7 tayangan

Diunggah oleh endoparasite

Shape optimization of micro channel heat sink

- OPTIMAL PIN FIN HEAT EXCHANGER SURFACE
- Thermal Re Sis Trance and Colling System
- LED akademi
- PPT OF DESIGN OF EXHAUST GAS HEAT EXCHANGER
- Experiment 8
- Air Flow Bench
- A Review of Computational Fluid Dynamics Analysis of Cylinder Head Water Jacket of a Diesel Engine
- 1.Forced External Convection
- Advanced Immersed Boundary Cartesian Meshing Technology in FloeFd
- A preliminary analysis for development of a high-temperature gas-phase solar receiver
- IRJET-Review On Fractal Image Compression Techniques
- Thesis 1
- amplificador 1 THD stk4432
- A numerical study of the impact of surface roughness.pdf
- EtraLED-9680 Modular Passive LED Star Heat Sink Φ96mm
- APDS100tips
- CLU03H- 25／ 55
- Intro. CheLab
- Numerical prediction of total resistance for fishing vessel
- EtraLED-LUME-9620 Lumens Modular Passive Star LED Heat Sink Φ96mm

Anda di halaman 1dari 9

2, JUNE 2008

Sink for Micro-Electronic Cooling

Afzal Husain and Kwang-Yong Kim

Abstract—A numerical investigation of 3-D fluid flow and heat Coolant flow rate.

transfer in a rectangular micro-channel has been carried out using Re Reynolds number.

water as a cooling fluid in a silicon substrate. Navier–Stokes and

energy equations for laminar flow and conjugate heat transfer are Thermal resistance.

solved using a finite volume solver. Solutions are first carefully Convective resistance.

validated with available analytical and experimental results; the

shape of the micro-channel is then optimized using surrogate Adjusted value of R square.

methods. Ratios of the width of the micro-channel to the depth Temperature.

and the width of the fin to the depth are selected as design vari-

ables. Design points are selected through a four-level full factorial Liquid velocity in micro-channel.

design. A single objective function thermal resistance, formulated Velocity vector.

using pumping power as a constraint, is optimized. Mass flow rate

is adjusted by the constant pumping power constraint. Response Micro-channel width.

surface approximation, Kriging, and radial basis neural network Fin width.

methods are applied to construct surrogates and the optimum

point is searched by sequential quadratic programming. Orthogonal coordinate system.

Index Terms—Electronic cooling, micro-channel, numerical sim- Greek Symbols

ulation, optimization, surrogate methods.

Normalized design variables, and ,

respectively.

NOMENCLATURE Micro-channel aspect ratio.

Cross section area of micro-channel. Density.

Surface area of substrate base. Dynamic viscosity.

Specific heat. Kinematic viscosity.

Hydraulic diameter. Subscripts

Friction factor. Liquid.

Convective heat transfer coefficient. Inlet.

Height of heat sink. Outlet.

Micro-channel depth. Substrate.

Thermal conductivity. max Maximum value.

Entry length. avg Average value.

Length of heat sink.

Width of heat sink. I. INTRODUCTION

Height of heat sink.

Number of micro-channels. R ECENT developments in micro electromechanical sys-

tems (MEMS) and advanced very large-scale integration

(VLSI) technologies and devices associated with micro minia-

Number of dimensions in design space.

turization have led to significant improvement in packing

Pressure. densities. These developments have helped satisfy growing

Perimeter. demand for higher dissipation of heat flux from electronic

Pumping power. devices. However, it has been observed that operation of most

electronic devices is strongly influenced by their temperature

Heat flux.

and their surrounding thermal environment. Micro-channel

heat sink, as an integrated part of silicon based electronic

Manuscript received December 21, 2006; revised August 15, 2007. This work device, is a potential solution to this problem. Sophisticated

was recommended for publication by Associate Editor S. Bhavnani upon eval- fabrication processes have yielded economically competitive

uation of the reviewers comments. micro-channels having a high surface area to volume ratio.

The authors are with the Department of Mechanical Engineering, Inha Uni-

versity, Incheon 402-751, Korea (e-mail: kykim@inha.ac.kr). The potential of micro-channel heat sinks as heat transfer de-

Digital Object Identifier 10.1109/TCAPT.2008.916791 vices has motivated many researchers to analyze micro-cooling

1521-3331/$25.00 © 2008 IEEE

HUSAIN AND KIM: SHAPE OPTIMIZATION OF MICRO-CHANNEL HEAT SINK 323

Pease [1] first realized the potential of this technology and

laid a foundation for silicon based micro-channel heat sink

experimentation. They experimented on a 56 m wide, 320 m

deep micro-channel fabricated by a chemical etching process.

Samalam [2] reported correlations for thermal resistance based

on a theoretical study of experiments of Tuckerman [3]. Fe-

dorov and Viskanta [4] carried out a numerical simulation

based on the experiments of Kawano et al. [5]. While numerous

investigations have been conducted using various techniques

[6]–[10], there are significant disparities between the existing

experimental data and the values predicted using classical

Fig. 1. Schematic diagram of micro-channel heat sink and design variables.

macro techniques [11].

The results of the aforementioned studies indicate that the

cross section parameters and fin width of the micro-channel can

thermal resistance of the micro-channel is then searched by

greatly influence the behavior of the fluid flow and resulting

sequential quadratic programming.

heat transfer capabilities of rectangular micro-channels [9],

[10], [12]. In addition, experimental investigations [1], [7], [9],

[10] reported a pressure drop in the convective heat transfer for II. PROBLEM DESCRIPTION AND GEOMETRIC CONSTRAINTS

water flowing in rectangular micro-channels of various aspect A schematic of the rectangular micro-channel heat

ratios. These results suggest that the aspect ratio is an important sink optimized in the current study is shown in Fig. 1.

parameter in the determination of flow friction and convective The dimensions of the heat sink under consideration are

heat transfer. 10 mm 10 mm 0.5 mm. The thickness of the base of the

Some analytical studies [13]–[15] have focused on modeling micro-channel is 100 m while the depth of the micro-channel

of the heat transfer and optimization of the micro-channel geom- is kept constant at 400 m. Simulations are performed

etry. Weisberg et al. [16] presented a design algorithm for the for varying fin width and channel width. A uniform heat flux

selection of a rectangular micro-channel heat exchanger using a ( 100 W/cm ) is applied at the bottom of the heat sink to

2-D conjugate heat transfer model. Qu and Mudawar [17] com- elucidate the effect of micro-channel geometry on the thermal

puted the 3-D fluid flow and heat transfer for a micro-channel resistance and friction factor. The flow is assumed to be lam-

with a rectangular cross section. Li et al. [18] numerically inves- inar and fully developed and is maintained by low flow rates

tigated 3-D conjugate heat transfer in a silicon based micro-heat and low Reynolds numbers. Since the focus of the study is to

sink. Toh et al. [19] carried out a detailed numerical study of optimize the microchannel geometry with surrogate analysis,

variation of local thermal resistance and friction factor along the thermodynamic and hydrodynamic properties of the substance

flow direction in micro-channels by solving 3-D Navier–Stokes are assumed to be constant with a reference temperature of

equations. They extended the length of the micro-channel to en- 27 C in all simulations.

sure realistic boundary conditions at the exit. As an attempt to One of the major challenges in micro-channel optimization is

optimize the shape of the micro-channel, Li and Peterson [20] manufacturing feasibility. The optimal design should not be im-

carried out a parametric study on the geometry of the micro- practical from manufacturing and design points of view. A sil-

channel. An exhaustive review of the literature, however, shows icon based micro-channel can be fabricated with an aspect ratio

that systematic optimization techniques have not yet been ap- of up to 20:1 using DRIE [23], [24] and 6:1 using

plied to this problem. KOH wet etching [1], [5], [24]. On the basis of strength, the

With the aid of high performance computers, the last decade constraint thickness of the base of the micro-channel is kept as

has witnessed rapid development of design optimization tech- 100 m, which is well above the minimum required thickness,

niques. Numerical optimization methods [21] are regarded as as suggested by Li and Peterson [20]. Li and Peterson [20] found

general design tools and offer a number of advantages, including that thermal resistance decreases with an increase in the aspect

automated design capability, varieties of constraints, and mul- ratio, and therefore is kept constant for all optimization cases

tidisciplinary applications. However, due to large computing in order to assess the effect of the micro-channel width and fin

time, coupling with Navier–Stokes analyses has recently proven width on the thermal resistance and friction factor. A constant

practical. Surrogate models are widely used in multidisciplinary pumping power is applied to the micro-channel heat sink, and

optimizations. Queipo et al. [22] reviewed various surrogate is defined as

based models used in aerospace applications.

The current investigation explores the application of surro- (1)

gate based optimization techniques for the shape optimization

of a rectangular micro-channel cross-section to minimize where is the volumetric flow rate across the heat sink and

thermal resistance. Response Surface Approximation (RSA), is the pressure drop. indicates the number of channels,

Kriging (KRG), and radial basis neural network (RBNN) is the average velocity, and is the cross-sectional area of the

methods are used to construct surrogates, and the minimum micro-channel.

324 IEEE TRANSACTIONS ON COMPONENTS AND PACKAGING TECHNOLOGIES, VOL. 31, NO. 2, JUNE 2008

sink and leaves at the outlet; the remainder of the heat sink is

occupied by the silicon substrate. The silicon part of the heat

sink at the inlet and outlet of the channel is maintained as an

adiabatic boundary. No-slip condition is applied at the interior

walls of the channel, i.e., 0. Average velocity obtained by

the method described later in this section is used at the inlet

of the microchannel, and velocity gradients are neglected at the

exit. The thermal conditions in the z-direction are

III. NUMERICAL ANALYSIS boundary conditions.

Formulation of the average velocity in terms of pumping

In the present study, steady, incompressible, laminar flow

power is described as follows. Fanning friction is defined as

and conjugate heat transfer are considered with constant

fluid (water) properties (density 997 kg/m , spe-

(6)

cific heat capacity 4.179 J/kg K, dynamic viscosity

8.55 10 kg/ms, and thermal conductivity

0.613 W/mK at a reference temperature of 27 C). Continuum where is the hydraulic diameter. The Reynolds number in

equations for conservation of mass, momentum, and energy for terms of average velocity can be defined as

the convective heat transfer in the micro-channel can be written

in vector form as (7)

Mass

where

(2)

Momentum

For a fully developed flow, Knight et al. [13] determined that

(3)

(8)

Energy

where

(4)

(5)

ature, and and are the substrate temperature and thermal

conductivity, respectively. The average velocity can be obtained from (6), (7), and (8) as

A numerical model is formulated to solve the 3-D conjugate follows:

heat transfer in micro-channels using commercial code CFX

5.7 [25]. The code uses finite volume discretization of gov- (9)

erning differential equations and the solution is based on the

SIMPLE algorithm [26]. In the numerical solution, the con-

Substituting (1) into (9) we obtain

vective terms are discretized using a first-order upwind scheme

for all equations. Due to symmetry of the problem, half of the

single micro-channel is selected for computation, as shown in (10)

Fig. 2. A hexahedral mesh is generated in the specified domain

through ICEM 5.7 [25] and a 401 61 16 grid is used for

where the pumping power should be realistic and can be taken

the current simulation cases. Grid independency is checked by

as being in a range of 0.01 to 0.8 considering the capacity of

taking different grids: it was found that for a 501 71 21 grid

state-of-the-art micro pumps [14]. For comparison of the results

the change in the highest temperature in the substrate was only

with experimental data, the thermal resistance is calculated as

0.03% while for a 301 41 11 grid the difference in highest

temperature attained was roughly 2%. A finite volume solver is

used under the following boundary conditions. (11)

HUSAIN AND KIM: SHAPE OPTIMIZATION OF MICRO-CHANNEL HEAT SINK 325

is the maximum temperature rise in the heat sink and is

defined as

(12)

is the fluid inlet temperature. Thermal resistance at the inlet is

defined as

and outlet, respectively.

Fig. 3 represents the various steps involved in the optimiza-

tion algorithm. Design points are selected using a four-level full

Fig. 3. Optimization procedure.

factorial design. Two design variables, and , are chosen for

the optimization methodology, where is defined as

and is defined as . Hence, 16 design points chosen TABLE I

from four-level full factorial design assisted with 4 more de- DESIGN VARIABLES AND DESIGN SPACE

sign points are used to construct the surrogates. These levels

are equally spaced within the design range for all variables. De-

sign space is specified performing some preliminary calcula-

tions within the geometric constraints discussed in Section II

to obtain minimum of thermal resistance. The design space is

given in Table I. Design variables are normalized from 0 to 1

to construct surrogate models. Objective functions are calcu-

lated by solving Navier–Stokes and heat conduction equations The KRG model, also known as Kriging metamodelling [28],

at specified design points and optimized using surrogate models. is a deterministic technique for optimization. A linear polyno-

In the current study, thermal resistance is the foremost con- mial function with a Gauss correlation function is used for the

sideration for optimization and is the objective function for the model construction. The Kriging postulation is a combination

surrogate based optimization i.e., . of a global model and departures of the following form:

Three surrogate models, response surface approximation

(RSA), Kriging (KRG), and radial basis neural network (14)

(RBNN), are applied to predict the optimal design point. In

RSA [27], a least square curve fitting by regression analysis where represents the unknown function, is the known

is performed on the data obtained by computational fluid dy- function of , and is the realization of a stochastic process

namic (CFD) calculations using Navier–Stokes equations. The with mean zero and non-zero covariance. A linear function,

following polynomial function is fitted to obtain the response , is fitted first, and real points are interpolated on it to ob-

surface function. If the regression coefficients are ’s, the tain the mean zero. Here, is the global design space while

polynomial function becomes is the localized deviations. The covariance matrix can be

written as

(13)

where is the number of design variables, and ’s are the de- (15)

sign variables. For a second order polynomial model, used in

the current study, the number of regression constants is where is the number of dimensions in the set of design vari-

. ables is the standard deviation of the population, and is

326 IEEE TRANSACTIONS ON COMPONENTS AND PACKAGING TECHNOLOGIES, VOL. 31, NO. 2, JUNE 2008

direction .

A radial basis neural network (RBNN) [29] uses a linear com-

bination of n radially symmetric functions, , for the re-

sponse function as

(16)

of radial basis functions (typically Gaussian), and is the set of

errors with equal variance, . Due to its linear nature, RBNN

has simpler mathematics and lower computational cost as com-

pared to back-propagation neural network (BPNN) [29]. RBNN

is a two layer network that consists of a hidden layer of the radial

basis function and a linear output layer. The design parameters

for this function are spread constant (SC) and a user defined

error goal (EG). The SC value is selected such that it should

not be so large that each neuron does not respond in the same

manner for the all inputs, and that it should not be so small that

the network becomes highly sensitive for every input within the

design space. EG or mean square error goal selection is also im-

portant. A very small error goal will produce over-training of the

network while a large error goal will affect the accuracy of the

model. The allowable error goal is decided from the allowable

error from the mean input responses. Cross-validation [22] is

performed to check the quality of constructed surrogates. These

surrogate methods are implemented using MATLAB [30].

Fig. 4. Comparison of numerically predicted fully developed velocity profiles

V. RESULTS AND DISCUSSION with analytical results: (a) on z middle plane and (b) on y middle plane.

developed laminar flow. The velocity obtained from (10) is ap-

plied at the inlet of the micro-channel. In order to ensure a Kawano et al. [5]. Fig. 5 shows a comparison of the numeri-

fully developed flow, lower pumping power, 0.05 W, to cally calculated inlet thermal resistance and outlet thermal resis-

a 10 mm 10 mm chip has been chosen for the optimization tance with available experimental results [5] for a wide range of

cases. To validate the assumption of fully developed flow, the Reynolds number. The present results show appreciable agree-

entrance length is calculated for the micro-channel flow. For ment with the experimental results, with most of the predictions

a typical case of 0.15 and 0.1, the entrance length lying within the experimental uncertainties (indicated by error

is found to be 6% of the total channel length which is in good bars in the figure). The present model underpredicts the inlet

agreement with the relation developed by Langhaar [31] for flow thermal resistance only for a low Reynolds number. One of the

in a circular tube. possible reasons for this underprediction may be heat loss to

the ambient by the solid substrate, as discussed by Qu and Mu-

0.057 (17) dawar [17]. Moreover, the low Reynolds number conditions are

unreliable because of the larger temperature-induced viscosity

It is found that except the extreme case the entrance length is gradients near the inlet portion [5]. Further validation of the nu-

less than one fifth of the length of micro-channel heat sink, merical model is done with experimental results of Tuckerman

therefore assumption of fully developed flow is acceptable at and Pease [1] for three different cases of different channel depth

low pumping power [20]. The numerical model is validated in and heat flux, as given in Table II. In light of the numerical re-

a number of ways to ensure the accuracy of the numerical solu- production of these experimental results by Toh et al. [19] and

tions and the design optimization. Fig. 4 presents a comparison Liu et al. [8], the present model shows good agreement with the

of the numerical model with available analytical results [32] at experimental results, even for higher heat flux.

two different planes perpendicular to each other. Non-dimen- Before carrying out the actual optimization, the effect of

sional velocity profiles are plotted along the y and the design variables on the thermal resistance and friction

z directions, respectively. This comparison reveals good agree- constant is assessed. Figs. 6 and 7 show the variations of the

ment between the numerical and analytical results for the ve- thermal resistance and friction constant with changes in the

locity distribution in different directions. Experimental valida- design variables. Convective thermal resistance is defined

tion of the numerical results is done with the experiment of as 1 , where is the convective heat

HUSAIN AND KIM: SHAPE OPTIMIZATION OF MICRO-CHANNEL HEAT SINK 327

inlet thermal resistance and (b) outlet thermal resistance.

Fig. 6. Variation of (a) thermal resistance and (b) friction constant with design

TABLE II variable .

COMPARISON OF THERMAL RESISTANCES BETWEEN

COMPUTATIONAL AND EXPERIMENTAL RESULTS

increase of . For lower values of , where the flow resistance

is significant, an increase in further increases the flow re-

sistance, resulting in increased thermal resistance. For higher

values of , an increase in reduces pressure drop but increases

the velocity, which results in lower thermal resistance. For

intermediate values of , a decrease in the thermal resistance is

observed with an increase in up to a certain value, because

an increase in the fin thickness results in higher velocity, which

in turn results in higher convective heat transfer. A reduced

area subjected to convective heat transfer results from a further

increase in , and decreases convective heat transfer through

the micro-channel. A higher velocity reduces the friction factor

but results in a higher effective friction constant. It is natural

that the friction constant appears to be unaffected by any change

in the value of over the entire range of . An increase in

transfer coefficient. With an increase of , the average velocity pressure drop is countered by an increase in average velocity.

increases. Furthermore, the area subjected to convective heat The RSA, KRG, and RBNN surrogates are constructed using

transfer reduces as the number of micro-channels decreases the training data from a four-level full factorial design. Some

in a specified heat sink width for a constant under constant additional CFD data is added to enhance the performance of the

pumping power. Therefore, the thermal resistance decreases as surrogate models. The results of optimization of the thermal re-

is increased. This trend continues until the velocity dominates sistance using the surrogate models are shown in Table III. In the

the convective heat transfer and then the thermal resistance RSA method, analysis of variance (ANOVA) and a regression

increases when the area subjected to convective heat transfer analysis, provided by t-statistics in [7], are implemented to mea-

is significantly reduced. The increased velocity and reduced sure the uncertainty in the set of coefficients in the polynomial.

328 IEEE TRANSACTIONS ON COMPONENTS AND PACKAGING TECHNOLOGIES, VOL. 31, NO. 2, JUNE 2008

optimal point.

goal and the spread constant in order to obtain the minimum

PRESS (Prediction Error Sum of Squares) for training data.

Prediction errors are calculated by cross-validation of the

constructed model predictions at the design points. In the

RBNN model, spread constant and error goals a set at 0.51

and 1 10 , respectively, to train the network. Consistent

performance of the network is checked with variation of the

spread constant. Objective function values are calculated at the

surrogate predicted optimal points, as shown in Table III. The

optimum values obtained by a Navier–Stokes analysis at the

design points predicted by all methods are almost identical.

RSA gives the optimum design as 0.671 0.430.

Fig. 7. Variation of (a) thermal resistance and (b) friction constant with design In contrast, the optimum design from the KRG and RBNN

variable .

models was 0.492 0.226 and 0.490 0.306,

respectively. The optimum design can help the designer find

TABLE III the design variables the geometry of his/her choice. For a

OPTIMAL POINTS (NORMALIZED) PREDICTED BY DIFFERENT SURROGATES micro-channel heat sink of channel depth 360 m, the

AND CORRESPONDING CFD CALCULATED VALUES adjusted number of channels for the optimum design calculated

from the RBF model is 120, which is in line with the findings

of Li and Peterson [20].

All the surrogate models under study predict almost same ob-

jective function value, although their predicted values of the de-

sign variables are different. Moreover, the CFD predicted objec-

tive function values are also very close. This characteristic of the

design space shows the relatively less sensitivity of the objective

function near the optimum point. Surrogate predictions depend

upon the nature and suitability of the problem, therefore each

In the present study, was maintained at 0.967, which can be surrogate predicts different optimum point. The use of multiple

considered reliable in reference to the value of 0.91 1.0 surrogates helps to understand the insight of the design space

suggested by Guinta [33] for accurate prediction of the response and suitability of the surrogate for the kind of problem under

surface model. The functional relationship between objective study.

function and design variables is established by RSA as A sensitivity analysis of the objective function is performed

by varying the design variables around the optimum design.

Each design variable is varied from the optimum point in both

(18) directions while keeping the other variables fixed. The objective

function values at these sets of design variables are calculated

The KRG model is prepared with the help of toolbox (DACE) using a surrogate model (typically RSA). The objective function

[34] in MATLAB. Correlation function parameters are adjusted (thermal resistance) increases sharply with a change in while

carefully to ensure consistent performance of the model. The keeping fixed. On the other hand, change of has a smaller

acceptable variance of the method was kept at 3.93 10 . effect on the objective function for a fixed , as shown in Fig. 8.

HUSAIN AND KIM: SHAPE OPTIMIZATION OF MICRO-CHANNEL HEAT SINK 329

Fig. 10. Temperature contours on the middle y–z plane: (a) for [1] case and (b)

for optimum case.

a silicon wafer is minimized for a constant heat source and

constant pumping power. Three different surrogate models, i.e.,

RSA, KRG, and RBNN, are employed for the optimization. De-

sign variables related to micro-channel depth and fin width are

selected to construct the surrogates, which are used to predict

the minimum of the objective function (thermal resistance of

the heat sink). The three surrogate models yielded somewhat

different optimum geometries, but predicted almost the same

objective function values. The objective function is found to be

more sensitive to channel width to depth ratio than fin width to

depth ratio around the optimal point.

Fig. 9. Temperature contours on the middle x–z plane: (a) for [1] case and (b)

for optimum case. REFERENCES

[1] D. B. Tuckerman and R. F. W. Pease, “High-performance heat sinking

for VLSI,” IEEE Electron Device Lett., vol. EDL-2, no. 5, pp. 126–129,

It can be seen that the optimal design is highly sensitive to May 1981.

[2] V. K. Samalam, “Convective heat transfer in microchannels,” J. Elec-

as compared to in the specified range. Therefore, the design tron Mater., vol. 18, pp. 611–617, 1989.

variable can be suitably adjusted for the optimum number of [3] D. B. Tuckerman, “Heat Transfer Microstructures for Integrated Cir-

cuits,” Ph.D. dissertation, Stanford Univ., Stanford, CA, 1984.

channels in order to obtain minimum thermal resistance for the [4] A. G. Fedorov and R. Viskanta, “Three dimensional conjugate heat

specified channel depth of the heat sink. transfer in the microchannel heat sink for electronic packaging,” Int.

J. Heat Mass Transf., vol. 43, pp. 399–415, 2000.

Figs. 9 and 10 present a comparison of the temperature dis- [5] K. Kawano, K. Minakami, H. Iwasaki, and M. Ishizuka, “Development

tributions between [1] and optimum shapes on x–z and y–z of microchannels heat exchanging,” in Application of Heat Transfer in

planes, respectively. These temperature distributions for the ge- Equipment Systems, and Education, R. A. Nelson, Jr., L. W. Swanson,

M. V. A. Bianchi, and C. Camci, Eds. New York: ASME, 1998, vol.

ometry of case 1 in [1] and for the optimum shape predicted by HTD-361-3/PID-3, pp. 173–180.

response surface approximation are calculated under constant [6] L. T. Hwang, I. Turlik, and A. Reisman, “A thermal module design for

advancing packaging,” J. Electron. Mater., vol. 16, no. 5, pp. 347–355,

pumping power and heat source. By minimizing the thermal re- 1987.

sistance, the optimum shape shows lower maximum tempera- [7] M. M. Rahman and F. Gui, “Experimental measurements of fluid flow

ture, which occurs at the bottom of the substrate near the exit of and heat transfer in microchannel cooling passages in a chip substrate.,”

Adv. Electron. Packag., vol. ASME EEP-4, no. 2, pp. 685–692, 1993.

the micro-channel, in comparison with the reference shape. [8] D. Liu and S. V. Garimella, “Analysis and optimization of the thermal

performance of microchannel heat sinks,” Int. J. Numer. Methods Heat

VI. CONCLUSION Fluid Flow, vol. 15, no. 1, pp. 7–26, 2005.

[9] H. Y. Wu and P. Cheng, “Friction factors in smooth trapezoidal silicon

A 3-D rectangular micro-channel heat sink has been geomet- micro channels with different aspect ratios,” Int. J. Heat Mass Transf.,

vol. 46, no. 14, pp. 2519–2525, Jul. 2003.

rically optimized for minimum thermal resistance using sur- [10] H. Y. Wu and P. Cheng, “An experimental study of convective heat

rogate models. Fluid flow and heat transfer analyses are con- transfer in silicon micro channels with different surface conditions,”

Int. J. Heat Mass Transf., vol. 46, no. 14, pp. 2547–2556, Jul. 2003.

ducted by solving 3-D Navier–Stokes and heat conduction equa- [11] S. V. Garimella and C. B. Sobhan, “Transport in micro channels-A

tions to find the overall thermal resistance of the heat sink. The critical review,” Annu. Rev. Heat Tansf., vol. 13, pp. 1–50, 2003.

330 IEEE TRANSACTIONS ON COMPONENTS AND PACKAGING TECHNOLOGIES, VOL. 31, NO. 2, JUNE 2008

[12] X. F. Peng and G. P. Peterson, “Convective heat transfer and flow [31] A. Bejan, Convective Heat Transfer, 1st ed. New York: Wiley, 1984.

friction for water flow in microchannel structures,” Int. J. Heat Mass [32] R. K. Shah and A. L. London, Laminar Flow Forced Convection in

Transf., vol. 39, no. 12, pp. 2599–2608, Jul. 1996. Ducts: A Source Book for Compact Heat Exchanger Analytical Data.

[13] R. W. Knight, D. J. Hall, J. S. Goodling, and R. C. Jaeger, “Heat New York: Academic, 1978.

sink optimization with application to micro channels,” IEEE Trans. [33] A. A. Guinta, “Aircraft Multi-Disciplinary Design Optimization Using

Compon., Hybrids, Manufact. Technol., vol. 15, no. 5, pp. 832–842, Design of Experimental Theory and Response Surface Modeling

Oct. 1992. Methods,” Ph.D. dissertation, Dept. Aerosp. Eng., Virginia Poly-

[14] X. Wei and Y. Joshi, “Optimization study of stacked micro-channel technic Institute and State University, Blacksburg, May 1997.

heat sinks for micro-electronic cooling,” IEEE Trans. Compon. Packag. [34] S. N. Lophaven, H. B. Nielsen, and J. Sondergaard, “DACE-A

Technol., vol. 26, no. 1, pp. 55–61, Mar. 2003. MATLAB Kriging Toolbox,” Technical University of Denmark,

[15] T. S. Fisher and K. E. Torrance, “Optimal shapes of fully embedded IMM-TR2002-12, 2002.

channels for conjugate cooling,” IEEE Trans. Adv. Packag., vol. 24,

no. 4, pp. 555–562, Nov. 2001.

[16] A. Weisberg, H. H. Bau, and J. N. Zemel, “Analysis of micro channels Afzal Husain received the B.E. and M.Tech. degrees

for integrated cooling,” Int. J. Heat Mass Transf., vol. 35, no. 10, pp. in mechanical engineering with specialization in

2465–2474, 1992.

thermal sciences from Aligarh Muslim University,

[17] W. Qu and I. Mudawar, “Analysis of three dimensional heat transfer

in micro-channel heat sinks,” Int. J. Heat Mass Transf., vol. 45, no. 19, Aligarh, India, in 2003 and 2005, respectively, and

pp. 3973–3985, 2002. is currently pursuing the Ph.D. degree in thermo-

[18] J. Li, G. P. Peterson, and P. Cheng, “Three-dimensional analysis of dynamics and fluid mechanics in Inha University,

heat transfer in a micro heat-sink with single phase flow,” Int. J. Heat Incheon, Korea.

Mass Transf., vol. 47, pp. 4215–4231, 2004. His research interests are thermal analysis of mi-

[19] K. C. Toh, X. Y. Chen, and J. C. Chai, “Numerical computation of fluid crosystems (MEMS), electronic cooling, and surro-

flow and heat transfer in microchannels,” Int. J. Heat Mass Transf., vol. gate based analysis and optimization.

45, pp. 5133–5141, 2002.

[20] J. Li and G. P. “Bud” Peterson, “Geometric optimization of a micro

heat sink with liquid flow,” IEEE Trans. Compon. Packag. Technol.,

vol. 29, no. 1, pp. 145–154, Mar. 2006.

[21] G. N. Vanderplaats, Numerical Optimization Techniques for Engi- Kwang-Yong Kim received the Ph.D. degree from

neering Design with Applications. New York: McGraw-Hill, 1984. the Korea Advanced Institute of Science and Tech-

[22] N. V. Queipo, R. T. Haftka, W. Shyy, T. Goel, R. Vaidyanathan, and nology (KAIST), Daejon, Korea, in 1987.

P. K. Tucker, “Surrogate-based analysis and optimization,” Progress

Presently, he is a Professor in School of Me-

Aerosp. Sci., vol. 41, pp. 1–28, 2005.

[23] F. Laermer and A. Urban, “Challenges, developments and applications chanical Engineering, Inha University, Incheon,

of silicon deep reactive ion etching,” Microelectron. Eng., vol. 67–68, Korea. He published more than 120 research pa-

pp. 349–355, 2003. pers in professional journals, and presented 76

[24] M. J. Madou, “MEMS Fabrication,” in MEMS Handbook, M. Gad-el- papers at international conferences and 156 pa-

Hak, Ed. Boca Raton, FL: CRC, 2002, vol. 16-1. pers at domestic conferences. He also published

[25] Solver Theory, CFX-5.7, ANSYS, 2004. 60 technical reports which were supported by a

[26] S. V. Patankar, Numerical Heat Transfer and Fluid Flow. New York: variety of research grants from government and

McGraw-Hill, 1980, pp. 124–134. industries, and has four domestic patents. He is presently the Editor-in-Chief

[27] R. H. Myers and D. C. Montgomery, Response Surface Method- of the Transactions of Korean Society of Mechanical Engineers (KSME). His

ology: Process and Product Optimization Using Designed Experi- recent research works have been concentrated on applications of numerical

ments. New York: Wiley, 1995. optimization techniques using computational fluid dynamics (especially the

[28] J. D. Martin and T. W. Simpson, “Use of Kriging models to approx-

three-dimensional Reynolds-averaged Navier–Stokes analysis techniques) to

imate deterministic computer models,” AIAA J., vol. 43, no. 4, pp.

853–863, 2005. designs of engineering systems or devices, such as heat transfer augmentation

[29] M. J. L. Orr, “Introduction to Radial Basis Function Networks,” Ctr. devices, components of thermal-hydraulics system in various nuclear reactors,

Cogn. Sci., Edinburgh Univ., Edinburgh, U.K., 1996 [Online]. Avail- turbomachinery blades, micro-mixers, micro heat sinks, etc., where the use of

able: http://anc.ed.ac.uk/RBNN/ numerical optimization techniques was not popular.

[30] The Language of Technical Computing, Release 14. The Math Works, Dr. Kim is Chief Vice President of the Korean Fluid Machinery Association

Inc., 2007. (KFMA).

- OPTIMAL PIN FIN HEAT EXCHANGER SURFACEDiunggah olehteguh hady a
- Thermal Re Sis Trance and Colling SystemDiunggah olehaksaltaf9137
- LED akademiDiunggah olehEuodias Edhy Wilieyanto
- PPT OF DESIGN OF EXHAUST GAS HEAT EXCHANGERDiunggah olehAnkit saxena
- Experiment 8Diunggah olehIamzura Abdullah
- Air Flow BenchDiunggah olehAhmad Syami
- A Review of Computational Fluid Dynamics Analysis of Cylinder Head Water Jacket of a Diesel EngineDiunggah olehIRJET Journal
- 1.Forced External ConvectionDiunggah olehvarshasdm1987
- Advanced Immersed Boundary Cartesian Meshing Technology in FloeFdDiunggah olehGerardo Franck
- A preliminary analysis for development of a high-temperature gas-phase solar receiverDiunggah olehMohamadreza Sedighi
- IRJET-Review On Fractal Image Compression TechniquesDiunggah olehIRJET Journal
- Thesis 1Diunggah oleheduardo_um
- amplificador 1 THD stk4432Diunggah olehJuan Moya Blas
- A numerical study of the impact of surface roughness.pdfDiunggah olehnle15
- EtraLED-9680 Modular Passive LED Star Heat Sink Φ96mmDiunggah olehleo
- APDS100tipsDiunggah oleh조용규
- CLU03H- 25／ 55Diunggah olehIgnacio Javier Ramírez González
- Intro. CheLabDiunggah olehpaolo
- Numerical prediction of total resistance for fishing vesselDiunggah olehTu Tran
- EtraLED-LUME-9620 Lumens Modular Passive Star LED Heat Sink Φ96mmDiunggah olehAlex Thomas
- Electrasil_1Diunggah olehhashusklc
- CFD Advance Physical ModelsDiunggah olehmatteo_1234
- 040301_1Diunggah olehlaukik
- Advanced Soft Switching Inverter for Reducing Switching and Power LossesDiunggah olehzeevko1249
- Analysis of a Novel CBHE - FGDiunggah olehDiego Barra
- 11891_010212162426Diunggah olehigor collins
- s00170-017-1387-6Diunggah olehRohit Mathur
- fulltextDiunggah olehReza KH
- SunapwarDiunggah olehAnonymous mXicTi8hB
- Untitled 1Diunggah olehMohamed Karam

- Performance OptimizationDiunggah olehendoparasite
- Experimental Investigation on Heat Transfer of finned tubeDiunggah olehendoparasite
- ExtrapolationDiunggah olehendoparasite
- interpolasi Lecture3Diunggah olehAnita Andriani
- High-performance heat sinking for VLSI.pdfDiunggah olehendoparasite
- 2008 LeePS Saturated Flow Boiling HT in MCHSDiunggah olehendoparasite
- Ada 395503Diunggah olehendoparasite
- Road Prince SBS FlyerDiunggah olehendoparasite
- Sloshing DynamicsDiunggah olehendoparasite
- Rocket Motor Design and Missile Flight Performance AnalysisDiunggah olehbabak3316
- AMOM Lecture4 - Viscoelasticity p1Diunggah olehendoparasite
- AMOM Lecture1 - FundamentalsDiunggah olehendoparasite
- Solving Nonlinear Governing Equations of Motion Using Matlab and Simulink in First Dynamics CourseDiunggah olehendoparasite
- Part5.pdfDiunggah olehendoparasite
- Linearization of non-linear ordinary diff equationDiunggah olehendoparasite
- AP-lec1Diunggah olehendoparasite
- DocDiunggah olehAdrian Mansur
- Wind Turbine Blade Element MomentumDiunggah olehCarloAlbertoCimatti
- Performance Optimization .pdfDiunggah olehendoparasite
- UCAV Configuration & Performance Trade-OffsDiunggah olehendoparasite
- Effect of Propellant Sloshing on the Design of Space Vehicle Propellant Storage SystemsDiunggah olehendoparasite
- The Evolution of Wind Turbine Design Analysis-A Twenty Year Progress ReviewDiunggah olehendoparasite
- Performance Evaluation MIMO Base Station Antenna DesignsDiunggah olehendoparasite
- FlightTrajectoryOptimizationUsingGeneticAlgorithmCombinedDiunggah olehendoparasite
- De_CompressorsDiunggah olehendoparasite
- Lecture RocketDynamicsDiunggah olehendoparasite
- Aerodynamic CoefficientsDiunggah olehDragan Lazic

- Colour SpacesDiunggah olehGigi Pedala
- MI 2088 Earth Insulation TesterDiunggah olehRommel Saavedra
- Camera ReadyDiunggah olehNay Win Htoon
- 01 TENDERDOCS (1).docxDiunggah olehKundan Shandilya
- Nokia N95 8GB Extended UG EnDiunggah olehterzobanco
- Knowledge ManagementDiunggah olehMuhammad Abuzar
- LiDAR Mapping and UAV Systems v11 OptDiunggah olehHassaan Ahmad
- Lms Proposal Moodle FinalDiunggah olehZain Ul Abedin Saleem
- Here Are Some Sample Functional Requirements for an HRISDiunggah olehSui Yui
- AT&T Reply to Wilkins 12-15-16Diunggah olehshgustin
- Loti IntegrationDiunggah olehadiba
- First YearDiunggah olehpiyamraza
- The Performance Analysis Of Four Tank System for Conventional Controller and Hybrid ControllerDiunggah olehIRJET Journal
- Design and Fabrication of Single-Walled Carbon Nanonet Flexible Strain SensorsDiunggah olehAzmul Fadhli
- Spanish Word Frequency ListDiunggah olehlizan13
- SYSMAC-XR Library Brochure en 201511 P102-E1-01 (1)Diunggah olehBianca Cerveira
- 860 Dspi Work Orders AppnoteDiunggah olehtrilithicweb
- Programming Without Coding Technology (PWCT) AbstractDiunggah olehMahmoud Samir Fayed
- SRS Template for Java Chat AppDiunggah olehshubhambindal
- Crypto Key ManagementDiunggah olehFachrizal Rizky Ferdiansyah
- Sensors PPTDiunggah olehanimesharma
- Jean-Marc Vesin and Touradj Ebrahimi- Trends in Brain Computer InterfacesDiunggah olehNoScript
- Nvd 16 Cv 01918 Jad Njk Document 55Diunggah olehActuaLitté
- ASCII Code TableDiunggah olehritagmrocha
- Data Warehousing ConceptsDiunggah olehSoumyajit Sarcar
- Yamaha G50 ManualDiunggah olehPsÿchø Six
- Rubric COE117LDiunggah olehEugene Martin
- i o Terminal Expansion Kinetix 300Diunggah olehMarco Antonio Alvarez Loza
- Distributed Information Systems Exam December 2006 - UK University BSc Final YearDiunggah olehTDiscover
- SFDC CRM Ready ReckonerDiunggah olehNirjhar Jajodia