can improve ing heat sink performance because both mass and cooling perfor-
mance are considered in the performance evaluation. Jang et al.
[12] compared various pin–fin radial heat sinks to achieve a lighter
⇑ Corresponding author. Tel.: +82 2 2220 0426; fax: +82 2 2295 9021. heat sink with a cooling performance similar to that of a plate-fin
E-mail address: ksleehy@hanyang.ac.kr (K.-S. Lee).

0306-2619/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.apenergy.2013.11.063
D. Jang et al. / Applied Energy 116 (2014) 260–268 261

Nomenclature

A heat transfer area (mm2) r standard deviation of elementary effects distribution


F body force vector per unit volume q density (kg/m3)
H height (mm) h angle, as in Fig. 3(b) (°)
I radiation intensity e emissivity
k thermal conductivity (W/m K) x weight factor
L length (mm) X hemispherical solid angle (°)
~
L periodic length vector
M mass of heat sink (kg) Subscripts
m _ mass flow rate (kg/s) A array
N number acr acrylic
n normal vector avg average
P pressure (Pa) D difference between fin heights
q_ heat flux (W/m2) f fin
RTH thermal resistance (°C/W) i inner
r radius (mm) L long fin
~
s ray vector M middle fin
T temperature (K) O outermost fin height
t fin thickness (mm) o outer
v velocity vector (m/s) s space between fins in the radial direction
w wall
Greek symbols 1 ambient
l dynamic viscosity (N s/m2)

l mean of elementary effects distribution

heat sink reported in previous studies [9–11]. Using this approach, the central region reduced the pressure drop in that region, and
they were able to reduce the mass by more than 30% while pre- improved the flow penetration to the outer region. As a result, a
serving a cooling performance similar to that of the optimum fin-height profile with reduced fin volume in the central region en-
plate-fin heat sink [10]. However, these studies were based on hanced both the thermal and hydraulic performance. Shah et al.
the assumption of uniform fin height, and hence their results are [16] extended their previous work and optimized the configuration
unsuitable for radial heat sinks with chimney-flow characteristics of the heat sink by investigating the removal of fin volume at the
because non-uniform heat transfer occurs along the radial direc- end fins, varying the number of fins, and reducing the size of the
tion in a chimney flow. Therefore, it is necessary to determine fan. Yang and Peng [17] studied impinging jets on pin–fin heat
fin-height profiles that are appropriate for chimney-flow charac- sinks with non-uniform fin heights. They found that the heat sink
teristics. Several recent studies have focused on the effects of temperature could be decreased by increasing the fin height
various fin-height profiles. Fabbri [13] studied the influence of around the central region of the heat sink. However, there is a limit
fin-height profiles with polynomial forms and obtained an opti- to how much the fin height around the central region of a heat sink
mum profile for every polynomial degree by using a genetic algo- can be increased because flow penetration into that region is weak-
rithm. Kanyakam and Bureeret [14] studied the forced convection ened by the resulting increased flow resistance. In addition, if the
of splayed pin–fin heat sinks by varying the fin height and azi- fins around the central region are too tall in comparison to the
muthal angle with the heat sink base as the center. They concluded other fins, heat transfer in the outer region will be poor and the
that a splayed heat sink offers outstanding cooling performance overall thermal performance will be degraded. Bello-Ochende
compared to a straight pin–fin heat sink, and Pareto fronts relating et al. [18] developed an optimized pin–fin heat sink design for
the cooling performance and the fan pumping power were maximizing forced convection with limited mass and found that
obtained via multi-objective optimization. Shah et al. [15] numer- cooling performance was maximized when the fin diameter and
ically analyzed the performance of a heat sink with an impinge- height were non-uniform. However, these studies [13–18] were
ment cooling system. The effect of the fin-height profile on the concerned with the forced convection of heat sinks with rectangu-
cooling performance (especially near the center of the heat sink) lar bases, and the optimum fin-height profile differed with respect
was analyzed. They found that reduction of fin volume around to the cooling method or shape. Therefore, a research of radial heat

Ambient air
Heat sink

PCB Printed circuit board

Slug
Epoxy

LED chip

Fig. 1. Diagram of a typical LED lighting application.


262 D. Jang et al. / Applied Energy 116 (2014) 260–268

sink under natural convection with a non-uniform fin-height pro-


file is still needed. In previous studies of radial heat sinks [10–
12], the thermo-flow characteristics around the heat sink were
analyzed and geometric optimization was carried out. The temper- (a)
ature contours of various fin arrays on a horizontal plane were also
compared, and the fin lengths and number of fin arrays were opti-
mized to retard the growth of a fully developed thermal boundary
layer. However, optimization results based on two-dimensional
thermo-flow characteristics on a horizontal plane are not appropri-
ate for the three-dimensional flow of a radial heat sink. To reflect
chimney-flow characteristics, an optimum design based on the
overall three-dimensional flow is necessary. LD
In the present research, a radial heat sink with pin fins is opti- LO
mized with respect to the fin-height profile. Fin-height profiles
are proposed that can improve cooling performance without
increasing the mass beyond that of a comparable plate-fin heat
sink. The cooling performances of three profiles (the LM type, a
pin–fin array with the tallest fins in the inner region, and a
pin–fin array with the tallest fins in the outer region) with equal
masses are analyzed and a reference profile is determined. Final-
ly, a heat sink is designed using multi-objective optimization con-
sidering both mass and thermal resistance to provide improved
cooling performance with a mass equivalent to that in previous Lf Ls
studies. (b)
LM
t
(a) ri
θ
LL
ro
g
Fig. 3. Computational domain. (a) Isometric view of the computational domain. (b)
gravity Top view of the computational domain.

2. Mathematical modeling
(b)
2.1. Numerical model

The LM type of heat sink addressed in previous studies [9–11],


a pin–fin array with the tallest fins in the inner region (Type 1),
g and a pin–fin array with the tallest fins in the outer region (Type
2) were compared (see Fig. 2). The fins were circularly arranged
gravity at consistent intervals. The heat sink base was oriented
horizontally, whereas the fins were arranged vertically. The fin
array was duplicated around the circumference of the base plate.
Due to the computational cost involved, only a single fin
array was used for the computational domain, as shown in
Fig. 3. The numerical analysis was based on the following
(c) assumptions.

(1) The flow is laminar, steady, and three dimensional.


(2) The density of air is computed using the ideal gas law.
(3) Except for the density of air, the fluid properties are
g constant.
(4) The heat sink surface is gray and diffuse.
gravity
2.2. Governing equations and boundary conditions

Table 1 enumerates the governing equations and boundary con-


ditions. Radiation heat transfer was computed via the discrete
Fig. 2. Test heat sinks. (a) LM plate-fin type. (b) Pin–fin array with the tallest fins on transfer radiation model [19,20], which can support periodic
the inside (Type 1). (c) Pin–fin array with the tallest fins on the outside (Type 2). conditions.
D. Jang et al. / Applied Energy 116 (2014) 260–268 263

Table 1
Governing equations and boundary conditions for the computational domain.

Wall Continuity equation Momentum equations Energy equation


r  ðqv Þ ¼ 0 q Dv
¼ rP þ lr v þ F 2
qC p DT DP
Dt ¼ r  ðkrTÞ þ Dt
Dt
(for z-direction F = qg)
Periodic face ui ð~ ri þ ~
r i Þ ¼ ui ð~ LÞ, Tð~ ri þ ~
r i Þ ¼ Tð~ LÞ
Fluid domain ~
rpðxi Þ ¼ g j~LLj þ rp ðxi Þ
Outer face Pressure inlet/Pressure outlet condition T inlet ¼ T outlet; back flow ¼ T 1

Heat sink base ui ¼ 0 s
ks @T @n heat sink base ¼ q
_
Solid domain 
@T s 
Symmetric face ui ¼ 0 ¼0
@n sectional wall

Interface between fluid and solid domain Interface ui = 0

T f;wall ¼ T s; wall ;
 
f s
kf @T
@n 
þ q_ out ¼ ks @T @n wall
þ q_ in
wall
R !
_qin ¼ ~s~n>0 Iin~ ~
s  ndX;
q_ out ¼ ð1  ew Þqin þ ew rT 4w

2.3. Numerical procedure tallest fin, and the radius beyond 1.5 ro, the temperature of the heat
sink changed by less than 0.5%; these results were used to set the
The numerical analysis was carried out using the finite volume size of the computational domain. A dense grid was generated in
method with Fluent V6.3, a commercial computational fluid the region where a boundary layer developed near the heat sink.
dynamics software package. The SIMPLE algorithm was adopted Grid dependence testing took place by increasing the number of
to couple the velocity and pressure fields. To enhance the accuracy grid points from 60,000 to 700,000, with 519,375 grid points being
of the result, a second-order upwind scheme was employed to the selected as a reference from our sensitivity analysis. The change in
convection terms of the governing equations. The result was deter- heat sink temperature with additional grid points was less than
mined to have converged when the relative error of all dependent 0.5%. ICEM CFD 14.5 was used to generate the mesh, and the grid
variables between two successive iterations was less than 105. system was shown in Fig. 4.
Taking into account both the convergence of the heat sink temper-
ature and the computational time, the height of the domain chan-
ged from two to ten times the height of the tallest fin, and the
radius of the domain changed from 1.3ro to 1.6ro. When the do-
main height was increased beyond five times the height of the

(a) Heat sink

Insulator
(polystyrene)

T3

T2 Film heater
Aluminum T1
plate
Acrylic
plate

(b)
Heat sink

Insulator
Thermocouple (polystyrene)

Fig. 5. Experimental setup. (a) Illustration for the experimental setup. (b) Photo for
Fig. 4. Computational grid system. (a) Top view. (b) Side view. the experimental setup.
264 D. Jang et al. / Applied Energy 116 (2014) 260–268

3. Experiments and validation An uncertainty analysis was carried out as part of the experi-
ment. The maximum uncertainty was estimated to be 2.7%. The
The LM type of heat sink [10], a pin–fin array with the tallest numerical results agreed well with the results of experiment with
fins in the inner region (Type 1), and a pin–fin array with the tallest an error of less than 6.8%. Accordingly, we validated that the
fins in the outer region (Type 2) were evaluated to validate the numerical model explained in Section 2 could correctly simulate
numerical model. The heat sinks were made of black anodized alu- natural convection and radiation heat transfer around a heat sink
minum (Al6061, e = 0.8). The parameters for the LM type were with a specific fin-height profile.
NA = 20, LL = 50 mm, LM = 20 mm, ro = 75 mm, ri = 10 mm,
t = 2 mm, and H = 21.3 mm. The parameters for Type 1 were 4. Results and discussion
NA = 20, LL = 45 mm, LM = 25 mm, Lf = 5 mm, LO = 24.3 mm, LD = 8 -
mm, Ls = 5 mm, ro = 75 mm, ri = 10 mm, and t = 2 mm. For Type 2, Pin–fin heat sinks with diverse fin-height profiles were com-
LO = 50.3 mm; the other parameters were the same as those of pared to the plate-fin heat sink suggested by Yu et al. [10] in terms
Type 1. As shown in Fig. 5, the experimental setup comprised a of cooling performance and mass. The elementary effects method
heat sink, a film heater (Kapton-coated stainless steel, 25 lm), an was utilized for the sensitivity analysis from which the design vari-
insulator, a data acquisition device (NI SCXI-1303, 1100, 1600), ables were determined. Finally, multi-objective optimization was
type-T thermocouples (gauge 36), a power supply, a personal com- conducted considering both mass and cooling performance, and
puter, and a wattmeter. Thin aluminum plates, 1 mm thick, were application of the optimized design to high-power LED products
placed beneath and on top the film heater to apply a uniform heat was examined.
flux. Thermal grease was employed to decrease the thermal con-
tact resistance between the heat sink and the film heater. An ac-
rylic plate, 5 mm thick, was inserted beneath the heater and used 4.1. Thermo-flow characteristics around a plate-fin radial heat sink
to evaluate the heat loss from the bottom of the heater as follows:
Fig. 7(a) illustrates the temperature contours around a radial
Q_ heat sink ¼ Q_ total  Q_ heat loss ð1Þ heat sink with the LM-type design (NA = 20, LL = 50 mm, LM = 30 -
mm, ro = 75 mm, ri = 10 mm, t = 2 mm, H = 21.3 mm, and e = 0.7,
ðT 1  T 2 Þ where NA is the number of fin arrays, LL is the long fin length,
Q_ heat loss ¼ kacr Aacr ð2Þ
tacr and LM is the middle fin length). The reference heat flux was
700 W/m2, which is equivalent to the power consumption of a typ-
To minimize the heat loss through the sides, the section of film
ical LED down light (20 W). To investigate the heat transfer charac-
heater was enclosed by an insulator. The heat sink temperature
teristics, the overall flow characteristics were examined from two
was measured with eight thermocouples, and the ambient air tem-
perspectives. The right-hand side of Fig. 7(a) shows the tempera-
perature was measured with two thermocouples. Constant current
ture contours for the first perspective, lying on a horizontal plane
and voltage were applied to the film heater during the experiment.
10 mm above the heat sink base. The spaces between the plate fins
When the temperature change was less than 0.1 °C over 30 min,
of the heat sink are comparatively wide in the outer region, and be-
the heat sink temperature was assumed to have reached a steady
come narrower in the inner region. The more the domain of inter-
state, and the thermal performance was calculated.
est penetrates into the inner region, the narrower these spaces
Fig. 6 compares the experimental and numerical results. The
become. In the inner region, the thermal boundary layers between
thermal resistance of the heat sink was adopted as the perfor-
the fins overlapped and become fully developed, leading to a de-
mance index, and defined by
  crease of the heat transfer coefficient.
T avg;heat sink  T 1 The left-hand side of Fig. 7(a) shows the temperature contours
RTH ¼     ð3Þ
p r2o  r2i  q_ heat sink for the second perspective, lying on a vertical plane (h = 9°). The
incoming cool air is warmed by the heat sink and rises. The overall
flow typically has a chimney-like pattern. Most of the heat transfer
occurs in the outer region and little heat transfer occurs in the cen-
tral region. Most of the mass flow does not reach the inner fins be-
3 cause the inflow rises in accordance with the chimney pattern.
Experiment Numerical Model Therefore, the mass flow rate for cooling decreases in the inner re-
LM type gion, resulting in non-uniform heat transfer.
Type 1 In previous studies that only considered the thermo-flow char-
Type 2 acteristics on a horizontal plane, the fin lengths and the number of
fin arrays that influence the thermal boundary layer development
2 on the horizontal plane were adopted as design variables for opti-
RTH (°C/W)

mum design. However, design variables pertaining exclusively to


two-dimensional thermo-flow characteristics are inappropriate
for three-dimensional radial heat sink flow analysis because non-
uniform heat transfer with respect to the radial direction must
be reflected in the design procedure. Taking into account the chim-
ney pattern of flow and the fully developed thermal boundary layer
1 in the inner region, the uniform fin height assumed in previous
studies is unsuitable for analyzing a non-uniform heat transfer
distribution.

0 400 800 1200


. 4.2. Non-uniform fin-height profiles
q (W/m2)

Fig. 6. Comparison of the computational and experimental results (ro = 0.075 m, To investigate the three-dimensional flow characteristics of a
e = 0.8). radial heat sink, the effects of diverse fin-height profiles on the
D. Jang et al. / Applied Energy 116 (2014) 260–268 265

(a) 53.64 (oC)

51.27

48.91

46.55

44.18

41.82

39.46

37.09

34.73

32.36
θ = 9° H = 10 mm
30.00

(b) 53.64 (oC)

51.27

48.91

46.55

44.18

41.82

39.46

37.09

34.73

32.36
θ = 9° H = 10 mm
30.00

53.64 (oC)
(c)
51.27

48.91

46.55

44.18

41.82

39.46

37.09

34.73

32.36
θ = 9° H = 10 mm
30.00

Fig. 7. Temperature contours at h = 9° and H = 10 mm (ro = 0.075 m, q_ ¼ 700 W=m2 , T1 = 30 °C, e = 0.7). (a) LM plate-fin model. (b) Pin–fin array with the tallest fins on the
inside (Type 1). (c) Pin–fin array with the tallest fins on the outside (Type 2).

heat sink mass and temperature were examined. The LM type of (Type 2) were compared. In general, cooling performance improves
heat sink, a pin–fin array with the tallest fins in the inner region with increasing heat sink mass. Accordingly, pin–fin heat sinks
(Type 1), and a pin–fin array with the tallest fins in the outer region with a mass equal to that of the LM type (M = 0.288 kg) were used
for the comparison. The parameters for the Type 1 pin–fin array
were NA = 20, LL = 50 mm, LM = 30 mm, Lf = 10 mm, LO = 22 mm,
Table 2
Comparison of various fin height profiles (ro = 0 075 m, q_ ¼ 700 W=m2 , T1 = 30 °C,
LD = 15 mm, Ls = 10 mm, ro = 75 mm, ri = 10 mm, t = 2 mm,
e = 0.7). Tair = 30 °C, e = 0.7, and q_ ¼ 700 W=m2 , where LO is the outermost
fin height and LD is the fin-height difference between the fins.
Model A (mm2) LO (mm) LD (mm) _ air (105 kg/s)
m Tavg (°C)
For the Type 2 pin–fin array, LO = 46 mm, and the other parameters
LM plate-fin 4499 21.3 0 3.04 53.12 were the same as those of Type 1. Fig. 7 shows the temperature
Type 1 5000 22 15 4.39 48.86
Type 2 5000 46 15 6.4 44.80
contours on the horizontal plane at H = 10 mm and the vertical
plane at h = 9° for the LM type [10] and the two fin-height profiles
266 D. Jang et al. / Applied Energy 116 (2014) 260–268

considered in this study. Table 2 shows the result of the numerical


(a) 1.0
analysis.
When the Type 1 profile was used in place of the LM type, the RTH
average heat sink temperature decreased by 4 °C. For the pin–fin Mass
0.8
array with the tallest fins in the inner region (Type 1), heated air
from the outermost fins rose in the direction of the chimney pat-
tern and heat transfer resumed at the second row of fins. The ther-
mal boundary layer of the flow developed steadily as the air 0.6
entered along the chimney flow path. Hence, a relatively high local

μ*
heat transfer coefficient was obtained on the upper sections in the
second row. In addition, a fresh inflow entered from the second 0.4
row of fins, since they were taller than the outermost fins. In the
radial direction, such flow characteristics were repeated, increas-
ing the mass flow rate by 44% compared to the LM type. Thus,
0.2
the cooling performance was better even though the mass of the
Type 1 was the same as that of the LM type.
When the Type 2 height profile was used in place of the LM
type, the heat sink temperature decreased by 8.3 °C. Compared to 0.0
NA LL LM LO LD
the Type 1 height profile, the ratio of the outermost fin area to
the total fin area increased by 90%. Under natural convection, a Geometric parameters
heat source causes the temperature of cooling air to rise, and thus
(b) 1.0
the air density decreases. A flow then develops toward the heat
source from the outer region. Therefore, the increased heat transfer RTH
area in the outermost region, which is the flow entrance of a radial Mass
0.8
heat sink, increased the driving force of the inflow. Consequently,
the mass flow rate increased by 46% and the cooling performance
was improved. Although the Type 1 profile offers the advantage
of repeated heat transfer along the chimney flow path, the in- 0.6
creased mass flow rate of the Type 2 profile had a greater influence
σ

on cooling performance. Therefore, the Type 2 model was the best


design among the fin arrays examined. 0.4
These results were compared with those of previous studies of
fin-height profiles. For impingement heat sinks [15–17], the pres-
sure drop in the central region due to flow resistance has a decisive 0.2
effect on flow penetration to the outer region, and the optimum
fin-height profile is obtained by considering this. However, natural
convection flow arises from the density difference due to increas-
0.0
ing temperature. Therefore, the cooling performance of the Type NA LL LM LO LD
2 profile in which the driving force was increased by the larger Geometric parameters
heat transfer area in the outer region was better than that of the
Type 1 profile, which had a lesser flow resistance in the outer re- Fig. 8. Design sensitivity analysis of the geometric parameters. (a) Mean of the
gion where the flow enters. elementary effects distribution (l⁄). (b) Standard deviation of the elementary
effects distribution (r).

4.3. Sensitivity analysis


Fig. 8 shows the sensitivity analysis of the design parameters
Since the optimization step entails multi-objective optimization with respect to mass and thermal resistance. The number of fin ar-
that takes into account cooling performance and mass, it is essen- rays (NA), the outermost fin height (LO), and the difference between
tial to select the influential parameters for the output. The influ- the fin heights (LD) were significant for thermal resistance and
ences of the number of fin arrays (NA), the long fin length (LL), mass. For these three design parameters, both the mean l⁄, which
the middle fin length (LM), the outermost fin height (LO), and the evaluates the overall effect of the design parameter on the outputs,
difference between fin heights (LD) on the mass and thermal resis- and the standard deviation r, which assesses the ensemble of the
tance were evaluated. The space between pin fins (Ls) and the sin- parameter’s effects due to interactions with other parameters,
gle pin–fin length (Lf) were excluded from the sensitivity analysis were significant. This means that the deviations of the elementary
because the influences of these parameters on the output were rel- effects were affected by the values of other parameters and exhib-
atively unimportant [12]. In addition, the smaller the space be- ited remarkable distinctions depending on the design points.
tween pin fins (Ls) and the single pin–fin length (Lf), the greater Although the difference between the fin heights (LD) was the
the leading-edge effects and the surface area per unit volume of third-most important parameter when only l⁄ was taken into con-
the fin. Therefore, Lf = 2 mm and Ls = 3 mm were fixed values in sideration, it was necessary to consider the parameters as design
the sensitivity analysis and the optimization procedure. The ele- variables rather than fixed parameters in the optimization step.
mentary effects method, which can be used to extract the impor- This is because the value of r for the difference between the fin
tant parameters from many design parameters, was employed for heights (LD) was largest with respect to cooling performance, and
the sensitivity analysis. This method [21,22] considers two sensi- thus was most affected by the values of the other parameters. In
tivity measures (l⁄, r) to select the design variables from many contrast, the long fin length (LL) and middle fin length (LM) had
interacting parameters. Normalized parameters and outputs were low r values and were nearly free of interactions. The three param-
adopted to exclude uncertainties induced by differing magnitudes. eters (NA, LO, LD) that markedly influenced the outputs were
D. Jang et al. / Applied Energy 116 (2014) 260–268 267

4.0 the weight factor for the normalized thermal resistance and x2 is
LM type Type 2 the weight factor for the normalized mass. The objective function
300W/m2 and constraints are described as follows:
700W/m2
1100W/m2
Minimize f ðX 1 ; X 2 ; X 3 Þ ¼ x1 RRTH ðXðX1 ;X 2 ;X 3 Þ
1 ;X 2 ;X 3 Þ
þ x2 MMðXðX1 ;X 2 ;X 3 Þ
1 ;X 2 ;X 3 Þ
TH;ref ref
RTH (°C/W)

subject to 20 6 X 1 6 28 ðX 1 is a natural numberÞ


2.5 40 6 X 2 6 56
LM type
X 3 6 NXL 1
2
ðNL ¼ Number of unit pin-fins in the long finÞ

The optimum design was carried out in the following sequence.


First, 25 design points were adopted for optimization using an
orthogonal array L25(53) to construct a Kriging model [25]. The Kri-
Type 2
ging model was constructed after numerical analysis of the 25 de-
1.0 sign points. The optimum point for X1, X2, and X3 was obtained
from the Kriging model by employing an evolutionary algorithm,
the specific parameters of which were as follows:
0.15 0.20 0.25 0.30
Mass (kg)  Population size: 50.
 Maximum number of generations: 1000.
Fig. 9. Comparison of the Pareto fronts for the LM type and Type 2 heat sinks
(ro = 0.075 m, T1 = 30 °C, e = 0).  Violated constraint limit: 0.003.
 Number of consecutive generations without improvement: 50.
 Mutation probability: 0.01.
 Selection probability: 0.15.
adopted as the design variables and the optimum design was based
on these results.
Fig. 9 compares the Pareto fronts of the Type 2 design and the
LM type for various heat fluxes. Multi-objective optimization was
4.4. Optimization carried out with various weight factors considering only natural
convection (e = 0). The thermal resistance was decreased by more
The number of fin arrays (X1 = NA), the outermost fin height than 50% with the Type 2 design, while preserving a mass similar
(X2 = LO), and the difference between the fin heights (X3 = LD) were to that of the LM type. This was because of the increased mass flow
adopted as the design variables. The pin–fin array with the tallest rate through the large heat transfer area in the outer region of the
fins in the outer region (Type 2) was selected as the reference mod- Type 2 model.
el. On the basis of a parametric study, the parameters for the refer- Table 3 shows the optimum results for the LM type and the
ence model were NA = 24, LO = 48 mm, LD = 3 mm, LL = 42 mm, Type 2 design with the same mass as the optimum LM type, taking
LM = 27 mm, Lf = 2 mm, Ls = 3 mm, NL = 9, ro = 75 mm, ri = 10 mm, both natural convection and radiation heat transfer into consider-
and t = 2 mm. Since the objective of this research was to design a ation. For the Type 2 heat sink, the thermal resistance was im-
radial heat sink for high-power LED products, it was necessary to proved by an average of 45%. At 1500 W/m2, which is more than
investigate various heat fluxes. In addition, the optimum configu- twice the reference heat flux of 700 W/m2, the heat sink tempera-
ration of a heat sink varies with the radiation heat transfer [11]. ture for the Type 2 model was approximately 20 °C lower than that
Therefore, the optimum design was performed with heat fluxes of LM type. Therefore, reliable cooling performance for high-power
of 300, 700, 1100, and 1500 W/m2, and emissivities of 0 and 0.9, LED products can be achieved by using a non-uniform fin-height
using commercial software, the process integration and design profile of a pin–fin heat sink and the same mass as the LM type. Al-
optimization (PIDO) tool known as PIAnO (Process Integration, most 25% of the total heat transfer was due to radiation. The opti-
Automation, and Optimization) [23]. mum configuration was obtained in the direction of maximizing
The objective function considers thermal resistance and mass radiation heat transfer, and thus the difference between the fin
simultaneously via a weighted sum method [24] in which x1 is heights (LD) was greater than in the case where only natural

Table 3
Optimization results when the mass is equal to that of the LM type heat sink (ro = 0075 m, T1 = 30 °C).

Model q_ (W/m2) e NA LO (mm) LD (mm) Mass (kg) Theatsink (°C) RTH (°C/W)

LM 300 0 – – – 0.291 45.1 2.899


Type 2 300 0 22 56 1 0.291 37.6 1.455
LM 300 0.9 – – – 0.240 39.6 1.843
Type 2 300 0.9 20 54.825 3.986 0.240 36.4 1.224
LM 700 0 – – – 0.308 57.6 2.268
Type 2 700 0 27 52.495 1.721 0.308 44 1.153
LM 700 0.9 – – – 0.269 50.6 1.695
Type 2 700 0.9 22 55.072 1.873 0.269 41.9 0.977
LM 1100 0 – – – 0.317 68.3 2.007
Type 2 1100 0 27 52.874 1.241 0.317 49.7 1.034
LM 1100 0.9 – – – 0.281 59.7 1.556
Type 2 1100 0.9 23 53.631 1.733 0.281 46.8 0.881
LM 1500 0 – – – 0.328 77.9 1.839
Type 2 1500 0 28 53.457 1.358 0.327 54.1 0.926
LM 1500 0.9 – – – 0.290 67.5 1.440
Type 2 1500 0.9 25 53.896 1.884 0.291 51.7 0.833
268 D. Jang et al. / Applied Energy 116 (2014) 260–268

convection was taken into account. Due to the radiation heat trans- [2] Leung CW, Probert SD. Heat-exchanger design: optimal uniform thickness of
vertical rectangular fins protruding perpendicularly outwards, at uniform
fer (e = 0.9), the thermal resistance diminished by an average of
separations, from a vertical rectangular ‘base’. Appl Energy 1987;26(2):111–8.
15% compared with the case where only natural convection was ta- [3] Leung CW, Probert SD. Heat-exchanger performance: effect of orientation.
ken into account (e = 0). However, there was a smaller improve- Appl Energy 1989;33(4):235–52.
ment in the thermal resistance in the case where only natural [4] Leung CW, Probert SD, Rapley CW. Natural convection and radiation from
vertically-based arrays of vertical, rectangular fins: a numerical model. Appl
convection was taken into consideration. This was because only Energy 1990;35(4):253–66.
natural convection was significantly improved, while the radiation [5] Leung CW, Probert SD. Heat-exchanger performance: influence of gap width
heat transfer did not change significantly. between consecutive vertical rectangular fin-arrays. Appl Energy
1997;56(1):1–8.
[6] Harahap F, Setio D. Correlations for heat dissipation and natural convection
5. Conclusion heat-transfer from horizontally-based, vertically-finned arrays. Appl Energy
2001;69(1):29–38.
[7] Elshafei EAM. Natural convection heat transfer from a heat sink with hollow/
In this paper, a pin–fin radial heat sink with a fin-height profile perforated circular pin fins. Energy 2010;35:2870–7.
was optimized. Natural convection and radiation heat transfer [8] Sertkaya AA, Bilir Sß , Kargıcı S. Experimental investigation of the effects of
orientation angle on heat transfer performance of pin-finned surfaces in
were taken into account, and an experiment was conducted to val-
natural convection. Energy 2011;36:1513–7.
idate the numerical model. Among the various fin-height profiles, [9] Yu SH, Lee KS, Yook SJ. Natural convection around a radial heat sink. Int J Heat
the pin–fin array with the tallest fins in the outer region (Type 2) Mass Transfer 2010;53:2935–8.
showed the best cooling performance. The sensitivity of various [10] Yu SH, Lee KS, Yook SJ. Optimum design of a radial heat sink under natural
convection. Int J Heat Mass Transfer 2011;54:2499–505.
parameters was investigated to determine the design variables, [11] Yu SH, Jang D, Lee KS. Effect of radiation in a radial heat sink under natural
which were the outermost fin height, the difference between fin convection. Int J Heat Mass Transfer 2012;55:505–9.
heights, and the number of fin arrays. Multi-objective optimization [12] Jang D, Yu SH, Lee KS. Multidisciplinary optimization of a pin-fin radial heat
sink for LED lighting applications. Int J Heat Mass Transfer 2012;55:515–21.
was carried out considering only natural convection, and the cool- [13] Fabbri G. A genetic algorithm for fin profile optimization. Int J Heat Mass
ing performance was improved by 50% with a design having the Transfer 1997;40:2165–72.
same mass as the LM type. In the case where both natural convec- [14] Kanyakam S, Bureerat S. Multi objective evolutionary optimization of splayed
pin-fin heat sink. Eng Appl Comput Fluid Mech 2011;5:553–65.
tion and radiation heat transfer were taken into account, the cool- [15] Shah A, Sammakia BG, Srihari H, Ramakrishna K. A numerical study of the
ing performance was improved by 15% compared with the case in thermal performance of an impingement heat sink-fin shape optimization.
which only natural convection was taken into account. However, IEEE Trans Compon Packag Technol 2004;27(4):710–7.
[16] Shah A, Sammakia BG, Srihari H, Ramakrishna K. Optimization study for a
no significant enhancement of the radiation heat transfer was ob-
parallel plate impingement heat sink. J Electron Packag 2006;128:311–8.
tained by using the Type 2 model instead of a plate-fin design, and [17] Yang Y, Peng H. Numerical study of pin-fin heat sink with un-uniform fin
thus the improvement in cooling performance was diminished in height design. Int J Heat Mass Transfer 2008;51:4788–96.
[18] Bello-Ochende T, Meyer JP, Bejan A. Constructal multi-scale pin–fins. Int J Heat
contrast to the case where only natural convection was taken into
Mass Transfer 2010;53:2773–9.
account. In summary, the cooling performance of a pin–fin radial [19] Carvalho MG, Faris T, Fontes P. Predicting radiative heat transfer in absorbing,
heat sink with a fin-height profile showed an improvement of emitting, and scattering media using the discrete transfer method. Fund Radiat
more than 45% while preserving a mass comparable to that of a Heat Transfer ASME HTD 1991;160:17–26.
[20] Bianco N, Langellotto L, Manca O, Nardini S. Radiative effects on natural
plate-fin heat sink. convection in vertical convergent channels. Int J Heat Mass Transfer
2010;53:3513–24.
Acknowledgements [21] Morris MD. Factorial sampling plans for preliminary computational
experiments. Technometrics 1991;33:161–74.
[22] Campolongo F, Cariboni J, Saltelli A. An effective screening design for
This research was supported by Basic Science Research Program sensitivity analysis of large models. Environ Model Software
through the National Research Foundation of Korea (NRF) funded 2007;22:1509–18.
[23] Framax INC. PIAnO (Process Integration, Automation and Optimization) User’s
by the Ministry of Education, Science and Technology (No. manual, Version 3.3. 2011.
2012R1A1B3000492). [24] Arora JS. Introduction to optimum design. United States of America: Elsevier
Academic Press; 2004.
[25] Sacks J, Welch WJ, Mitchell TJ, Wynn HP. Design and analysis of computer
References
experiments. Stat Sci 1989;4:409–35.

[1] Naik S, Probert SD, Wood CI. Natural-convection characteristics of a


horizontally-based vertical rectangular fin-array in the presence of a shroud.
Appl Energy 1987;28(4):295–319.

You might also like