Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

Int J Adv Manuf Technol (2011) 52:739–749

DOI 10.1007/s00170-010-2754-8

ORIGINAL ARTICLE

Multi-objective optimization of flat plate heat sink


using Taguchi-based Grey relational analysis
S. Manivannan & S. Prasanna Devi & R. Arumugam &
N. M. Sudharsan

Received: 31 January 2010 / Accepted: 25 May 2010 / Published online: 8 June 2010
# Springer-Verlag London Limited 2010

Abstract This paper presents an approach for the multi- width of the heat sink, fin height, base height, number of
objective optimization of the flat plate heat sink using fins, and fin thickness. The multi-objective optimization
Taguchi design of experiments-based Grey relational problem is then converted into single-objective optimization
analysis. The responses studied were electromagnetic problem using Grey relational analysis and the optimum
emitted radiation, thermal resistance, average heat transfer design settings of the heat sink geometry were obtained. Also,
coefficient, pressure drop, and the mass of the flat plate heat ANOVA test was carried out for finding out the contribution
sink. The heat sink is modeled using Ansoft High and impact of each heat sink design factor towards the
Frequency Structure Simulator (HFSS) software version multiple responses of the heat sink.
12 and the value of the emitted radiation is obtained by
simulation. The same heat sink is modeled using Flotherm Keywords Heat sink . Design optimization . Taguchi design
V7.2 software for finding the thermal resistance, pressure of experiments (DOE) . Grey relational analysis (GRA) .
drop, and average heat transfer coefficient. Experimental HFSS simulation . CFD simulation . Experimentation
investigation was performed to find the thermal resistance
and emitted radiations from the heat sink and thus the Nomenclature
simulation model was validated with the experimental l wavelength, m
results. The simulations were continued for the combina- ρAl density of the extruded aluminum, kg/m3
tions generated by the L27 (6 factors, three levels) a fin height, mm
Taguchi’s design of experiments using Minitab software. b base height of the heat sink, mm
The factors considered for optimization are the length and d fin thickness, mm
f frequency, Hz
k thermal conductivity of fin material, W/mK
S. Manivannan (*)
Department of Electrical and Electronics Engineering, L heat sink length, mm
Anna University, N number of fins
Chennai, India Q heat dissipation, W
e-mail: asmani25@yahoo.co.in
Rsin total thermal resistance, 0K/W
S. P. Devi Ta ambient temperature, 0K
Department of Industrial Engineering, Anna University, Tb temperature of fin base, 0K
Chennai, India Vf fan velocity, m/s
W heat sink width, mm
R. Arumugam
Department of Electrical and Electronics Engineering, SS sum of squares
SSN College of Engineering, DOF degrees of freedom
Chennai, India F, T test statistics
MS mean square
N. M. Sudharsan
Sarvajit–CAE, P value of probability
Chennai, India Pcritical table value of probability
740 Int J Adv Manuf Technol (2011) 52:739–749

1 Introduction and specified effect of more EMI problems when the


components become much closer.
The heat sinks are used in electronic systems to remove Richard Georgerian and Mantrose [8] studied the
heat from the chip and effectively transfer it to the ambient concept of wavelength-based heat sink fin length selection
[1]. The heat sink geometry is designed by the mechanical for the minimization of radiated emissions and simultaneously
engineers with the primary aim of reducing the thermal they presented an approach for the base temperature rise with
resistance of the heat sink [2, 3] for better cooling in the the variation in fin length, for maximizing the cooling of
electronic systems. Due to the proximity of the heat sink electronic packages.
with the integrated circuits (ICs), the radio frequency (RF) Bruce Archambeault et al. [9] discussed the effects of
fields created by RF currents in the ICs/PCBs gets coupled EMI by heat sink and presented a method to suppress its
to heat sinks. The fin of the heat sink effectively behaves as effects by grounding techniques.
monoples at high frequency. And hence, the coupled RF Junwei Lu, Xian Duan [10] presented a approach of
current can cause radiated emission. This radiated noise applying finite element frequency domain analysis for
from the device can couple and disturb the functioning of electromagnetic radiation emitted from high power micro-
the nearby electronic systems. Also, this radiated emission electronic circuits connected to heat sink and also investi-
from the device poses a problem to the system compliance gated various grounding options of the heat sink.
with respect to electromagnetic interference (EMI)/electro- Junwei Lu and Dawson [11] studied the EMC
magnetic compatibility (EMC) regulations. The internation- computer modeling techniques for CPU heat sink simula-
al EMI/EMC standards require the radiated emission from tion and presented an approach to model and simulate
the electronic devices to be kept below the specified limits. a heat sink. Philippe Sochoux et al. [12] modeled heat
Various techniques are practiced by the EMC engineers to sink using Micro stripes (Flomerics) to find radiations
suppress the radiated emission but reducing the emission emitted from the heat sink also found the variation in
from the source point is the most preferred. Hence, EMC radiation for the variation in design factors like height and
engineers need to understand the factors contributing to number of fins.
radiated emission from heat sink. It is important to analyze Visser et al. [13] described the use of mathematical
and optimize these factors, mainly the geometrical param- optimization techniques to minimize the heat sink mass
eters of heat sink to minimize the radiated emissions at the and thermal resistance using five design variables namely,
design stage itself. heat sink height, thickness, extrusion length, base thick-
Das and Roy [4] observed that the radiating efficiency ness, and number of fins for the heat sink. Culham and
not only depends on the clock frequency of the ASIC Muzychka [14] presented a procedure for optimization of
but also on the shape of the heat sink. The radiated heat sink design parameters based on minimization of the
emissions are most severe when the monopole length entropy generation associated with heat transfer and fluid
becomes quarter of the wavelength and vertical polarized friction.
emissions were found to be higher than the horizontal Shah et al. [15] examined the effect of the shape
polarization. of the heat sink fins, particularly near the center of the
Brench [5] examined the variations in the radiation heat sink. An optimum heat sink shape is reported that
characteristic of heat sink with respect to their geometries results in a lower operating temperature and pressure
by applying 3D-FDTD technique and found that there gradient.
is an additional increase in radiated emissions at the Khan et al. [16] studied the entropy generation minimiza-
resonance of the heat sink. John Parry [6] presented tion procedure and employed it to optimize the overall
approach for modeling heat sink and studied the effect of performance of micro channel heat sinks using Newton-
grounded heat sink and examined the common mode RF Raphson method.
coupling between die and heat sink and concluded that heat Bar-Cohen and Rohsenow [17] analyzed the thermally
sink radiating clock harmonics throughout the frequency optimum spacing of vertical natural convection cooled
spectrum. parallel plates. Bar-Cohen, Kraus and Bar-Cohen [18, 19]
Golkhab and Tavakoli Bina [7] studied the effect of has analyzed the state of art and trends in thermal analysis
reduction in space for the want of compactness in geometry of electronic packages. Chyi-Tsong Chen et al. [20] has

Table 1 Heat sink geometry


used for simulation Length of the Width of the Fin height Base height (mm) Number Fin thickness
heat sink (mm) heat sink (mm) (mm) of fins (mm)

90 70 30 6 17 1.2
Int J Adv Manuf Technol (2011) 52:739–749 741

done the parameter optimization of plate fin heat sink (with


side cooling fan) using genetic algorithm.
Lin [21] has studied the use of the Taguchi method and
Grey relational analysis (GRA) to optimize turning oper-
ations with multiple performance characteristics. Ko-Ta
Ching et al. [22] has studied the optimum design
parameters of pin-fin heat sink using the Grey fuzzy logic
based on the orthogonal arrays. The effects of design
parameters and the optimum design parameters for a pin-fin
heat sink with the multiple thermal performance character-
istics have been investigated by using the Grey fuzzy logic
based on the orthogonal arrays. Chih-Chung-Chou et al. Fig. 2 Far-field radiation pattern of heat sink
[23] studied the optimization of a parallel-plain fin heat
sink using the Grey-based fuzzy algorithm with the predicting the radiations emitted using Ansoft software.
orthogonal arrays. The study done by Arularasan and Velraj Section 3 illustrates the CFD modeling, simulation and
[24] presents the modeling and simulation of parallel plate experimentation of heat sink for finding multiple thermal
heat sinks using computational fluid dynamics (CFD) and performance characteristics. Section 4 deals with optimiza-
DOE considers the factors like number of fins (N), fin tion of heat sink geometry using Taguchi L27 design of
length (L), fin height (a), base height (b) of the heat sink to experiments based Grey relational analysis. Section 5
optimize the thermal resistance of the heat sink. presents the results and discussions followed by the
In the present study, the authors focus in optimizing conclusions in Section 6.
heat sink geometry factors such as length (L), width (W) of
the heat sink, height of the fin (a), base height of the heat
sink (b), number of fins (N), and fin thickness (d) to 2 HFSS modeling, simulation, and experimentation
minimize the total radiations emitted, thermal resistance, of heat sink for finding radiated emissions
pressure drop, mass, and to maximize the average heat
transfer coefficient of a heat sink. In this paper, the Taguchi A typical aluminum heat sink with the geometry specifica-
design of experiments is used to generate the various tions available in Table 1 is modeled using Ansoft HFSS
geometry combinations of the heat sink and the various version 12 [25]. The model essentially consists of three
performance characteristics for each combination were regions namely the heat sink region, source region and the
found using HFSS and Flotherm simulations and cor- ground plane. The aluminum ground plane is modeled with
responding experimental investigations were performed in the dimensions 150×135×1 mm below the heat sink at a
order to validate the simulation results. The simulation is distance of 5 mm. The typical flat plate heat sink with “N”
continued for L27 orthogonal array to perform the multi- number of fins is shown in Fig. 1.
objective optimization of the flat plate heat sink using Grey Lumped port of impedance 50Ω is defined to represent
relational analysis. the coupling of processor to the heat sink. Also, there is no
In this paper, Section 2 explains the HFSS modeling, any additional grounding point for this processor to
simulation and experimentation of the heat sink for suppress the radiations to avoid additional cost and space

Fig. 1 Typical flat plate heat sink model Fig. 3 Schematic diagram of experimental setup
742 Int J Adv Manuf Technol (2011) 52:739–749

Fig. 4 Photograph of experimental setup

restrictions. The radiation boundary is defined to specify an


open boundary condition that allows waves to radiate Fig. 6 CFD modeling of heatsink using flotherm
infinitely into space. The radiation boundary in the present
problem is defined as box with faces positioned at least a test setup consist of the components such as (1) log periodic
quarter wave length (l/4) from the objects in the model. antenna, (2) signal generator, (3) turn table, (4) insulation
The total emissions from the heat sink are evaluated in the material, (5) heat sink and ground plane, and (6) spectrum
far-field region by defining an infinite sphere that surrounds analyzer.
the radiating object. The source power level to excite heat The heat sink is excited using the Agilent-MXG-analog
sink is kept at 1 mW. The simulation was carried for the signal generator and swept for the range of frequencies
frequencies ranging from 1 to 10 GHz with a interval of between 1 and 10 GHz with a RF output of 1 mW (0 dBm)
500 MHz. The emissions are obtained in terms of dBmV/m at 50 Ω impedance. RF SMA cable was used to connect the
at a distance of 3 m from the source. signal generator to the N connector of the base plate via
The 3D far-field radiation pattern for the E field at SMA-N adaptor. A log periodic antenna of SAS make 200/
3 GHz of the heat sink and the boundary are shown in 518 was placed at a distance of 3 m from the heat sink to
Fig. 2. The same aluminum heat sink with the specifica- receive the radiations emitted. The antenna is vertically
tions given in Table 1 is used for the experimental polarized and connected to Agilent spectrum analyzer
investigation. An aluminum ground plane of dimensions which is kept in the control room via N-type RF cable.
150×135 ×1 mm is placed at 5 mm below the heat sink, Figure 4 shows the photograph of the experimental setup.
with an N-type male connector fixed at the center for the Signal generator is set to provide RF signal with a range of
excitation. The center pin of the N-type connector is 1-10 GHz with a step size of 500 MHz. The radiated
extended and soldered to the base of the heat sink with a emissions from the heat sink are measured using spectrum
copper wire. The radiation from a heat sink is experimen- analyzer. The radiated emission from the heat sink obtained
tally measured in a shielded semi-anechoic chamber. The via experimentation and simulation were plotted with
set-up for the experimental analysis is shown in Fig. 3. The frequency and emitted radiation (dBmV/m) along X and
Y, respectively, and is shown in Fig. 5.

Table 2 Inlet boundary conditions and material properties

Parameter Value

Ambient temperature 301 0K


Thermal conductivity of extruded 209 W/m0K
aluminum heatsink
Density of extruded aluminum 2684.9 kg/m3
Conductivity of air 0.026274 W/m0K
Viscosity of air 0.00001824 kg/ms
Density of air 1.1758 kg/m3
0
Specific heat capacity of air 1005.49967 Ws/kg K
Fig. 5 Comparison of simulation vs. experimental results
Int J Adv Manuf Technol (2011) 52:739–749 743

Fan

Heat sink

J Type
Thermocoupl

Heater

Insulating
base
Fig. 7 Temperature distribution plot of original heat sink

Fig. 9 Schematic diagram for the experimental arrangement


The non-parametric Wilcoxon signed-ranks test is con-
ducted to validate the presently available simulation and transfers in the heat sink with a fan of 70 mm opening is
experimental results using SPSS software. The null hy- shown in Fig. 6.
pothesis assumes that the median of difference between the The heat sink with specifications in Table 1 is modeled
experimental and simulation results is zero. The results using Flotherm. A heat source of size 3×3 cm is modeled
indicate that the null hypothesis is accepted with a and attached to the bottom of heat sink and a layer of
confidence interval of 95%. Hence, the simulation results Wakefield [27] thermal compound with a thermal resistivity
are validated with experimental results. of 0.735 W/mK is applied between heat sink and heat
source to provide a proper surface contact between them.
The inlet boundary conditions and material properties are
3 CFD modeling, simulation, and experimentation given to Flotherm software as inputs are given in Table 2.
of heat sink for finding multiple thermal performance The program is set to run for 500 iterations. The base
characteristics temperature of the heat sink is measured at various six
points at heat sink base. The heater load is varied from 10
The thermal performance characteristics of a heat sink such to 80 W at an interval of 10 W. For each heat loads, the
as the thermal resistance, pressure drop, and average heat base temperatures are recorded and average base tempera-
transfer coefficient are found using CFD simulations. A ture is found. The corresponding thermal resistance values
numerical model was formulated using commercial CFD are obtained using Eq. 1. The inlet and outlet pressures are
package Flotherm Version 7.2 [26] for solving 3D heat recorded to find the pressure drop in the heat sink. Also, the

1. Computer

2. Power Supply for the


6 heater set up.

3. Heat Sink With fan


attached to Heater.

1 4. J Type Thermocouple
2 3 5. Data Logger
5
6. Power Supply for the
4 fan arrangement

Fig. 10 Experimental setup for measuring the base temperature of the


Fig. 8 Pressure distribution plot of original heat sink heatsink
744 Int J Adv Manuf Technol (2011) 52:739–749

Table 3 Levels of various factors of the heat sink Where


Design Parameter Level 1 Level 2 Level 3
Δ T ¼ T b  Ta ð2Þ
(low) (medium) (high)

Length of the heat 70 80 90 Mass ¼ ½ðbLWÞ þ ðLdaNÞ  109  rAl ð3Þ


sink (L) mm
Width of the heat 70 80 90
sink (W) mm Figures 7 and 8 shows the temperature distribution and
No of fins (N) 10 20 30 pressure distribution plots, respectively, of the CFD
Fin height (a) mm 10 20 30 simulation using Flotherm software, with a heat load of
Base height (b) mm 4 6 8 80 W. From the average values of base temperature, the
Fin thickness (d) mm 0.8 1 1.2 simulated value of thermal resistance of a heat sink is found
as 0.2368 0 K/W.
Full-scale experimentation is performed for the heat sink
value of average heat transfer coefficient is recorded from with geometry specified in Table 1 by mimicking the Watt
the CFD simulations. density, i.e., the heat dissipated by various processors from
10 to 80 W. An experimental setup is arranged with an
electric heater of size 30×30 mm as a heat source to mimic
Rsin ¼ ΔT =Q ð1Þ a processor and it is supplied by DC power supply. The

Table 4 Taguchi L27 heat sink combinations with corresponding response

Heat sink design factors Multiple responses studied

L27 Length Width Fin Base Fin Number h pressure Rsin Mass Radiation
experiments height Height thickness of fins drop

1 70 70 10 4 0.8 10 38.67926 31.61299 0.881965 67.65948 11.303


2 70 70 10 4 1 20 43.87016 41.82888 0.432559 90.21264 11.379
3 70 70 10 4 1.2 30 49.2269 62.23041 0.293341 120.2835 12.93
4 70 80 20 6 0.8 10 26.99681 9.272992 0.702373 120.2835 12.144
5 70 80 20 6 1 20 33.72965 14.49167 0.323024 165.3898 11.94
6 70 80 20 6 1.2 30 30.88444 24.47018 0.212195 225.5316 13.371
7 70 90 30 8 0.8 10 23.11338 3.185494 0.645489 180.4253 14.852
8 70 90 30 8 1 20 23.37154 5.79142 0.299137 248.0848 15.364
9 70 90 30 8 1.2 30 23.01993 11.65755 0.191255 338.2974 15.702
10 80 70 20 8 0.8 20 38.59814 16.74324 0.269675 189.017 14.207
11 80 70 20 8 1 30 46.27413 34.55881 0.179465 249.1587 14.525
12 80 70 20 8 1.2 10 35.74262 13.72804 0.551366 171.8336 14.43
13 80 80 30 4 0.8 20 32.31988 6.708182 0.271091 171.8336 12.626
14 80 80 30 4 1 30 24.78749 12.08745 0.177363 262.0462 13.327
15 80 80 30 4 1.2 10 22.80031 4.340119 0.528742 146.0586 12.122
16 80 90 10 6 0.8 20 34.93076 18.25632 0.448907 150.3544 15.29
17 80 90 10 6 1 30 37.0762 24.11352 0.299522 180.4253 15.65
18 80 90 10 6 1.2 10 35.02478 18.07321 0.818506 141.7627 15.02
19 90 70 30 6 0.8 30 25.1672 14.47585 0.163617 275.4707 16.716
20 90 70 30 6 1 10 24.008 5.435299 0.455578 173.9815 15.738
21 90 70 30 6 1.2 20 27.42544 17.54251 0.202934 275.4707 16.406
22 90 80 10 8 0.8 30 36.81901 40.7776 0.241451 212.6441 15.96
23 90 80 10 8 1 10 30.28531 28.62623 0.650075 178.8143 15.02
24 90 80 10 8 1.2 20 30.28531 28.62623 0.650092 212.6441 15.73
25 90 90 20 4 0.8 30 23.10386 10.0281 0.212911 202.9784 15.746
26 90 90 20 4 1 10 22.89866 5.845856 0.591928 135.319 15.06
27 90 90 20 4 1.2 20 24.3832 9.193638 0.287743 202.9784 15.47
Int J Adv Manuf Technol (2011) 52:739–749 745

heater is fixed with heat sink using Anabond compound from 10 to 80 W, in the interval of each 10 W. The
with thermal conductivity of 0.437 W/mK [28] and Wake- temperatures at various base locations of heat sink are
field thermal compound with thermal conductivity of measured and recorded using data logger and computer.
0.735 W/mK is applied to the contact surface of the sink Also, the values of voltage and current of the heaters power
to make proper surface contact between heat sink and the supply are noted to find the total power applied to the heater.
heater. The heat sink rejects the heat into the air which is Anemometer is used to find the fan velocity at 12 V and it is
enhanced by placing the axial fan above the heat sink. The found to be 5.5 m/s.
axial fan with an opening of 70 mm diameter is placed on From the average values of base temperature, the experi-
top the heat sink and the fan is charged by 12 V DC supply. mental value of thermal resistance of a heat sink is found as
The bottom side of the heater is insulated so as to ensure 0.2447 0K/W and the deviation with simulation results are
that all the heat from the heater is dissipated through the very minimal (3.22%). Hence, simulation model is found to
heat sink only. Six numbers of J-type thermocouples are be in good agreement with the experimental model.
suitably placed in the base of the heat sink so as to measure
the base temperature of the heat sink and a separate
thermocouple is placed on the experimental bench to 4 Multi-objective optimization of heat sink geometry
measure the ambient temperature. These thermocouples using Taguchi-based Grey relational analysis
are connected to the data logger and computer setup (data
acquisition system). Figure 9 shows the schematic repre- To provide efficient optimization, traditional optimization
sentation of the arrangement for the experimentation. techniques like design of experiments are generally chosen
The photograph of experimental setup is shown in Fig. 10. as a tool. The heat sink geometry design factors like length
The experiment was conducted by varying the heat loads of the heat sink (L), width of the heat sink (W), height of

Table 5 Data processing of each performance characteristic

Normalized h Normalized pressure drop Normalized Rsin Normalized mass Normalized emitted radiation

0.601031897 0.518581 0 1.000000074 1


0.797507983 0.345547 0.625652234 0.916666734 0.98596
1.000261166 −6.9E-06 0.819467841 0.805555615 0.699427
0.158849563 0.896968 0.250024363 0.805555615 0.844633
0.413688545 0.808576 0.778144577 0.638888936 0.88232
0.305997153 0.639563 0.932438396 0.416666697 0.617957
0.0118614 1.000076 0.329216205 0.583333376 0.344356
0.021632906 0.955938 0.811399485 0.333333358 0.249769
0.008324507 0.856579 0.961589865 0 0.187327
0.597961477 0.77044 0.852415773 0.551587342 0.463514
0.888498348 0.468685 0.97800362 0.329365104 0.404766
0.489879648 0.82151 0.460251984 0.615079411 0.422317
0.360328419 0.94041 0.850444452 0.615079411 0.755588
0.075226619 0.849298 0.980930322 0.281746053 0.626085
1.16677E-05 0.98052 0.491748225 0.710317513 0.848698
0.459150557 0.744812 0.60289329 0.694444496 0.26344
0.540355946 0.645604 0.810863149 0.583333376 0.196933
0.462709229 0.747913 0.088346095 0.72619053 0.31332
0.089598961 0.808844 1.000067172 0.232142874 0
0.045722817 0.96197 0.593605736 0.607142902 0.180676
0.175073512 0.756902 0.94533099 0.232142874 0.05727
0.530621103 0.363354 0.891707852 0.464285749 0.139664
0.283320006 0.56917 0.322832382 0.589285758 0.31332
0.283320006 0.56917 0.322808019 0.464285749 0.182154
0.011501273 0.884179 0.931440554 0.500000037 0.179198
0.003734223 0.955016 0.403783238 0.750000055 0.30593
0.059924152 0.898312 0.827261242 0.500000037 0.230187
746 Int J Adv Manuf Technol (2011) 52:739–749

the fin (a), base height of the heat sink (b), number of fins the heat sink by simulation and specified in Table 4.
(N), fin thickness (d) are chosen as the design parameters Similarly, CFD simulations is set to run for the 27
(six factors with three levels (low, medium, high)) for the combinations at a heat load of 80 W to find the thermal
optimization. The minimum number of experimental performance characteristics such as the thermal resistance,
combinations (MNE) for conducting simulations are given pressure drop, average heat transfer coefficient. The values
by MNE=kn, where k is the number of levels and n is the of emitted radiation with multiple thermal performance
number of factors. Full factorial method gives 729 (36) characteristics are specified along with its mass in Table 4.
experiments. Since it is difficult and tedious to do all the In the GRA, the multiple responses obtained are first
experiments/simulations, Taguchi orthogonal array method normalized in the range between zero and one, which is
is used instead of full factorial method to reduce the number also called the Grey relational generation shown in Table 5.
of experiments/simulations which gives similar results as in Next, the Grey relational coefficient is calculated from
full factorial method [29]. Hence, we use the L27 the normalized experimental data to express the relationship
orthogonal array with the six factors at three levels between the desired and actual experimental data. In our
generated using Minitab software [30] and it is given in study, a linear data preprocessing method [31] for the
Table 3. emitted radiations, thermal resistance, pressure drop and
Since it is not feasible to fabricate all the heat sink mass is chosen as lower-the-better and is calculated using
geometry combinations for the experimental investigations, Eq. 4 and for the average heat transfer coefficient, higher-
we have opted for the HFSS simulations for the combina- the-better is chosen and calculated using Eq. 5.
tions generated using L27 Taguchi method. HFSS simu-
lations are set to run for 3 GHz for all the 27 combinations
generated by design of experiments. The total radiations max yi ðkÞ  yi ðkÞ
xi ðkÞ ¼ ð4Þ
from the heat sink are obtained at a distance of 3 m from max yi ðkÞ  min yi ðkÞ

Table 6 Grey relational


coefficient of each performance h Pressure drop Rsin Mass Radiation
characteristic
0.556193 0.509466 0.333333 1.00E+00 1
0.711752 0.433106 0.571855 8.57E-01 0.972686433
1 0.333333 0.734719 7.20E-01 0.624552902
0.372814 0.829144 0.400008 7.20E-01 0.762931642
0.460273 0.723145 0.692659 5.81E-01 0.809481083
0.418759 0.5811 0.880962 4.62E-01 0.56686564
0.33599 1 0.427064 5.45E-01 0.43265926
0.338211 0.919013 0.72611 4.29E-01 0.399926117
0.335194 0.777097 0.92866 3.33E-01 0.380902118
0.5543 0.685344 0.7721 5.27E-01 0.482399073
0.817659 0.484818 0.957861 4.27E-01 0.456523573
0.494991 0.736931 0.480886 5.65E-01 0.463958173
0.438723 0.893512 0.769757 5.65E-01 0.671671423
0.350933 0.768401 0.963262 4.10E-01 0.572138252
0.333333 0.9625 0.495908 6.33E-01 0.767692526
0.480377 0.662087 0.557347 6.21E-01 0.404347501
0.521027 0.585209 0.725545 5.45E-01 0.383710215
0.482025 0.664817 0.354194 6.46E-01 0.421343504
0.354509 0.723426 1 1.00E+00 0.333333333
0.343813 0.929316 0.551636 5.60E-01 0.378982007
0.377379 0.672859 0.901438 3.94E-01 0.346565081
0.515794 0.439891 0.821973 4.83E-01 0.367556189
0.410954 0.537155 0.424748 5.49E-01 0.421343504
0.410954 0.537155 0.42474 4.83E-01 0.379407023
0.335909 0.811924 0.879416 5.00E-01 0.378557941
0.334165 0.917458 0.456114 6.67E-01 0.418735979
0.347204 0.830996 0.743231 5.00E-01 0.393758638
Int J Adv Manuf Technol (2011) 52:739–749 747

yi ðkÞ  min yi ðkÞ Table 8 Response table for the Grey relational grade (grand
xi ðkÞ ¼ ð5Þ mean=0.585603)
max yi ðkÞ  min yi ðkÞ
Level 1 Level 2 Level 3 Max - Min average

Where xi (k) is the value after the Grey relational L 0.620592 0.590175 0.546399 0.074193 0.585722
generation, min yi (k) is the smallest value of yi (k) for the W 0.634777 0.579177 0.552814 0.081963 0.588923
kth response, and max yi (k) is the largest value of yi (k) for a 0.569304 0.585285 0.594939 0.025635 0.583176
the kth response. Table 6 gives the sequences for Grey b 0.626403 0.581828 0.5427 0.083703 0.583644
relational coefficient for the performance characteristics d 0.605666 0.588773 0.562728 0.042938 0.585722
which is calculated using Eq. 6. N 0.571631 0.587828 0.599825 0.028195 0.586428

Δmin þ z Δmax  
xi ðkÞ ¼ ð6Þ
Δoi ðkÞ þ z Δmax Δmin ¼ 8jmin 2 i8k min x0 ðkÞ  xj ðkÞ is the smallest

value of Δoi and Δmax ¼ 8jmax 2 i8k max x0 ðkÞ  xj ðkÞ is
the largest value of Δoi.
Where Δoi ¼ kx0 ðkÞ  xi ðkÞk is the difference of the Then, the Grey relational grade is computed by
absolute value between x0 (k) and xi (k). ζ is the averaging the Grey relational coefficient corresponding to
distinguishing coefficient between 0 and 1 (we have chosen each response using Eq. 7.
it to be equal to 0.5).
1X n
gi ¼ x ðkÞ ð7Þ
n k¼1 i
Table 7 Grey relational grades
and its order L27 Grey relational
combinations grade The overall evaluation of the multiple responses is based
on the Grey relational grade. As a result, optimization of
1 0.679798569 the complicated multiple process responses can be con-
2 0.709308359 verted into optimization of a single Grey relational grade. In
3 0.682521087 other words, the Grey relational grade can be treated as the
4 0.61697951 overall evaluation of experimental data for the multi-
5 0.653240833 response process. Table 7 shows the Grey relational grade
6 0.581845105 for the L27 combinations.
7 0.548233675 The mean of the Grey relational grade for each level of
8 0.562366375 the heat sink factors can be calculated by averaging the
9 0.551037141 Grey relational grade for Length of heat sink for experiment
10 0.604268083 number 1-9 as level 1, experiment number 10-18 as level 2,
11 0.62879614 and experiment number 19-27 as level 3. Similarly, it is
12 0.548357527 calculated for the respective levels for the other factors and
13 0.667737099 in addition, the total mean of the Grey relational grade for
14 0.613031453 the 27 experiments is also calculated and is shown in
15 0.63851999 Table 8.
16 0.54496958
17 0.552189112
18 0.513706719
19 0.682253623
20 0.552749553
21 0.538521582
22 0.525594607
23 0.468644128
24 0.447002874
25 0.581161207
26 0.558628041
27 0.563037823
Fig. 11 Graph of Grey relational grade
748 Int J Adv Manuf Technol (2011) 52:739–749

Table 9 Shows the ANOVA


table for the heat sink geometry Factors DOF SS MS F P %Contribution
design parameters
Length 2 0.02505 0.012525 15.56542 0.000276 21.87881
Width 2 0.03153 0.015765 19.5913 0.000088 27.53769
Fin height 2 0.00303 0.001515 1.882763 0.188743 2.646427
Base height 2 0.03158 0.01579 19.62299 0.000087 27.58223
Fin thickness 2 0.00843 0.004215 5.238183 0.02003 7.362831
No of fins 2 0.00361 0.001805 2.24316 0.142877 3.153004
Error 14 0.01127 0.000805 9.839256
Total 26 0.114494 100.0002

5 Results and discussions geometry. From the results, it is inferred that combining
Taguchi method and GRA yields better results for the
The Grey relational grade graph showing the different optimal design of heat sink geometry.
factors at different levels contributing to the mean Grey
relational grade is shown in Fig. 11. The larger the value of
the Grey relational grade, the better is the multi-response
characteristics. The optimal combination of parameter 6 Conclusions
values are: L1W1a3b1d1N3.
Also the conceptual approach from Table 9 is confirmed Optimization of multiple performance characteristics of the
by the analysis of variance test shown in Table 9 which heat sink such as the average heat transfer coefficient,
indicates that the factors such as the length (21.87%), base pressure drop, thermal resistance, mass and the radiations
height (27.58%), and width (27.53%) of the heat sink are emitted has been carried out using Taguchi-based GRA
the most influencing factors affecting the responses, as technique. The factors considered for the L27 orthogonal
shown from their percentage contribution. Also, the factors array construction are the length, width, height of the fin,
such as fin height, fin thickness, and number of fins base height, number of fins, and fin thickness of the heat
contribute only to a small extent to the combined sink. The contribution of each factor towards the multi-
optimization of the performance characteristics of the heat response optimization has been studied and it is inferred
sink. Table 10 shows the original vs. optimum design that combining Taguchi method and GRA are best suitable
settings of the heat sink parameters obtained from the for the optimal design of heat sink geometry.
Taguchi-based GRA. Also, Table 11 indicates the percent-
age improvement obtained for each performance character-
istics using HFSS and CFD simulations for the optimum

Table 11 Comparison of performance parameters for original vs.


optimum heat sink

Table 10 Comparison of original vs. optimum heat sink geometry Performance Parameters Original Optimum %Improvement
of the heat sink design design
Heat sink geometry factors Original Optimum
design design Average heat 26.82 28.90 7.19
transfer coefficient
Length (L), mm 90 70 (W/m2K)
Pressure drop (Pa) 16.59 16.24 2.10
Width (W), mm 70 70
Thermal resistance 0.2368 0.1903 19.63
Fin height (a), mm 30 30
(0K/W)
Base height (b), mm 6 4 Mass (gm) 249.37 187.943 24.63
Number of fins (N) 17 30 Radiations emitted, 16.12 11.869 26.37
Fin thickness (d), mm 1.2 0.8 dBmV/m
Int J Adv Manuf Technol (2011) 52:739–749 749

References 15. Shah A, Sammakia B, Srihari H (2002) A numerical study of the


thermal performance of an impingement heatsink fin shape
optimization. ITHERM 298–306
1. Tummala R (2001) Fundamentals of microelectronic packaging. 16. Khan WA, Yovanovich MM, Culham JR (2006) Optimization of
McGraw-Hill Professional microchannel heatsinks using entropy generation minimization
2. Shih CJ, Liu GC (2004) Optimal design methodology of plate-fin method. IEEE International Conference on Semiconductor Thermal
heat sinks for electronic cooling using entropy generation strategy, Measurement and Management Symposium 78–86
components and packaging technologies. IEEE Transactions 27 17. Bar-Cohen A, Rohsenow WM (1984) Thermally optimum spacing
(3):551–559 of vertical, natural convection cooled, parallel plates. J Heat
3. Teertstra P, Yovanovich MM, Culham JR, Lemczyk T (1999) Transfer 106:116–123
Analytical forced convection modeling of plate fin heat sinks. in 18. Bar-Cohen A (1992) State-of-the-art and trends in the thermal
Proc. 15th IEEE SEMI-THERM Symp packaging of electronic equipment. J Electron Package 114:257–
4. Das R (1998) An investigation on radiated emissions from heat 270
sink. IEEE Int Symp Electromagn Compat 2:784–789 19. Kraus AD, Bar-Cohen A (1995) Design and analysis of heatsinks.
5. Brench (1994) Heat sink radiation as a function of geometry. in Wiley, New York
proceedings. IEEE Symp, Electromoagnetic Compatibility 105– 20. Chen C-T, Wu C-K, Hwang C (2008) Optimal design and control of
109 CPU heat sink processes, components and packaging technologies.
6. Parry J (2000) Multiphysics modeling for electronics design. IEEE Transactions 31(1):184–195
Therm Thermo Phen Elect Sys, ITTHERM 2:86–93 21. Lin CL (2004) Use of the Taguchi method and Grey relational
7. Mohammad G, Mohammad TB (2008) “Multilevel converter analysis to optimize turning operations with multiple performance
objectives: a critical evaluation and combination of available characteristics. Mater Manuf Process 19(2):209–220
natural commuted topologies with restructured iron cores”, 22. Chiang K-T, Chang F-P, Tsai T-C (2006) Optimum design
Proceedings of world congress on Engineering and computer parameters of pin-fin heat sink using the grey-fuzzy logic based
science WCEC 428–433 on the orthogonal arrays. Int Commun Heat Mass Transf 33:744–
8. Georgerian R, Mantrose I (2003) Product safety and the heat sink— 752
dilemma of minimizing radiated emissions and maximizing thermal 23. Chou C-C, Liu N-M, Horng J-T, Chiang K-T (2009) Designing
cooling. IEEE Int Symp Electromagn Compat 1:134–137 parameter optimization of a parallel-plain fin heat sink using the
9. Archambeault B, Pratapneni S, Zhang L, Wittwer DC, Chen J grey-based fuzzy algorithm with the orthogonal arrays. Int J
(2001) A proposed set of specific standard EMC problems to help Therm Sci 48:2271–2279
engineers evaluate EMC modeling tools. IEEE Int Symp Electro- 24. Arularasan R, Velraj R (2008) Modeling and simulation of a
magn Compat 2:1335–1340 parallel plate heat sink using computational fluid dynamics. Int J
10. Lu J, Xiao D (2007) Comparative analysis of Intel Pentium IV and Adv Manuf Tech. doi:10.1007/s00170-008-1867-9
IEEE/EMC TC-9/ACEM CPU heat sinks. IEEE International 25. HFSS V12.0, Ansoft corporation http://www.ansoft.com/products/
Symposium on Electromagnetic Compatibility 1–6 hf/hfss/
11. Lu J, Dawson F (2006) EMC computer modeling techniques for 26. Flotherm - Fluent software tutorial available at www.fluent.com
CPU heat sink simulation. IEEE Trans Magn 42(10):3171–3173 27. Wakefield thermal compound datasheet available at http://www.
12. Sochoux P, Yu J, Bhobe A, Centola F (2008) Heat sink design wakefield.com/pdf/accessories.pdf
flow for EMC. Design Con, IEC publications 1–27 28. Anabond adhesive data sheet available at http://www.anabond.
13. Visser, JA, de Kock DJ, Conradie FD. Minimisation of heatsink com/anabondImages/anabond.pdf
mass using mathematical optimization. IEEE Semiconductor 29. Douglas C. Montgomery, Design and analysis of experiments,
Thermal Measurement and Management Symposium 252–259s Fifth edition. Wiley: New York. ISBN 9971-51-329-3
14. Culham JR, Muzychka YS (2000) Optimization of plate fin 30. Minitab 15 software tutorial available at: http://www.minitab.com
heatsinks using entropy generation minimization. In Proc. IEEE 31. Wang Z, Zhu L, Wu JH (1996) Grey relational analysis of
Intersociety Conf. Thermal Phenomena correlation of errors in measurement. J Grey Syst 8(1):73–78

You might also like