International Journal of Heat and Mass Transfer 103 (2016) 715–724

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer

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

Design of microchannel heat sink with wavy channel and its

time-efficient optimization with combined RSM and FVM methods
Jiandong Zhou a, M. Hatami a,b, Dongxing Song a, Dengwei Jing a,⇑
a
International Research Center for Renewable Energy, State Key Laboratory of Multiphase Flow in Power Engineering, Xi’an Jiaotong University, Xi’an 710049, China
b
Department of Mechanical Engineering, Esfarayen University of Technology, Esfarayen, North Khorasan, Iran

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

Article history: In this study, a sinusoidal wavy structure of microchannel heat sink intended for active cooling of com-
Received 28 June 2016 pact electronic devices such as insulated-gate bipolar transistor (IGBT) has been designed. By combining
Received in revised form 22 July 2016 with the finite volume method (FVM), the geometric parameter of the wavy wall, as the key factor to
Accepted 27 July 2016
improve the heat transfer efficiency, has been optimized by a novel response surface methodology
(RSM), which has found to be time-efficient and accurate. Furthermore, we use a comprehensive heat
transfer index b to study whether the heat transfer enhancement outweighs the increased pressure drop.
Keywords:
After investigating the Re, h, DP, f, b, it is concluded that the best case is occurred when wave amplitude
Heat transfer enhancement
MCHS
value is 40 and wavelength is 100. For our optimized wavy channel, the heat transfer can be enhanced by
Wavy channel a maximum of 2.8 times compared to regular straight channel. CFD simulation demonstrates that under
RSM such case, the existence and disturbance of vortex can lead to thinning of boundary layer and hence more
effective heat transfer. Our results should have practical value for designing of compact heat exchanger
and the proposed optimization method is supposed to have wide application for the time-efficient opti-
mization of heat transfer through irregular configurations.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction Reynolds number and the media type, etc., were present in this
review. Tuckerman and Pease [5] are among the pioneering
Miniaturization and compactness of heat sink for electronic researchers in applying the concept of microchannel heat sink with
devices are crucial for devices such as PC, mobile phone and even forced convection. They designed a new compact, water-cooled
artificial satellite. In order to ensure the working reliability of those integral heat sink with microchannel and found that it showed
electronic instruments with higher power density, even small heat an excellent efficiency of heat exchange. Generally, microchannel
emission problem must be solved. Other than electronic equip- heat exchangers are composed by a lot of parallel microchannels
ment, large machinery such as locomotive traction system also with a hydraulic diameter ranging from 10 lm to 1000 lm. In such
faces difficulty in heat dissipation. For this system, insulated-gate design, the heat exchangers are usually bonded with the heat-
bipolar transistor (IGBT) is its key component unit, which gener- dispersing surfaces by heat conduction silica gel. The coolant flows
ates a huge amount of heat due to high switching frequency and through the microchannels and carries out the needless heat for
flow through of huge current. To maintain the temperature of IGBT generating uniform temperature distributions. Recently, the heat
in safety zone is therefore crucial for its steady and safe operation. transfer and fluid flow processes in rectangular microchannel has
For most of such devices, active cooling by air or water are the been studied by Lee and Garimella [6]. They pointed that conven-
most commonly methods for dissipating additionally generated tional analysis approach can be employed in predicting heat trans-
heat. Among numerous types of heat exchangers, microchannel fer behavior in microchannels. However, the entrance and
heat exchangers have received great attention in recent years as boundary conditions imposed in the experiment need to be care-
noted in several reviews [1–3]. For instance, recent development fully matched in the predictive approaches. Tazraei et al. [7,8]
of microchannel heat sink (MCHS) was summarized by Smakulski investigated the features of fast transient non-Newtonian fluids
et al. [4]. An elaborate comparison on the technology, operating flow and found that the non-Newtonian behavior of fluids has sig-
fluids, flow character, heat transfer coefficient, pressure drop, nificant influence on the velocity and shear stress profiles and also
on the magnitude of pressure heat and wall shear stress. On the
⇑ Corresponding author. other hand, some researchers designed MCHS with different struc-
E-mail address: dwjing@mail.xjtu.edu.cn (D. Jing). tures in order to enhance heat transfer performance. Cui et al. [9]

http://dx.doi.org/10.1016/j.ijheatmasstransfer.2016.07.100
0017-9310/Ó 2016 Elsevier Ltd. All rights reserved.
716 J. Zhou et al. / International Journal of Heat and Mass Transfer 103 (2016) 715–724

investigated the characteristics of fluid flow and heat transfer in a will calculate the Nu, DP, sw, f values for the convections of wavy
novel microchannel heat sink with triangular cavities and rectan- walls with different amplitudes and wavelengths under fixed val-
gular ribs (TC-RR). The results showed that TC-RR microchannel ues of Re. Then the relationships between convective heat transfer
significantly enhanced heat transfer efficiency due to the interrup- coefficient and the wave amplitude, wavelength, are analyzed.
tion and redevelopment of thermal boundary layer. Xia et al. [10] Pressure drop of wavy microchannel of against various wave
studied the effects of geometric parameters on water flow and heat amplitudes and wavelengths are also analyzed. In the RSM opti-
transfer characteristics in microchannel heat sink with triangular mization, we will use b as the reference value to find the optimal
reentrant cavities by numerical simulation method. By comparing geometry. By calculating the b under different Re, the optimal
four different geometries of the triangular reentrant cavity, they geometry can be confirmed.
found that the vortices in the triangular reentrant cavities could The novelty of our present work compared to previous studied
lead to chaotic advection which dramatically enhanced the convec- are as follows. Firstly, a sinusoidal wavy structure of microchannel
tive fluid mixing. heat sink for potential cooling of electronic devices has been
Kim et al. [11] studied the heat transfer enhancement by cross- designed. Furthermore, we use a comprehensive heat transfer
cut induced flow control in a wavy fin heat exchanger. Parametric index b to study whether the heat transfer enhancement out-
study was conducted to find the optimal position and length of the weighs the increased pressure drop. Finally, RSM was combined
cross-cut. The results showed that heat transfer performance of with the finite volume method (FVM) to find the optimized perfor-
optimized cross-cut applied plain fin could be enhanced by mance for heat transfer. However, in the previous studies, only sin-
23.81% compared to a typical wavy fin. Sabaghan et al. [12] simu- gle indexes such as Nusselt number or Pressure drop, were used for
lated the heat exchange rates in a rectangular microchannel con- evaluation of the heat transfer performance.
sisting of six longitudinal vortex generators (LVG) based on the
two-phase approach. Besides, they also used the nanofluids as
the coolant. It was demonstrated that the maximum normalized 2. Computational model
efficiency of the LVG-enhanced microchannel, compared to the
plain channel, could be enhanced by around 14%. There are also 2.1. Response surface method (RSM)
many research works about the heat transfer in the microchannel
[13–16] carried out by both numerical and experimental methods. Response surface methodology (RSM) is a collection of mathe-
Considering the more and more complicated heat transfer phe- matical and statistical techniques, which was applied to establish
nomena and/or increased power-density for electron devices in a mathematical model between independent variable and depen-
recent years, most of the already existing heat transfer strategies dent variable, and find the effect of parameters affecting a response
for MCHS become less effective and more efficient MCHSs are in a process. The time taken to solve the targeted problem can be
apparently highly desired. Therefore, it is still essential to focus reduced and a lot of computation resources are saved by using
on the heat transfer characteristics in microchannel and try to find RSM. Generally, the structure of the relationship between the
optimal structures for MCHS that is suitable for certain dependent variable and the independent variables is unknown.
applications. The first step in RSM is to find a suitable approximation close to
Kim et al. [11] have performed simulation using non dimen- the true relationship. The most common forms are low-order poly-
sional governing equations for a steady laminar flow. The paramet- nomials (first or second order). A second-order model can signifi-
ric study was conducted to find the optimum position and length cantly improve the optimization process when a first-order
of the cross-cut wavy fin. Tazraei et al. [17] put forward a novel model suffers from some lack of fit due to the interaction between
method, named as the QL method, for solving the radiative transfer variables and surface curvatures. A general second-order model is
equation. They also extended the QL method to 3D problems in defined as [18]
non-grey media and combined it with FVM, ClAM scheme and
X
n X
n X
n X
n 
SLW model to solve the radiative transfer equation. The predictions y ¼ a0 þ ai xi þ aii x2i þ aij xi xj i<j ð1Þ
by this method were found to be accurate and computationally low i¼1 i¼1 i¼1 j¼1
cost, therefore the results can be used in general code with full reli-
ability. Gorji et al. [18] have investigated the optimal microchannel where xi and xj are the design variables, a the tuning parameter and
heat sink geometry by response surface methodology (RSM). They n the number of parameters which is 2 in this case. A Box-Wilson
found that RSM can significantly reduce the calculation time and Central Composite Design, also called as a ‘‘central composite
cost. Jafaryar et al. [19] also applied the response surface method- design” or CCD commonly, is one of options in RSM which helps
ology (RSM) based on central composite design (CCD) to obtain an the user in defining the factor levels.
optimization design for the asymmetric blades geometry of a ver- An imbedded factorial or fractional factorial design is contained
tical axis wind turbine (VAWT). In recent studies, Ganji et al. have in CCD with center points that are augmented with a group of ‘star
investigated the finned-tube heat exchangers for diesel exhaust points’ which allow an estimation of the curvature. If the distance
waste heat recovery using CFD and CCD techniques [22]. We have from the center of the design space to a factorial point is ±1 unit for
also optimized a circular-wavy cavity filled by nanofluid under the each factor, the distance from the center of the design space to a
natural convection heat transfer condition by this efficient tech- star point is ±a for |a| > 1 [20–22]. In the RSM method with CCD,
nique [23]. optimization is based on a parameter called ‘desirability’. Desir-
In our study, taking IGBT as a typical example, we will demon- ability is an objective function ranging from 0.0 outside of the lim-
strate the advantage of RSM for the optimization the wavy its to 1.0 which corresponds to the goal [23]. The numerical
microchannel heat sink for heat dissipation of such electronic optimization finds a point that maximizes the desirability function.
devices. Specifically, in our work a new 3D heat transfer of wavy The characteristics of the goal may be altered by adjusting the
MCHS has been designed and compared with rectangular MCHS. weight or importance. For several responses and factors, all goals
In the application of wavy-wall heat transfer systems, an impor- get combined into one desirability function. In this study, one
tant problem is to determine the optimum geometry. In our study, response is defined as comprehensive heat transfer index b. The
we will employ the RSM method to optimize the geometry of the goal of optimization is to find a set of conditions that meet all
microchannel in a time-efficient way. There are two steps to inves- the goals, instead of to get a desirability value of 1.0. Desirability
tigate the performance of wavy channel heat transfer. Firstly, we reflects the preferred ranges for each response (di). The simultane-
 J. Zhou et al. / International Journal of Heat and Mass Transfer 103 (2016) 715–724 717

ous objective function is a geometric mean of all transformed 2.4. Assumption and simplification
responses:
In order to simplify the analysis in this study, several assump-
!1n tions have been made regarding the operating conditions of a wavy
1 Y
n
D ¼ ðd1  d2  :::  dn Þn ¼ di ð2Þ channel MCHS.
i¼1

1. The fluid, which is water in our study, is considered as incom-

where n is the number of responses in the measurement (in this pressible laminar flow across a channel and maintains the con-
case, n = 1). If any of the responses or factors falls outside their dition of single phase,
desirability range, the overall function becomes zero. For simultane- 2. We will consider both thermally and hydraulically fully devel-
ous optimization, each response must have a low and high value oped flow.
assigned to each goal. On the CCD worksheet, five option can be 3. The thermophysical properties of the solid are constant.
chosen in the ‘‘Goal” field for responses: ‘‘none”, ‘‘maximum”, ‘‘min- 4. The surface of the MCHS is well insulated.
imum”, ‘‘target”, or ‘‘in range”. In this study, we choose the ‘‘maxi-
mum” as the goal parameter for comprehensive heat transfer 3. Mathematical modeling
efficiency, as shown below:
Based on the assumptions above, the governing equations for
di ¼ 0; Y i 6 Lowi this model are as follows.
h iwti
Y i Lowi
di ¼ High ; Lowi < Y i < Highi ð3Þ The continuity equation:
i Lowi

di ¼ 1; Y i P Highi @u @ v @w
þ þ ¼0 ð5Þ
@x @y @z
where Yi is the ith response value and wt is the weight of that where u, v, and w are defined as the velocity components in the x, y,
response. Weight adds emphasis to the goal. In this paper, the and z directions, respectively.
weight has a negligible effect on the final results because just one Momentum equations for the coolant:
response is defined.  
þ v @u ¼  @p
2 2 2
qf u @u
@x @y
þ w @u
@z @x
þ lf ð@@x2u þ @@yu2 þ @@z2uÞ
 
qf u @@xv þ v @@yv þ w @@zv ¼  @p þ lf ð@@xv2 þ @@yv2 þ @@zv2 Þ
2 2 2
2.2. Straight microchannel heat sink @y
ð6Þ
 
þ v @w ¼  @p
2 2 2
For comparison, the geometric dimensions of straight qf u @w
@x @y
þ w @w@z @z
þ lf ð@@xw2 þ @@yw2 þ @@zw2 Þ
microchannel are taken from optimally designed one in literature
[24] as show in Table 1. Its bottom size is 28  100 mm and the where qf and lf are the density and dynamic viscosity of the cool-
heat flux of 100w=cm2 is received from the bottom surface ant, respectively, and p is the coolant pressure.
(Fig. 1a). In order to save the computational resources, only one Energy equation for the coolant:
of the straight microchannels is taken as the computational   !
@T f @T f @T f @2T f @2T f @2T f
domain as shown in Fig. 1b. In this model, both the two sides of qf cpf u þv þw ¼ jf þ þ 2 ð7Þ
@x @y @z @x2 @y2 @z
the microchannel contain a half-wall. The geometry parameters
of each microchannel are as follows: wall thickness Energy equation for solid region:
W r =2 ¼ 28 lm; channel width W c ¼ 85 lm; channel height
Hc ¼ 700 lm and cover plate thickness Hr ¼ 100 lm.The rectangu- !
@2T s @2T s @2T s
lar straight microchannel heat sink is used as the benchmark to 0 ¼ js þ 2 þ 2 ð8Þ
evaluate the heat transfer performances of wavy channel of various @x2 @y @z
geometries.
Generally, the thermal properties of water, such as density, vis-
cosity, heat capacity and thermal conductivity, can vary with the
2.3. Wavy microchannel heat sink temperature change. The expressions of temperature dependent
thermophysical properties are given [25] by,
In this study, single layer wavy microchannel heat sink is ini-  
tially considered to be an improved design compared to rectangu- T þ 15:9414
qðTÞ ¼ 1000 1  ðT  276:9863Þ2
lar straight microchannel, as shown in Fig. 1c. All of the dimensions 508929:2ðT  204:87037Þ
are same as the rectangular straight microchannel other than that ð9Þ
straight microchannel becomes wavy. The wave function along
x-axis can be determined by,
 8:9
T
lðTÞ ¼ 1:005  103 EXP½4700ðT 1  2931 Þ ð10Þ
 293
x
y ¼ A sin 2p  ð4Þ
k
kf ðTÞ ¼ 1:579 þ 0:01544T  3:515  105 T 2 þ 2:678  108 T 3
where A and k are defined as amplitude and wavelength, respec- ð11Þ
tively. Also, x is defined as the flow direction.
The velocity inlet is defined with uniform temperature of 298 K
while velocity inlet is specified as 0.6 m/s, 0.8 m/s, 1 m/s, 1.2 m/s,
Table 1 1.4 m/s, which corresponds to Reynolds number ranges from 99
Optimal geometric dimension for straight microchannel heat sink. to 232. The pressure outlet is adopted with atmosphere pressure.
Lx (mm) Ly (lm) Lz (mm) W c (lm) W r (lm) Hc (lm) d (lm) q(W=cm2 ) Constant heat flux 100 W/cm2 is applied at the bottom of MCHS.
Symmetrical boundary conditions are assumed on the left and right
10 900 10 85 56 700 100 100
sides. The boundary conditions are listed as below.
718 J. Zhou et al. / International Journal of Heat and Mass Transfer 103 (2016) 715–724

Fig. 1. Schematics of microchannel heat sinks for (a) heat sink, (b) straight microchannel and (c) wavy microchannel. The meanings of all the symbols can be found in Table 1.

Inlet: The resistance coefficient is,

u ¼ uin ; v ¼ 0; w ¼ 0; T ¼ T in ð12Þ DP A
f ¼1 ð19Þ
Outlet: 2
qu2 L
p ¼ pout ð13Þ In order to investigate the comprehensive effects of enhanced
heat transfer, the comprehensive heat transfer coefficient b is given
Coolant-solid interface:
[26] by
u ¼ 0; v ¼ 0; w ¼ 0; T ¼ T in ð14Þ
1=2
b ¼ j=f ð20Þ
Bottom wall of the heat sink:
Here b is used to judge whether the increment of heat exchange
@T s
qw ¼ ks ð15Þ capability is greater than the increment of flowing resistance at the
@n
same pressure drop.
Other solid walls and symmetric boundaries: To evaluate the heat transfer characteristics and fluid flow
within the wavy channel, finite volume method (FVM) is imple-
@T s mented with the assistance of a CFD commercial package (ANSYS
ks ¼0 ð16Þ
@n Fluent). The SIMPLEC (semi implicit method for pressure linked
equation consisted) algorithm is used to model pressure–velocity
Reynolds number is defined as coupling. The prescribed convergence criterions of velocities and
quav Dh continuity are 104 , and for energy, it is set to be 106 .
Re ¼ ð17Þ
l
where Dh is the hydraulic diameter defined as 4. Results and discussions
Dh ¼ 2Hc W ch =ðHch þ W ch Þ.
The dimensionless surface heat transfer coefficient, i.e., j factor 4.1. Grid independency test and accuracy of calculation validation
is defined as,
Based on ANSYS Fluent, the 3D conjugate heat transfer problem
Nu
j¼ 1=3
ð18Þ is numerically solved. As shown in Fig. 2, we use the hexagonal
Re  Pr structured grid to deal with geometry, a grid of 1000  104  20
 J. Zhou et al. / International Journal of Heat and Mass Transfer 103 (2016) 715–724 719

Fig. 2. The computational grids (a) for the calculated domain in the wavy microchannel and (b) the enlarged view near the boundary wall.

was used for the calculation of the flow channel, as shown in Fig. 2 1.00E+08
(a). The dense grid cells with successive ratio of 1.2 distribute near
the boundaries, as shown in Fig. 2(b) for the enlarged view near the 1.00E+07
Experimental data
boundary wall. To identify the least mesh-dependence, four mesh
Pressure Gradient, Pa/m
1.00E+06 Simulation data
numbers of structured grids are considered for grid independence
test. The grid independence test is undertaken for the Reynolds
1.00E+05
number Re = 150 considered in this study, with pure water as cool-
ant inside the microchannel. Fig. 3 shows the results of heat trans-
1.00E+04
fer coefficient h calculated over different grid sizes. Since the
relative differences of calculated h are less than 0.1% over cell num- 1.00E+03
ber 4 and 5.5 million, respectively, 5 million cells is deemed appro-
priate for all the computational simulations carried out in this 1.00E+02
study.
In order to validate the accuracy of our model, the experimental 1.00E+01
10 100 1000 10000
data of DP from the literature [27] were compared with the simu-
lation of our model. The results are depicted in Fig. 4. As can be Hydraulic diameter of circular tube, μm
clearly observed, there is a good agreement between our simulated
Fig. 4. Validation of the model accuracy with the data available in the literature, S.
results and the tendency of the experimental results. G. Kandlikar et al. [26].

4.2. The results of CFD simulation CCD is applied for three levels for each of the two factors, A and
k, as shown in Table 2. The wave amplitude A is changed from
As described, a wavy channel is studied under the laminar flow 10 lm to 40 lm and wavelength k from 100 lm to 1000 lm. By
convective heat transfer. The wall function is y ¼ A sin 2p  xk , CCD analysis, 9 different geometries with various amplitudes and
where A and k are the wave amplitude and wavelength, respec- wavelength have been studied, as show in Table 2.
tively. Sakanova et al. [28] considered constant numbers of 25, Fig. 5 shows the streamlines for nine proposed cases at flow rate
50, 75 for A, and 250, 550 for k, respectively. In this investigation, u = 1 m/s by the FLUENT. From the streamline contours, it can be
we want to find optimal constants for these numbers which have observed that the maximum velocity in x direction increases with
a high comprehensive heat transfer coefficient (b). For this aim, the increment of wave amplitude at the same wavelength. For
instance, the maximum X velocity is shifted from 1.5 m/s to
2.8 m/s when A is adjusted from 10 lm to 40 lm at constant wave-
21000 water (havg) length of k = 100 lm. In addition, obvious vortex appears in the
cases of A = 25, k = 100 and A = 40, k = 100. It means that the vortex
20000 is more possibly occurred under the larger A and the smaller k. The

19000
Table 2
h W/m K
2

Variation of b in different CCD proposed cases when Re = 150.

18000
Case number A k b
17000 1 10 100 0.046034
2 10 550 0.0422877
16000 3 10 1000 0.0428088
4 25 100 0.0414412
15000 5 25 550 0.0477744
1 2 3 4 5 6 6 25 1000 0.0476693
Number of Cells (million) 7 40 100 0.0556904
8 40 550 0.0554782
9 40 1000 0.0417381
Fig. 3. Grid independency examination for the rectangular microchannel.
720 J. Zhou et al. / International Journal of Heat and Mass Transfer 103 (2016) 715–724

X Velocity X Velocity X Velocity

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4

Case 1 A=10 μm Case 2 A=10 μm Case 3 A=10 μm

λ=100 μm λ=550 μm λ=1000 μm

X Velocity X Velocity X Velocity

0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2 2.2 2.4
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4

Case 5 A=25 μm Case 6 A=25 μm

Case 4 A=25 μm
λ=550 μm λ=1000 μm
λ=100 μm

X Velocity X Velocity X Velocity

-0.2 0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2 2.2 2.4 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4

Case 8 A=40 μm Case 9 A=40 μm

Case 7 A=40 μm λ=1000 μm
λ=100 μm λ=550 μm

Fig. 5. Velocity vector streamlines along X direction for the typical section of the 9 different geometries of microchannel obtained by RSM.

obtained b values over 9 proposed cases are also given in Table 2. It rectangular channel for low Reynolds number and it becomes 2.81
can be found that the highest and secondly highest heat transfer times with high Reynolds number. From Fig. 6(b), it can be
can be achieved over case 7 and case 8, where vortex can be observed that the relation between pressure drop and Re number.
observed in case 7. The reasons of higher b in case 7 as follows, With the increase of Re number, pressure drop increases for all
on the one hand, the higher velocity can accelerate the flow next geometries. The variation trends for all the curves are in agreement
to wavy peaks, the hydrodynamic and thermal boundary layers with the fluid streamlines in Fig. 4. It is apparent that wavy channel
are therefore getting thinner. On the other hand, the vortex can has higher pressure drop than rectangular channel. Therefore, it is
cause the disturbance of boundary layer, and make it much thin- necessary to investigate the temperature contour and the relation-
ner. Therefore, there is an increment in heat transfer coefficient ship between h, DP and value of wave amplitude, wavelength
for cases such as case 7. further.
Fig. 6 shows the comparison of wavy-enhanced microchannel Fig. 7 presents the temperature contours (K) along X direction
with straight microchannel for convective heat transfer coefficient for the typical section of the 3 different geometries of microchan-
(a) and pressure drop (b), respectively. Fig. 6(a) shows that the nel obtained by RSM, which correspond to the normal (case 1),
relation of convective heat transfer coefficient and Reynolds num- worst (case 4) and optimal (case 7) heat transfer cases, respec-
ber. As the Re numbers increases, the convective heat transfer coef- tively. For all the three cases, one can see that when fluid flows
ficient increases. Also, it is obvious that the convective heat through the channel, both the temperature and the temperature
transfer coefficient of wavy channels can be significantly improved gradients are high near the wall and low in the central region of
compare to straight channel. Moreover, it can be seen that for the channel. It can also be found that for case 7, the average tempera-
cases of A = 40, k = 100; A = 40, k = 550 and A = 25, k = 100, they ture is high and also the temperature gradient is quite low. It indi-
have higher heat transfer coefficients compared to other cases. cates that geometry of case 7 should be the preferred case in term
The results are listed in the size order of h: case 7, case 8, case 4, of heat transfer capacity of the microchannel.
case 5, case 6, case 1, case 3, case 2, case 9. For example, in the case Fig. 8 shows the heat transfer coefficients and pressure drop
of A = 40 and k = 100, the h value is about 2.54 times compared to against Re numbers over cases of various wave amplitudes and
 J. Zhou et al. / International Journal of Heat and Mass Transfer 103 (2016) 715–724 721

(a) (b)
case 1 case 2 case 3 10
6 case 1 case 2 case 3
100000 case 4 case 5 case 6 case 4 case 5 case 6
90000 case 7 case 8 case 9
80000 case 7 case 8 case 9
70000 rectangular channel rectangular channel
havg W/m K60000
2

Δ Pa
5
50000 10
40000

30000
4
10
20000

100 125 150 175 200 225 100 125 150 175 200 225
Re Re

Fig. 6. Comparison of wavy-enhanced microchannel and straight microchannel for convective heat transfer coefficient (a) and pressure drop (b), respectively.

Temperature Temperature Temperature

Case 1 A=10 μm Case 4 A=25 μm Case 7 A=40 μm

λ=100 μm λ=100 μm λ=100 μm

Fig. 7. Temperature contours (K) along X direction corresponding to normal (case 1), worst (case 4) and optimal (case 7) heat transfer cases, respectively.

wavelengths. It can be observed that with the same A and under contours (2D and 3D) are presented in Fig. 10, which is obtained
different k, the lower the k of wavy channel, the higher the h and from DOE software based on CCD analysis. Considering the effects
DP will be. However, with the same k and under different A of wavy of main factors and also the interactions between the two factors of
curved channels, the higher the A of wavy channel, the higher the h A and k, Eq. (1) takes the form,
and DP become. For example, comparing the Figs. 8a–c(1), it has
the largest h when A = 40 under the same range of wavelength. ðbÞ ¼ 0:038936 þ 8:01987  105 ðAÞ þ 2:00901  105 ðBÞ
Also, when A = 40, increase of h is much faster than the cases for
A = 10 and A = 25 when k changes from 100 lm to 1000 lm. Corre-  3:973  107 ðABÞ þ 7:60548  106 ðA2 Þ  1:29208  108 ðB2 Þ
spondingly, DP also shows similar characteristic in Fig. 8a(2)–c(2). ð21Þ
What’s more, it should be noted that the performance of DP shows
From the 3D graphs for above formulas as shown in Fig. 8, it can
less difference between k = 550 lm and k = 1000 lm. Figs. 8d–f
be found that the maximum b occurs in the high A (wave ampli-
show the effect of wavelength on heat transfer and fluid flow.
tude) and low B (wave length).
We can see that the same amplitude range with shorter wave-
As described in Section 2.1 for CCD technique, optimization is
length of wavy channel lead to a higher h and DP. Also, it can be
based on the desirability with high desirability indicating the bet-
observed that the higher the wavelength, the lower the increasing
ter operating condition. Desirability contour for the optimum point
rate of DP will be, in the range between A = 10 to A = 40.
can also be found in Fig. 10. One can see that the desirability for the
optimal operating point is about 0.927 which corresponds to a
4.3. Optimal design of MCHS by RSM good optimization result. The desirability of the selected confirms
that the best case (maximum desirability) occurs when A = 40 and
In order to comprehensively evaluate the performances of heat k = 100. In this case, could be as high as 0.0525327. It is therefore
exchanger, b is defined to judge whether the increment of heat chosen as the best case from the comprehensive heat transfer
exchange capability is greater than increment of flowing resistance property. To further confirm the accuracy of maximum desirability,
under the same pressure drop, considering the complicated rela- all cases of b are investigated in Re numbers ranging from 90 to
tionships between h and DP over various geometrical parameters 210. The results are shown in the Fig. 11. It can be seen that best
of wave channels. Table. 2 show the results of RSM analysis of case for comprehensive heat transfer coefficient b, i.e., A = 40 and
the 9 cases. k = 100, is obviously higher than other cases almost over all Re
Based on the results of RSM analysis from Figs. 6, 8 and 9, a ranges. Also b is found to be decreased with the increase in Re
polynomial model with quadratic order is applied for the response for all cases. It is supposed that with the increment of flow rate,
(b) and the equation of the b versus A and k can be found and its the increase of resistance is quick than heat transfer capacity.
722 J. Zhou et al. / International Journal of Heat and Mass Transfer 103 (2016) 715–724

a(1) 40000 wavy A=10μm λ=100μm a(2) 22000 wavy A=10μm λ=100μm
38000 wavy A=10μm λ=550μm 20000 wavy A=10μm λ=550μm
wavy A=10μm λ=1000μm

h avg W/m K
36000 wavy A=10μm λ=1000μm 18000

2
34000 16000

ΔP/Pa
14000
32000
12000
30000
10000
28000
8000
26000 6000
24000 4000

75 100 125 150 175 200 225 100 125 150 175 200 225
Re Re

b(1) 44000 wavy A=25μm λ=100μm b(2) 120000 wavy A=25μm λ=100μm
wavy A=25μm λ=550μm wavy A=25μm λ=550μm
42000 wavy A=25μm λ=1000μm 100000 wavy A=25μm λ=1000μm
havg W/m K

40000
2

ΔP/Pa
38000 80000
36000
60000
34000
32000 40000
30000
28000 20000
26000
0
100 125 150 175 200 225 100 125 150 175 200 225
Re Re

wavy A=40μm λ=100μm

c(1) 70000 c(2) 250000 wavy A=40μm λ=100μm
wavy A=40μm λ=550μm
65000 wavy A=40μm λ=550μm
wavy A=40μm λ=1000μm
60000 200000 wavy A=40μm λ=1000μm
havg W/m K

55000
2

ΔP/Pa

50000 150000
45000
40000 100000
35000
30000 50000
25000
20000 0
100 125 150 175 200 225 100 125 150 175 200 225
Re Re

wavy A=10μm λ=100μm

d(1) 65000 wavy A=25μm λ=100μm
d(2) 250000 wavy A=10μm λ=100μm
wavy A=40μm λ=100μm wavy A=25μm λ=100μm
60000
200000 wavy A=40μm λ=100μm
havg W/m K

55000
2

ΔP/Pa

50000 150000
45000
100000
40000
35000
50000
30000
25000 0
100 125 150 175 200 225 100 125 150 175 200 225
Re Re
e(1) 48000 wavy A=10μm λ=550μm e(2) 20000 wavy A=10μm λ=550μm
46000 wavy A=25μm λ=550μm 18000 wavy A=25μm λ=550μm
44000 wavy A=40μm λ=550μm
16000 wavy A=40μm λ=550μm
havg W/m K

42000
40000 14000
2

ΔP/Pa

38000 12000
36000
34000 10000
32000 8000
30000
28000 6000
26000 4000
24000
100 125 150 175 200 225 100 125 150 175 200 225
Re Re

f(1) 34000 wavy A=10μm λ=1000μm wavy A=10μm λ=1000μm

wavy A=25μm λ=1000μm f(2) 14000 wavy A=25μm λ=1000μm
wavy A=40μm λ=1000μm wavy A=40μm λ=1000μm
32000 12000
havg W/m K
2

ΔP/Pa

30000 10000

28000 8000

26000 6000

4000
24000

100 125 150 175 200 225 100 125 150 175 200 225
Re Re

Fig. 8. Heat transfer coefficient and pressure drop at the same amplitude under different wavelength/ at the same wavelength under different amplitude.
 J. Zhou et al. / International Journal of Heat and Mass Transfer 103 (2016) 715–724 723

Fig. 9. Effect of A and B parameters on b (a) 2D and (b) 3D.

5. Conclusion

In this paper, a new type of MCHS with wavy microchannel

structure is put forward. The heat transfer in heat sink is investi-
gated by a numerical method. The effects of wavy amplitude,
wavelength, Reynolds number and b are studied. We employed
water as the coolant in our design. The results indicate that the
heat transfer can be enhanced by a maximum of 2.8 times com-
pared to regular straight channel. It is also found that the higher
amplitude and shorter wavelength for wavy channel can lead to
a more convective heat transfer coefficient.
In order to obtain the maximum comprehensive heat transfer
index b, instead of traditional time-consuming methods, we use
RSM to find the optimal geometry for the wavy wall in a time-
efficient and accurate way. By combining the RSM and FVM, it is
concluded that the best case is when amplitude value is 40 and
wavelength is 100. The proposed method, FVM combined with
RSM, is supposed to have wide application for the time-efficient
optimization of heat transfer through irregular geometry. More-
over, the results of this study are also considered to be of practical
Fig. 10. Desirability contour for the optimum point.
value for the optimal designing of active cooling of electronic
devices, considering that a small improvement of heat transfer
coefficient for these devices is expected to result in great savings
in their long-term operation.

rectangular channel wavy A=10μm λ=100μm

0.08 wavy A=10μm λ=550μm wavy A=10μm λ=1000μm
wavy A=25μm λ=100μm wavy A=25μm λ=550μm Acknowledgments
wavy A=25μm λ=1000μm wavy A=40μm λ=100μm
wavy A=40μm λ=550μm wavy A=40μm λ=1000μm
The authors gratefully acknowledge the financial supports of
the National Natural Science Foundation of China (Nos.
0.06 51422604, 21276206) and National 863 Program of China (No.
2013AA050402). This work was also supported by the China Fun-
β

damental Research Funds for the Central Universities.

0.04
References

[1] G. Goss Jr, J.L.G. Oliveira, J.C. Passos, Pressure drop during condensation of
R-134a inside parallel microchannels, Int. J. Refrig. 56 (2015) 114–125.
[2] X.W. Yin, W. Wang, V. Patnaik, Evaluation of microchannel condenser
0.02 characteristics by numerical simulation, Int. J. Refrigeration 54 (2015) 126–
100 125 150 175 200 225 141.
Re [3] J. Zhang, Y.H. Diao, Y.H. Zhao, Y.N. Zhang, An experimental study of the
characteristics of fluid flow and heat transfer in the multiport microchannel
Fig. 11. The comprehensive heat transfer index in 9 cases. flat tube, Appl. Therm. Eng. 65 (2014) 209–218.
724 J. Zhou et al. / International Journal of Heat and Mass Transfer 103 (2016) 715–724

[4] P. Smakulaki, S. Pietrowicz, A review of the capabilities of high heat flux [17] P. Tazraei, The extension of the QL method to solve the radiative heat transfer
removal by porous materials, microchannels and spray cooling techniques, problem in a 3D square enclosure containing non-grey gas, Am. J. Mech. Eng. 4
Appl. Therm. Eng. (2016). (2) (2016) 71–81.
[5] D.B. Tuckerman, R.F.W. Pease, High-performance heat sinking for VLSI, IEEE [18] M. Rahimi-Gorji, O. Pourmehran, M. Hatami, D.D. Ganji, Statistical
Electron Device Lett. 2 (1981) 126–129. optimization of microchannel heat sink (MCHS) geometry cooled by
[6] P.S. Lee, S.V. Garimella, D. Liu, Investigation of heat transfer in rectangular different nanofluids using RSM analysis, Eur. Phys. J. Plus 130 (2015) 1–21.
microchannels, Int. J. Heat Mass Transfer 48 (2005) 1688–1704. [19] M. Jafaryar, R. Kamrani, M. Gorji-Bandpy, M. Hatami, D.D. Ganji, Numerical
[7] P. Tazraei, A. Riasi, Quasi-two-dimensional numerical analysis of fast transient optimization of the asymmetric blades mounted on a vertical axis cross-flow
flows considering non-Newtonian effects, J. Fluids Eng. 138 (1) (2016) 011203. wind turbine, Int. Commun. Heat Mass 70 (2015) 93–104.
[8] P. Tazraei, A. Riasi, B. Takabi, The influence of the non-Newtonian properties of [20] W. Kalaoka, S. Witayangkurn, Numerical simulation of natural convection in a
blood on blood hammer through the posterior cerebral artery, Math. Biosci. partially cooled square enclosure filled with porous medium, Int. J. Pure Appl.
264 (2015) 119–127. Math. 96 (2014) 507–522.
[9] G. Xia, Y. Zhai, Z. Cui, Numerical investigation of thermal enhancement in a [21] G.A. Florides, P. Christodoulides, P. Pouloupatis, An analysis of heat flow
micro heat sink with fan-shaped reentrant cavities and internal ribs, Appl. through aborehole heat exchanger validated model, Appl. Energy 92 (2012)
Therm. Eng. 58 (2013) 52–60. 523–533.
[10] G. Xia, L. Chai, H. Wang, Optimum thermal design of microchannel heat sink [22] M. Hatami, M. Jafaryar, D.D. Ganji, M. Gorji-Bandpy, Optimization of finned-
with triangular reentrant cavities, Appl. Therm. Eng. 31 (2011) 1208–1219. tube heat exchangers for diesel exhaust waste heat recovery using CFD and
[11] G.W. Kim, H.M. Lim, G.H. Rhee, Numerical studies of heat transfer CCD techniques, Int. Commun. Heat Mass 57 (2014) 254–263.
enhancement by cross-cut flow control in wavy fin heat exchangers, Int. J. [23] M. Hatami, D. Song, D. Jing, Optimization of a circular-wavy cavity filled by
Heat Mass Transfer 96 (2016) 110–117. nanofluid under the natural convection heat transfer condition, Int. J. Heat
[12] A. Sabaghan, M. Edalatpour, M.C. Moghadam, Nanofluid flow and heat transfer Mass Transfer 98 (2016).
in a microchannel with longitudinal vortex generators: two-phase numerical [24] Z.H. Wang, X.D. Wang, W.M. Yan, Multi-parameters optimization for
simulation, Appl. Therm. Eng. 100 (2016) 179–189. microchannel heat sink using inverse problem method, Int. J. Heat Mass
[13] L. Chai, G.D. Xia, H.S. Wang, Parametric study on thermal and hydraulic Transfer 54 (2011) 2811–2819.
characteristics of laminar flow in microchannel heat sink with fan-shaped ribs [25] D. Susa, M. Lehtonen, H. Nordman, Dynamic thermal modelling of power
on sidewalls – Part 1: heat transfer, Int. J. Heat Mass Transfer 97 (2016) 1069– transformers, IEEE Trans. Power Delivery 20 (2005) 197–204.
1080. [26] Y.L. He, W.Q. Tao, Y. Wang, J.F. Fan, The research progress of heat exchange
[14] P. Li, Y.H. Xie, D. Zhang, Laminar flow and forced convective heat transfer of equipment comprehensive evaluation index, in: Proceedings of the Annual
shear-thinning power-law fluids in dimpled and protruded microchannels, Int. Conference of Chinese Society of Engineering Thermophysics (2011).
J. Heat Mass Transfer 99 (2016) 372–382. [27] S.G. Kandlikar, W.J. Grande, Evolution of microchannel flow passages-
[15] H.E. Mghari, H. Louahlia-Gualous, Experimental and numerical investigations thermohydraulic performance and fabrication technology, Heat Transfer Eng.
of local condensation heat transfer in a single square microchannel under 24 (2003) 3–17.
variable heat flux, Int. Commun. Heat Mass 71 (2015) 197–207. [28] A. Sakanova, C.K. Chan, J. Zhao, Performance improvements of microchannel
[16] G.D. Xia, D. Ma, Y.L. Zhai, Experimental and numerical study of fluid flow and heat sink using wavy channel and nanofluids, Int. J. Heat Mass Transfer 89
heat transfer characteristics in microchannel heat sink with complex (2015) 59–74.
structure, Energy Convers. Manage. 105 (2015) 848–857.