Journal of the Brazilian Society of Mechanical Sciences and Engineering (2023) 45:178

https://doi.org/10.1007/s40430-023-04097-x

TECHNICAL PAPER

Coupling CFD and RSM to optimize the flow and heat transfer
performance of a manifold microchannel heat sink
Farzad Pourfattah1 · Mohsen Faraji Kheryrabadi2 · Lian‑Ping Wang1,3

Received: 27 September 2022 / Accepted: 7 February 2023 / Published online: 1 March 2023
© The Author(s), under exclusive licence to The Brazilian Society of Mechanical Sciences and Engineering 2023

Abstract
Maintaining the operating temperature within the allowable range for electronic components is crucial. This work aims to
optimize the design of a heatsink manifold microchannel where the working fluid is MWCNT/water-nanofluid. The design
parameters include inlet width (Linlet ), outlet width (Loutlet ), heatsink height (hf ), and MWCNT nanoparticle volume fraction
in the working fluid (𝜑). Minimum pressure drop and minimum thermal resistance are selected as the objective functions.
The finite volume method simulates the flow field and heat transfer at each design point. A regression model between the
objective functions and the design variables is derived by utilizing the response surface method, and the sensitivity analysis
of objective functions is performed by Pareto chart analysis. Finally, the response optimization method configures the optimal
design points as Linlet , Loutlet , hf being 85, 91, 245 𝜇m, respectively, and 𝜑 0.016, corresponding to a pressure loss at 2677 Pa
and thermal resistance at 0.8281 K/W. According to the results, the outlet width and heatsink height significantly affect the
pressure drop and thermal resistance. Moreover, the physics of the flow field shows that the strength of the corner vortex and
separation on the manifold can play a significant role in the thermal and hydraulic performance of the manifold microchannel
heat sink. A numerical simulation has been performed to assess the regression model’s accuracy in predicting the thermal and
fluid performance at the optimum point, showing a good agreement between the model prediction and the simulation results.

Keywords CFD · Manifold microchannel · Optimization · Thermal management

Abbreviations
cp,f Specific heat of the working flow (J ­kg−1 ­K−1)
hf Height of the fin (μm)
Technical Editor: Ahmad Arabkoohsar.
hm Height of the manifold (μm)
* Farzad Pourfattah hs Height of the heatsink (μm)
farzadpourfattah@gmail.com kf Thermal conductivity of the working flow (W
Mohsen Faraji Kheryrabadi ­m−1 ­K−1)
mohsen_faraji70@yahoo.com ks Thermal conductivity of the manifold (W
Lian‑Ping Wang ­m−1 ­K−1)
wanglp@sustech.edu.cn Linlet Inlet width (μm)
1
Lm Manifold width (μm)
Guangdong Provincial Key Laboratory of Turbulence
Research and Applications, Center for Complex Flows
Loutlet Outlet width (μm)
and Soft Matter Research and Department of Mechanics MMC Manifold Microchannel
and Aerospace Engineering, Southern University MWCNT Multi-walled carbon nanotubes
of Science and Technology, Shenzhen 518055, n Normal to boundary
People’s Republic of China
nf Nanofluid
2
Department of Mechanical Engineering, Faculty P Pressure of the working flow (N ­m−2)
of Engineering, University of Isfahan, Hezar Jerib Ave.,
Isfahan 81746‑73441, Iran
Pin Inlet pressure (N ­m−2)
3
Pout Outlet pressure (N ­m−2)
Guangdong‑Hong Kong‑Macao Joint Laboratory
for Data‑Driven Fluid Mechanics and Engineering
Q Applied heat flux (W/cm2)
Applications, Southern University of Science RSM Response surface method
and Technology, Shenzhen 518055, China RT Thermal resistance (K/W)

13
Vol.:(0123456789)
178 Page 2 of 20 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2023) 45:178

T Temperature of the working flow (K)

Tinlet Temperature of the coolant fluid (K)
Tmax Maximum temperature of heated surface (K)
Ts,interface Solid–fluid interface, fluid side
Tf ,interface Solid–fluid interface, solid side
u Velocity of the working flow (m/s)
wc Width of the heatsink (μm)
wf Width of the fin (μm)
Greek symbols
ΔP Pressure drop (N ­m−2)
𝜇 Viscosity of the working flow (kg/s)
𝜌f Density of the working flow (kg/m3)
𝜌p Density of the nanoparticle (kg/m3)
𝜑 Nanoparticle volume fraction

Fig. 1  Schematic illustration of a manifold microchannel heat sink

1 Introduction
experimental investigations have been conducted [10–14].
Continuous heat dissipation has been the most common To enhance the heat transfer performance, researchers have
problem in electronic components. The applied heat flux investigated the effect of geometrical parameters such as
on the electronic components increases their operating inlet–outlet width ratio, channel depth [15, 16], manifold
temperature, and exceeding the operating temperature above depth [17], the gap between the manifold and channel [18],
the allowable temperature adversely affects the performance and the choice of working fluid [19] on the flow and heat
of the electronic devices, lifetime [1] and reliability [2]. It transfer performance. In addition, some investigations have
can also weaken their optimal performance and even cause focused on the effect of the secondary flow on the flow field
failure [3]. Over 50% of electronic component failures are and heat transfer rate in the MMC [20, 21]. Researchers have
attributed to operating temperatures above the allowable found that the secondary flow causes the re-development of
limit [4]. Therefore, an effective cooling system is an thermal boundary layer and enhances flow mixing, leading to
essential requirement for electronic components. thermal resistance reduction and heat transfer enhancement.
Recently, two factors have made the cooling of electronic In a numerical study, Ryu et al. [9] showed that the
components more complicated. The first is the increased inlet–outlet width ratio and channel depth significantly
power input of the electronic system due to enhanced affected the thermal resistance under a given pumping
capabilities, and the second is the trend to miniaturize power. They regarded water as a working fluid and simulated
electronic equipment [5]. Therefore, innovative cooling the laminar flow in a manifold microchannel. Their results
systems must be designed to remove the applied high heat indicated that the optimal inlet–outlet width ratio remains
flux in the small cross section. Cooling systems based on constant as the Reynolds number varies. Comparison of the
traditional microchannel heatsinks are commonly used to thermal performance of the microchannel manifold with
cool electronic components due to their high heat transfer traditional microchannels has shown that in the manifold
coefficient [6]. They provide acceptable heat transfer microchannel heat sink, the thermal resistance is reduced
performance; however, due to their long length, they by half and the temperature uniformity on the heated wall
suffer high-pressure loss [7]. Consequently, to improve is improved by tenfold. By employing the computational
the heat transfer and flow performance of the traditional fluid dynamics (CFD) approach, Luo et al. [15] simulated
microchannels, using a manifold microchannel (MMC) as a the boiling process in a manifold microchannel under heat
heat sink has been proposed [8]. A schematic illustration of fluxes ranging from 50 W/cm2 to 400 W/cm2 and investigated
a MMC is exhibited in Fig. 1. As it can be seen, the coolant the effect of manifold divider dimension. They reported 43%
is divided by manifold dividers, and it enters to the heat pressure reduction when the ratio of manifold inlet to outlet
sink through the vertical downward coolant path. After length ranges from 1 to 2, suggesting that this ratio has the
passing through the horizontal path, a vertical upward path minimum pressure loss and maximum heat transfer. Bae et al.
is directed to the outlet. In the manifold microchannel, the [18] performed a numerical simulation to investigate the
shortened flow path reduces the pressure drop [9]. MMCs manifold-microchannel heat sink flow-thermal performance
have recently attracted the researchers’ attention in the field for cooling electronic components. They examined the effect
of cooling electronic components that both numerical and of a gap between the manifold and the microchannel during

13
Journal of the Brazilian Society of Mechanical Sciences and Engineering (2023) 45:178 Page 3 of 20 178

assembly. Their results showed that the gap (in micron size) nanofluid. They reported that the thermal resistance of the
could drastically reduce manifold-microchannel performance. heatsink as coolant was MWCNT-based nanofluid is lower
Their study also revealed the necessity of uniform and than the case that coolant was oxide-metallic nanofluid.
hermetic bonding between the manifold and the micro fins' Although high volume fraction MWCNT enhances heat
tips. transfer rate which is desirable from thermal performance
While water is considered a working fluid in most studies point of view, on the other hand increasing nanoparticle
and solid material is supposed to be fixed (usually copper or volume fraction cause to a significant increase in viscosity,
silicon), Pan et al. [22] examined the effect of the thermal which dramatically increases the pumping power inside
conductivity of silicon wall on the optimum inlet–outlet the MHS [41]. Therefore, to increase heat transfer with
width ratio and investigated the manifold thermal and fluid a minimum increase in pressure loss, the nanoparticle
performance of microchannel in the case of two different volume fraction should follow an optimum range. Hence,
working fluids, water and HFE7100. The coolant fluid optimization methods need to be used to find the optimum
and thermophysical properties of walls were considered volume fraction.
constant, and the applied heat flux was 120 W/cm2. Based Based on our knowledge and literature review, the
on their results, the optimal inlet–outlet width ratio rises by optimization and parametric analysis of microchannel
increasing the dynamic viscosity of the working fluid and manifold arrangement when the working fluid is MWCNT/
the channel wall's thermal conductivity. Also, their results water nanofluid has not been investigated that is covered
showed that the optimal aspect increases by decreasing the in the current investigation.
coolant's specific heat capacity and thermal conductivity. In this study, the flow field and heat transfer in a MMC
In the optimal design of heatsinks, several geometrical are simulated by finite volume method, the effect of the
parameters affect thermal and fluid performance. Designing design variables including the inlet and outlet width,
an optimum heatsink (maximum heat transfer rate and manifold height, and MWCNT nanoparticle volume fraction
minimum pressure drop) requires statistical and optimization on the hydrothermal performance of the MMC have been
methods [23, 24]. In recent years, researchers have examined. The thermal resistance and pressure loss are
emphasized the importance of coupling numerical results considered objective functions. In the current investigation
with optimization methods to maximize the heat transfer the numerical results coupled with the response surface
rate and minimize the pressure loss [25–31]. Reducing method to optimize the MMC configuration to achieve the
computational costs, sensitivity analysis, and determining maximum heat transfer and minimum pressure loss. As
the most effective and insignificant parameters, providing the working fluid is water-MWCNT. The remaining sections of
relationship between the dependent variables and objective the paper are arranged in the following sections. Section 2
functions are among the advantages of using optimization provides the problem statement, followed by a presenting
methods. governing equation in Sect. 3. Simulation results are
Another method to increase the heat transfer rate is discussed in details in Sect. 4, including variance analysis,
enhancing the thermophysical properties of working response surface, sensitivity analysis, optimization and
fluid [32, 33]. Adding nanoparticles to the working physics of the optimized flow. Finally, conclusions are
fluid in the last decade has been widely investigated as summarized in Sect. 5.
one of the methods to increase the rate of heat transfer
in microchannels [34–37]. Some studies have examined
nanofluid use in MMCs and have concluded that although
nanofluids increase the pressure drop, they reduce 2 Problem statement
the operating temperature and enhance heat transfer
performance [10, 16]. In addition to investigating the In the present study, by coupling the numerical simulation
input/output width ratio, Pourfattah et al. [16] investigated of the flow field and heat transfer with the response surface
the effect of the volume fraction of ­Al2O3 nanoparticles method (RSM), manifold microchannel's flow and heat
in a laminar flow within MMC. Their results show that transfer performance are optimized. As shown in Fig. 1,
an ­Al2O3 volume fraction of 2% in the laminar flow can a manifold microchannel heatsink is composed of several
provide maximum heat transfer rates with a minimal microchannel manifolds. According to the literature,
pressure loss. One of the nanoparticles that has attracted investigating the thermal and fluid performance of one
many researchers' attention is MWCNT (multi-walled of the manifold microchannels is sufficient [15, 42]. The
carbon nanotubes) because its thermophysical properties, schematic illustration of a microchannel manifold is shown
such as high thermal conductivity [38, 39]. Sarlak et al. in Fig. 2. As seen, the coolant fluid enters from the inlet
[40] examined the hydrothermal performance of a heatsink section and exits from the outlet section. In the current
by numerical simulation as the coolant was MWCNT based research, in a constant mass flow rate of the coolant fluid,

13
178 Page 4 of 20 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2023) 45:178

ideal for assigning the individual operation variables into

various evaluations through the rationalized number of
design points and a reliable curvature estimation to obtain
practical information for testing lack-of-fit [43].
Table 2 presents the nominal and coded values of 31
proposed designs. At each design point, the laminar flow
field and heat transfer are simulated by the finite volume
method, and the values of the objective functions including
the pressure loss and thermal resistance are calculated
(based on Eqs. 11, 12) by the numerical simulation; more
details are presented in the next sections.

3 Governing equations and numerical

simulation

In this investigation, the steady laminar flow field and

fluid–solid coupling heat transfer are simulated by ANSYS-
FLUENT 19. The governing equations for an incompressible
and Newtonian fluid in a laminar flow (Reynolds number in
Fig. 2  Geometrical parameters of MMC (L_inlet and L_outlet are
all cases is less than 100) are given by:
the width of the inlet and outlet section, respectively, L_mis manifold
width; h_s, h_f, and h_m are the height of heatsink, fin and manifold,
∇𝜌f u = 0 (1)
respectively, and w_f and w_c are the width of the fin and heatsink,
respectively)
(u.∇)𝜌f u = −∇P + 𝜇∇2 u (2)

Table 1  Upper and lower bounds of the design parameters 𝜌f cp,f (u.∇T) = kf ∇2 T (3)
Parameter Level
where u is the velocity, P stands for the pressure, T shows the
−1 0 +1 temperature, and 𝜌f , 𝜇 cp,f and kf illustrate density, viscosity,
hf (μm) 100 200 300 specific heat and thermal conductivity of the working flow,
Linlet (μm) 50 125 200 respectively.
Loutlet (μm) 50 125 200 For solid fin and the substrate, the energy equation is:
φ(Volume fraction) 0 0.01 0.02
ks ∇2 T = 0 (4)

where ks is the thermal conductivity of manifold and

the effect of geometric parameters and volume fraction substrate.
of nanoparticles in the working fluid on the thermal and At the solid–fluid interface, the continuity of temperature
fluid performance has been investigated. Inlet and outlet and heat flux are applied as:
section width ( Linlet and Loutlet ), the heatsink height hf , and
MWCNT nanoparticle volume fraction are considered key Ts,interface = Tf ,interface (5)
design variables.
𝜕Tf ,interface 𝜕Ts,interface
2.1 Materials and methods kf = ks (6)
𝜕n 𝜕n
Table 1 determines the lower and upper bounds of each The boundary conditions for solving the above-mentioned
design parameter. In the present study, the RSM-Central governing equations are specified as follows. Inlet velocity
composite design (CCD) algorithm has been used to and outlet pressure are applied for inlet and outlet sections.
determine the design points. CCD is a standard, effective, The constant heat flux is imposed on the lower wall (100W/
and the most commonly used design of RSM. The CCD is

13
Journal of the Brazilian Society of Mechanical Sciences and Engineering (2023) 45:178 Page 5 of 20 178

Table 2  Design points and the Design Point hf (μm) Linlet (μm) Loutlet (μm) φ(Volume fraction) Numerical results
value of the objective functions
in each design point K
RT ( W ) ΔP(Pa)

1 300 (1) 200 (1) 200 (1) 0 (−1) 1.16 1491.9

2 100 (−1) 50 (−1) 200 (1) 0 (−1) 1.06 3701
3 100 (−1) 50 (−1) 50 (−1) 0.02 (1) 0.76 5834
4 200 (0) 50 (−1) 125 (0) 0.01 (0) 0.92 3126.7
5 200 (0) 125 (0) 125 (0) 0.01 (0) 0.89 1887.4
6 200 (0) 125 (0) 125 (0) 0 (−1) 0.92 1870.5
7 100 (−1) 200 (1) 50 (−1) 0.02 (1) 0.68 12,015.1
8 300 (1) 200 (1) 200 (1) 0.02 (1) 1.11 1517.4
9 200 (0) 125 (0) 125 (0) 0.01 (0) 0.89 1887.4
10 200 (0) 125 (0) 50 (−1) 0.01 (0) 0.75 5164.8
11 300 (1) 125 (0) 125 (0) 0.01 (0) 0.93 1540.2
12 100 (−1) 200 (1) 200 (1) 0 (−1) 1.08 4728.9
13 200 (0) 125 (0) 200 (1) 0.01 (0) 1.04 1506
14 200 (0) 125 (0) 125 (0) 0.01 (0) 0.89 1887.4
15 300 (1) 50 (−1) 200 (1) 0.02 (1) 0.98 2887.8
16 100 (−1) 200 (1) 50 (−1) 0 (−1) 0.71 11,761.4
17 300 (1) 200 (1) 50 (−1) 0 (−1) 0.85 7487.4
18 200 (0) 125 (0) 125 (0) 0.02 (1) 0.88 1904.3
19 200 (0) 125 (0) 125 (0) 0.01 (0) 0.89 1887.4
20 200 (0) 125 (0) 125 (0) 0.01 (0) 0.89 1887.4
21 300 (1) 200 (1) 50 (−1) 0.02 (1) 0.81 7648.8
22 300 (1) 50 (−1) 50 (−1) 0.02 (1) 0.79 4061.4
23 100 (−1) 125 (0) 125 (0) 0.01 (0) 0.79 4000.7
24 200 (0) 125 (0) 125 (0) 0.01 (0) 0.89 1887.4
25 200 (0) 125 (0) 125 (0) 0.01 (0) 0.89 1887.4
26 100 (−1) 50 (−1) 50 (−1) 0 (−1) 0.79 5723.6
27 100 (−1) 50 (−1) 200 (1) 0.02 (1) 0.99 2965.5
28 300 (1) 50 (−1) 50 (−1) 0 (−1) 0.83 4302.9
29 200 (0) 200 (1) 125 (0) 0.01 (0) 0.91 2589.8
30 300 (1) 50 (−1) 200 (1) 0 (−1) 1.02 2829.2
31 100 (−1) 200 (1) 200 (1) 0.02 (1) 1.03 4928

Table 3  Thermophysical properties of working fluid and solid to a constant level below 1­ 0–5 and 1­ 0–6, respectively, the
materials convergence of the numerical solution is assured [16].
parameters Water MWCNT Solid- MWCNT/eater nanofluid is the considered working fluid,
nanoparticle [44] Silicon and the material of the substrate and fins is Silicon. The
[19] thermophysical properties of the materials are presented in
ρ(kg∕m3 ) 997 1600 2300 Table 3.
μ(Pa.s∕m) 0.001003 – – Based on the literature, it can be found that nanofluids
Cp (kg∕m3 ) 4185 796 700 thermophysical properties depend on the particle volume
k(W∕m.K) 0.65 3000 150
fraction as follows [45]:
𝜌nf = 𝜑𝜌p + (1 − 𝜑)𝜌f (7)
2
cm ). To simulate a unit cell of the MMC, on the sidewalls,
symmetry boundary conditions are implemented. The 𝜇f
𝜇nf = (8)
above governing equations are discretized by the second- (1 − 𝜑)2.5
order upwind method, and the pressure and velocity fields
are coupled by SIMPLE method [9]. By monitoring the
scaled residuals for the momentum and energy calculation

13
178 Page 6 of 20 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2023) 45:178

(𝜌Cp )nf = 𝜑(𝜌cp )p + (1 − 𝜑)(𝜌cp )f (9) the experimental data as working fluid is MWCNT (0.1%
volume fraction)-water. As can be seen, the present results
[( ) ] show good consistency with the experimental data (the
kp + 2kf − 2𝜑(kf − kp ) maximum error is about 4%). In conclusion, the assumptions
knf = kf ( (10)
and numerical methods lead to satisfactory accuracy.
)
kp + 2kf + 𝜑(kf − kp )

where subscripts nf, p and f are related to nanofluid, 4.2 Response surface

nanoparticle, and working fluid and 𝜑 is the MWCNT
volume fraction. This section shows the effect of changes in the input
parameters on the objective function by presenting the
response surface. Figure 4 illustrates the response surface
4 Results of pressure loss with inlet–outlet width when the height
of heatsink is minimum and MWCNT volume fraction is
The flow field and heat transfer were simulated at each maximum. As already mentioned, the mass flow rate is
design point, and then, the objective functions were fixed in all design points. It can be seen from Fig. 4a that
calculated. Pressure loss (ΔP) [46] and thermal resistance when the heatsink height is minimum, in each outlet width
(RT ) [47] were considered the objective functions: ( Loutlet ), the pressure loss increases when inlet width ( Linlet )
ΔP = Pin − Pout (11) increases. Furthermore, it has an inverse relation with the
outlet width ( Loutlet ). When the outlet width is constant
Tmax − Tinlet ( Loutlet ), increasing the inlet width ( Linlet ) dramatically
RT = (12) impacts the pressure loss significantly when the outlet width
Q
is minimized. The trend shows that the minimal pressure loss
where Pin , Tinlet are the inlet pressure (Pa) and temperature corresponds to the minimum inlet section and maximum
(K) of the coolant fluid and Pout stands for the outlet static outlet width. The trend can be the effect of the blockage ratio
pressure (Pa). Tmax refers to the maximum temperature of (width of inlet section/width of outlet section)—the blockage
heated surface and Q depicts the applied heat flux (W). The ratio increases as the inlet width increases or the output
values of the objective functions for each design point are width decreases. When the heatsink height is minimum, the
presented in Table 2. flow to reach the outlet cross section faces severe velocity
field gradient and more pressure loss. Comparing pressure
loss variation in Fig. 4a, b shows that as heatsink height
4.1 Grid convergence and validation
increases, the effect of inlet width on the pressure loss is
weakened. The growth of the cross section in the horizontal
A 3-D structured mesh was applied for the discretization
path decreases velocity, causing less pressure reduction.
of the computational domain. Grid convergence study was
In Fig. 5, the response surface of the thermal resistance
conducted, and after using 100,000 elements, a negligible
is shown. As seen, without paying attention to the height
deviation was seen in the heat transfer coefficient and
of the heatsink, in a constant inlet width ( Linlet ), increasing
pressure drop, therefore, this grid number was used for all
the outlet width ( Loutlet ) leads to an increase in the thermal
simulations [16]. Accuracy of assumed numerical method
resistance that is not desirable from a heat transfer point
in predicting the flow field and heat transfer characteristics
of view. Therefore, the negative effect of decreasing outlet
was validated by Kermani’s experimental data [14]. The
width ( Loutlet ) on thermal resistance and the positive
pressure loss and heat transfer coefficient in a manifold
effect of increasing outlet width ( Loutlet ) on pressure drop
microchannel heat sink were experimentally studied by
(which reduces pressure drop, Fig. 4) demonstrate that a
Kermani [14]. Figure 3a, b compares the current numerical
multi-objective optimization method for optimal design is
results with the experimental data (heat transfer coefficient
necessary.
and pressure drop). Although the experimental results have
uncertainty, the results show satisfactory agreement between
4.3 Variance analysis results
the experimental data and the simulation results. Also, in
the current simulation, the working fluid is MWNCT-water
When the values of objective functions are determined, it is
nanofluid. Present results are validated by experimental data
possible to find a fitting relationship between independent
presented by Diao et al. [48], which examined the flow field's
variables and objective functions by the response surface
characteristics and heat transfer in the microchannel as the
methodology.
working fluid was MWCNT-water. Figure 3c compares the
Nusselt number of the current numerical simulation and

13
Journal of the Brazilian Society of Mechanical Sciences and Engineering (2023) 45:178 Page 7 of 20 178

Fig. 3  Validation of the numeri- (a) 80000

cal results: a. heat transfer coef-
Kermani Exp. Results
ficient and b. pressure drop of 70000 Current results
a MMC, coolant: water, c. the

Heat transfer coefficient (W/m K)

Nusselt number of a microchan-

2
nel, coolant: MWCNT-Water 60000

50000

40000

30000

20000 Error: 8%

10000
0.2 0.4 0.6 0.8 1.0 1.2

Mass flow (gr/s)

(b) 25

Kermani - Exp. Results

Current results
20
Pressure drop (kPa)

15

10

0
0.2 0.4 0.6 0.8 1.0 1.2

Flow rate (gr/s)

MWCNT volume fraction: 0.1 %

5.0
(c).

4.5
Diao et al. - Exp.Results
Current results
Nusselt number

4.0
max Error: 4 %

3.5

3.0

2.5
100 200 300 400 500 600

Reynolds number

13
178 Page 8 of 20 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2023) 45:178

(a). hf min (-1)

(a).
hf min (-1)
ϕ max (1) ϕ max (1)

14 1.05

1.00
12

ce (K/R)
0.95
)
Pressure loss (kPa

10
0.90

Therma resistan
8 0.85

0.80
6
200 200
0.75
180 180
4 160 160
0.70
140 140

)
120

µm
)
120

µm
2 0.65
100

in (
200 100

in (
180 180
160

L
160 80

L
140 80 140
120 120 60
60 100
Lo (µ
100
80 Lo (µ 80
60 40
ut m) 60 40 ut m) 40
40
(b). hf max (-1)
ϕ max (1)
hf max (1)
(b).
ϕ max (1)
1.15

1.10
10
ance (K/R)

1.05

1.00
8
Therma resist

0.95
)
Pressure loss (kPa

0.90
6 200
0.85
180
160
0.80 140
4

)
120

µm
0.75
100

in (
180
200 160

L
140 80
2 180 120
100 60
160 Lou (µ 80
60 40
140 t m) 40
)

120
µm

0
200 100
in (

180
160 Fig. 5  Response surface of thermal resistance a h_f max (1), φ max
L

140 80
120 60
100 (1) and b h_f min (− 1), φ max (1)
Lo (µ 80
60 40
ut m) 40

depicts the difference between the test data and the value
Fig. 4  Response surface of pressure loss a h_f max (1), φ max (1) obtained from the relation.
and b h_f min (− 1), φmax (1)
In order to find a proper relationship between
independent variables and objective functions, the variance
analysis is utilized to determine which parameters should
The multivariate model based on considering all the
be retained in the response surface correlation. To detect
linear, square and the cross-interaction terms can be written
the most significant parameters that should be considered
as follows [5].
in fitting relationship between independent variables and
n
∑ n
∑ n
∑ n
∑ objective functions, the variance analysis has been carried
Y = 𝛽0 + 𝛽i xi + 𝛽ij xi xj + 𝛽ii xi2 + 𝜀 (13) out. The variance analysis results of the pressure loss and
i=1 i=1 j=1,j≠i i=1
thermal resistance are presented in Table 4 and Table 5.
where 𝛽0 is the intercept, 𝛽i shows linear regression The analysis of variance table (ANOVA) presents the F
coefficient of ith factor, 𝛽ii illustrates quadratic regression value and P value (Probability value). P value < 0.05 shows
coefficients of ith factor, 𝛽ij stands for the interaction of the the parameter or interaction between them is statistically
ith and jth factors, Y refers to the objective function, and 𝜀 considerable [49] and more than 0.05 are meaningless

13
Journal of the Brazilian Society of Mechanical Sciences and Engineering (2023) 45:178 Page 9 of 20 178

Table 4  Analysis of variance Source DOF Adj. sum of Adj. mean F value P value
table for the thermal resistance squares (SS) squares (MS)
(RT )
Model 14 0.40950 0.029251 123.81 0.000 Significant
hf 1 0.019586 0.019586 82.90 0.000 Significant
Linlet 1 0.002222 0.002222 9.41 0.007 Significant
Loutlet 1 0.345488 0.345488 1462.34 0.000 Significant
𝜑 1 0.007275 0.007275 30.79 0.000 Significant
hf × hf 1 0.001478 0.001478 6.25 0.024 Significant
Linlet × Linlet 1 0.002222 0.002222 9.41 0.007 Significant
Loutlet × Loutlet 1 0.000142 0.000142 0.60 0.450 Not Significant
𝜑×𝜑 1 0.000304 0.000304 1.29 0.273 Not Significant
hf × Linlet 1 0.011624 0.011624 49.20 0.000 Significant
hf × Loutlet 1 0.003713 0.003713 15.72 0.001 Significant
hf × 𝜑 1 0.000002 0.000002 0.01 0.920 Not Significant
Linlet × Loutlet 1 0.013010 0.013010 55.07 0.000 Significant
Linlet × 𝜑 1 0.000002 0.000002 0.01 0.920 Not Significant
Loutlet × 𝜑 1 0.000295 0.000295 1.25 0.280 Not Significant
Error 16 0.003780 0.000236
Lack-of-Fit 10 0.003780 0.000378 * *
Pure Error 6 0 0
Total 30 0.413288
R2 = 99.09%—R2_adj = 98.29%—R2_pred = 96.01%

Table 5  Analysis of variance Source DOF Adj. sum of Adj. mean F Value P Value
table for the pressure loss (ΔP) squares (SS) squares(MS)

Model 14 225,465,025 16,104,645 61.21 0.000 Significant

hf 1 26,623,591 26,623,591 101.19 0.000 Significant
Linlet 1 19,503,343 19,503,343 74.12 0.000 Significant
Loutlet 1 77,890,593 77,890,593 296.03 0.000 Significant
𝜑 1 1005 1005 0.00 0.951 Not Significant
hf × hf 1 2,092,178 2,092,178 7.95 0.012 Significant
Linlet × Linlet 1 2,521,356 2,521,356 9.58 0.007 Significant
Loutlet × Loutlet 1 5,553,290 5,553,290 21.11 0.000 Significant
𝜑×𝜑 1 571 571 0.00 0.963 Not Significant
hf × Linlet 1 7,763,328 7,763,328 29.51 0.000 Significant
hf × Loutlet 1 1,121,746 1,121,746 4.26 0.056 Not Significant
hf × 𝜑 1 1943 1943 0.01 0.933 Not Significant
Linlet × Loutlet 1 21,874,563 21,874,563 83.14 0.000 Significant
Linlet × 𝜑 1 130,990 130,990 0.50 0.491 Not Significant
Loutlet × 𝜑 1 33,884 33,884 0.13 0.724 Not Significant
Error 16 4,209,853 263,116
Lack-of-Fit 10 4,209,853 420,985 * *
Pure Error 6 0 0
Total 30 229,674,878
R2 = 98.17%—R2_adj = 96.56% −R2_pred = 91.82%

[50]. Based on Table 4, the effect of Linlet × Linlet , 𝜑 × 𝜑 , presented in Table 5, the pressure loss is not affected by 𝜑,
hf × 𝜑 , Linlet × 𝜑 , and Loutlet × 𝜑 on the response surface of 𝜑 × 𝜑 , hf × Loutlet , hf × 𝜑 , Linlet × 𝜑 , Loutlet × 𝜑.
the thermal resistance can be ignored. Based on the results As it is seen, the highest values for multiple determination
coefficient R2 (99.09% and 98.17%) are obtained by the

13
178 Page 10 of 20 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2023) 45:178

Fig. 6  Pareto chart for a thermal pareto chart for RT

resistance and b pressure loss (a)
L inlet
H ×ϕ
(f)*Phi
inlet × ϕ
LL(in)*Phi
LL
outlet × L (out)
(out)*L outlet
outlet × ϕ
LL(out)*Phi
Phi*Phi
ϕ×ϕ
h f × (f)
H (f)*H hf
L Linlet × L(in)
(in)*L inlet

Term
LL(in)
inlet
Hh (f)*L
f × L (out)
outlet
ϕ
Phi
Hh (f)*L
f × L (in)
inlet
LLinlet × L(out)
(in)*L outlet
H (f)
hf
L outlet
Lmout

0 2.20 10 20 30 40

t value

pareto chart for ∆P

(b) 2.20
ϕ×ϕ
Phi*Phi
ϕ
Phi
f×ϕ
H h(f)*Phi
outlet × ϕ
LL (out)*Phi
L inlet × ϕ
L (in)*Phi
Hh (f)*L
f × L outlet
(out)
h f × h(f)f
H (f)*H
Term

L inlet × L (in)
L (in)*L inlet
L outlet × L outlet
L (out)*L (out)
Hh (f)*L
f × L inlet
(in)
L inlet
(in)
L inlet × L (out)
L (in)*L outlet
hf
H (f)
LL outlet
(out)

0 2 4 6 8 10 12 14 16 18

t value

testing methods and statistical analysis of the model. It and the vertical axis is the parameters, their squares, and
should be mentioned that a value of multiple determination their interactions. Based on the literature [52], if the value of
coefficient R2 being more than 0.8 shows acceptable fitting the standardized effect is less than 2.23, the parameters can
[51]. be omitted from the regression model. As can be seen, Loutlet
A statistical sensitivity analysis methodology called has the most significant impact on the pressure loss and
Pareto chart analysis is used to quantify the effects of the thermal resistance. Also, hf and Linlet × Loutlet influenced the
design parameters on the objective functions [21]. The flow field and heat transfer considerably, respectively. Based
Pareto bar chart for thermal resistance and pressure loss on Fig. 6, the 𝜑 × 𝜑 , Loutlet × Loutlet , Linlet × 𝜑 and hf × 𝜑
is presented in Fig. 6a, b, respectively. In these charts, the can be omitted from the regression model of the thermal
horizontal axis is the standardized effect of each parameter,

13
Journal of the Brazilian Society of Mechanical Sciences and Engineering (2023) 45:178 Page 11 of 20 178

resistance and hf × Loutlet,Linlet × 𝜑, hf × 𝜑, 𝜑 and 𝜑 × 𝜑 can Loutlet = −2080 + 2963 × Loutlet − 1169 × Linlet (22)
be ignored in the regression model of the pressure loss.
The variance analysis results led to the following
𝜕ΔP
regression models for thermal resistance and pressure loss: =0 (23)
𝜕𝜑
RT (K/W) = 0.89304(K/W) + 0.03298 hf + 0.01Linlet
Figure 7 shows the sensitivity of pressure drop to
+ 0.1385Loutlet − 0.02 𝜑 − 0.023hf × hf independent variables. In this figure, the horizontal
+ 0.03 × Linlet + 0.027hf × Linlet axis is the independent parameter, and the vertical axis
− 0.015hf × Loutlet + 0.028Linlet Loutlet is the pressure drop sensitivity. It is assumed that the
values of the other parameters are constant when the
(14)
sensitivity of one parameter is checked. Figure 7a shows
ΔP(Pa) = 1881 (Pa) - 1216 hf + 1041 Linlet - 2080 Loutlet the sensitivity of pressure drop to design variables by
+ 898 hfin × hfin + 986 Linlet × Linlet + 1463 Loutlet changing the hf when Linlet and Loutlet is the maximum.
× Loutlet - 697 hfin × Linlet - 1169 Linlet × Lout According to Fig. 7a, when the heatsink height is the
(15) minimum ( hf = −1), the pressure is mainly affected by the

It should be noted that to calculate the value of thermal (inlet section width and in the maximum heatsink height
hf = 1 , hf , plays a significant role on the pressure loss.
)
resistance and pressure drop in the desired point (in the Figure 7b shows the sensitivity of the pressure drop in
range of input data), the value of the input parameter different values of Linlet to the design parameters when hfin
should be used in the coded format based on Table.1. and Loutlet are maximum. As seen, the prominent effect on
the pressure loss is related to theLinlet , and by increasing
4.4 Sensitivity analysis the Linlet , the pressure loss decreases.
Figure 7c shows the sensitivity of the pressure drop in
After obtaining the regression model and the relationship different values of Loutlet to the design parameters when
between inputs parameters and output variable, hfin and Linlet are maximum. Based on Fig. 7c, the most
sensitivity analysis can show the effect of the changes critical parameter that affects the pressure loss is Loutlet ,
of each input parameter on the output variable. A high by increasing the Loutlet the pressure drop sensitivity to the
sensitivity coefficient for a given input parameter shows Linlet increases. It is worth mentioning that when the Linlet is
its considerable effects on the results, while lower ones constant (maximum value), the pressure loss relative to the
can be ignored in modeling. The derivatives of the output hf is not significantly sensitive.
variable (thermal resistance and pressure loss), Eqs. (16 Figure 8 illustrates the sensitivity of the manifold
and 17), with respect to input parameters were calculated microchannel heat sink thermal resistance to the design
as follows: variables. As seen, the Loutlet is the critical parameter that
𝜕RT significantly affects the thermal resistance. Based on Fig. 8
= 0.03298l − 0.0477 × hf + 0.027 × Linlet − 0.015 × Loutlet the sensitivity of the thermal resistance to the nanoparticle
𝜕hf
(16) volume fraction is a negative value that shows by adding the
nanoparticles to the working fluid, the thermal resistance
decreases. Figure 8 indicates the need for configuration
𝜕RT
= 0.01 + 117 × Linlet + 0.027 × hf + 0.028 × Lout
𝜕Linlet optimization because the inlet/outlet width and heatsink
(17) height have a greater impact on the thermal resistance
𝜕RT reduction than the nanoparticle volume fraction.
= 0.1385 − 0.01523 × hf + 0.028 × Linlet (18)
𝜕Lout
4.5 Optimization
𝜕RT
= −0.02 (19) After obtaining the relationship between the objective
𝜕𝜑
functions (response surface) and independent parameters
(design variables), the response optimization method
𝜕ΔP
= −1216 + 1796 × hf − 697 × Linlet (20) proposes the optimal case. As mentioned before, the
𝜕hf objective functions in the present investigation are minimal
thermal resistance and pressure loss. The response
𝜕ΔP optimization method-desirability function approach is a set
= 1041 + 1972 × Linlet − 697 × hf − 1169 × Loutlet
𝜕Linlet of statistical and mathematical relationships used to find the
(21)

13
178 Page 12 of 20 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2023) 45:178

Fig. 7  Sensitivity analysis of Linlet= 1, Loutlet= 1, ϕ=-1,0,1

the response–pressure loss (∆P) (a).
relative to a hf, b L_inlet, and sensitivity to h f
c L_outlet 4000 sensitivity to Linlet
sensitivity to Loutlet
sensitivity to ϕ

2000

Sensitivity
0

-2000

-4000
-1 0 1
hf
hf= 1, Loutlet= 1, ϕ= -1,0,1
(b).
sensitivity to hf
4000 sensitivity to L inlet
sensitivity to Loutlet
sensitivity to ϕ

2000
Sensitivity

-2000

-4000
-1 0 1
Linlet

hf= 1, Linlet= 1, ϕ= -1,0,1

(c).
sensitivity to hf
4000
sensitivity to Linlet
sensitivity to Loutlet
2000 sensitivity to ϕ
Sensitivity

-2000

-4000

-6000
-1 0 1
Loutlet

13
Journal of the Brazilian Society of Mechanical Sciences and Engineering (2023) 45:178 Page 13 of 20 178

Fig. 8  Sensitivity analysis of (a). Linlet= 1, Loutlet= 1, ϕ= -1,0,1

the response–thermal resistance
(R_T) relative to a hf, b L_inlet, 0.20 sensitivity to hf
and c L_outlet sensitivity to Linlet
sensitivity to Loutlet
sensitivity to ϕ
0.15

Sensitivity
0.10

0.05

0.00

-1 0 1
hf

(b). hf= 1, Loutlet= 1, ϕ= -1,0,1

sensitivity to hf
0.20
sensitivity to Linlet
sensitivity to Loutlet
sensitivity to phi
0.15
Sensitivity

0.10

0.05

0.00

-1 0 1
Linlet

(c). hf= 1, Linlet= 1, ϕ= -1,0,1

sensitivity to hf
0.20 sensitivity to Linlet
sensitivity to Loutlet
sensitivity to ϕ
0.15
Sensitivity

0.10

0.05

0.00

-1 0 1

Loutlet

13
178 Page 14 of 20 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2023) 45:178

Table 6  Individual desirability Time RT (K∕W) ΔP(Pa) Individual Individual Composite

and composite desirability of desirability for RT desirability for ΔP desirability
design points and the optimal
point 1 1.16 1491.9 0.00 1.00 0.00
2 1.06 3701 0.21 0.79 0.41
3 0.76 5834 0.83 0.59 0.70
4 0.92 3126.7 0.50 0.84 0.65
5 0.89 1887.4 0.56 0.96 0.74
6 0.92 1870.5 0.50 0.96 0.69
7 0.68 12,015.1 1.00 0.00 0.00
8 1.11 1517.4 0.10 1.00 0.32
9 0.89 1887.4 0.56 0.96 0.74
10 0.75 5164.8 0.85 0.65 0.75
11 0.93 1540.2 0.48 1.00 0.69
12 1.08 4728.9 0.17 0.69 0.34
13 1.04 1506 0.25 1.00 0.50
14 0.89 1887.4 0.56 0.96 0.74
15 0.98 2887.8 0.38 0.87 0.57
16 0.71 11,761.4 0.94 0.02 0.15
17 0.85 7487.4 0.65 0.43 0.53
18 0.88 1904.3 0.58 0.96 0.75
19 0.89 1887.4 0.56 0.96 0.74
20 0.89 1887.4 0.56 0.96 0.74
21 0.81 7648.8 0.73 0.41 0.55
22 0.79 4061.4 0.77 0.76 0.76
23 0.79 4000.7 0.77 0.76 0.77
24 0.89 1887.4 0.56 0.96 0.74
25 0.89 1887.4 0.56 0.96 0.74
26 0.79 5723.6 0.77 0.60 0.68
27 0.99 2965.5 0.35 0.86 0.55
28 0.83 4302.9 0.69 0.73 0.71
29 0.91 2589.8 0.52 0.90 0.68
30 1.02 2829.2 0.29 0.87 0.50
31 1.03 4928 0.27 0.67 0.43
Optimal point 0.8271 2566.0379 0.69 0.90 0.783

optimal design point [53]. In this method, for each design To verify the accuracy of optimization method in
point, the individual desirability and composite desirability predicting the thermal resistance and pressure loss at
have been calculated as given in Table 6. More details about the optimum point, the flow field and heat transfer at the
that approach can be found in [54]. Based on Table 6, one optimum candidate point conditions have been simulated. In
optimum point is proposed (the last row) that the composite Table 7, the results obtained from numerical simulation are
desirability of the optimum point is maximum (0.783). compared with the RSM prediction. As it can be seen, the
Figure 9 illustrates the behavior of the objective functions error of pressure loss and the thermal resistance prediction
(response surface) in terms of independent variables. by RSM at the optimum point are equal 0.64% and 4.14%,
According to Fig. 9, the values of the composite, pressure respectively, which shows an acceptable consistency.
loss, and thermal resistance desirability are 0.78, 0.89 and
0.68. If the value of this function is greater than 0.6, it 4.6 Physics of the optimized flow
will be acceptable and shows accuracy of the optimization
[55]. Based on Fig. 9, the coded format of the optimized In this section, effor ts are made to show the
point for Linlet , Loutlet , hf and 𝜑 is -0.45, -0.55, 0.45 and 0.8, effect of the MMC configuration on the physics
respectively, which, based on Table 1, corresponds to Linlet ,
Loutlet , hf being 85, 91, 245 𝜇m and 𝜑 0.016.

13
Journal of the Brazilian Society of Mechanical Sciences and Engineering (2023) 45:178 Page 15 of 20 178

Fig. 9  Optimization of L_inlet, L_outlet, h_f, and φ in terms of composite desirability, minimum pressure loss, and minimum thermal resistance

Table 7  Compilation of numerical results and RSM prediction for the not reattach to the manifold, and the dead zone (Zone D1—
optimum candidate point ( Linlet , Loutlet , hf being 85, 91, 245 𝜇m, and pointed in Fig. 10b) on the outlet side (left side) reduces
𝜑 0.016)
the heat transfer between the working fluid and solid walls.
RT (K∕W) ΔP(Pa) Moreover, the coolant fluid and substrate interface in the
horizontal cooling path can play a significant role in the
RSM prediction 0.8271 2566
cooling performance. As seen in Fig. 10b, the corner vortex
CFD results 0.8218 2677
causes separation at Point S (z = 380 μm). Velocity reduction
Error 0.64% 4.14%
before the separation point decreases the convection heat
transfer. The temperature distribution is shown in Fig. 10c.
As seen in the separation point (Point S—pointed in
o f f l ow. F o r t h i s r e a s o n , d e s i g n p o i n t 1 2 Fig. 10b), the working fluid temperature has been increased
(Linlet = Loutlet= 200𝜇m, hf = 100𝜇mand𝜑 = 0) and optimum due to the momentum reduction of the flow. Based on
point (Linlet = 85𝜇m, Loutlet= 91𝜇m, hf = 245𝜇mand𝜑 = 0.016) Fig. 10 b and c, in the corner vortex (Zone D2), the mixing
are selected to compare. Based on the numerical results and turbulent intensity increases, enhancing heat diffusion
presented in Table 2, at design point 12, the pressure loss into the cooing fluid.
and thermal resistance are 4729 (Pa) and 1.08 (K/W), The flow pattern (counter of velocity magnitude and
respectively. According to Table 7, the pressure loss and streamlines) and the temperature distribution of the optimum
thermal resistance are 2677 (Pa) and 0.82 (K/W) at the case are shown in Fig. 11a, b and c, respectively. As seen in
optimum point. Figure 10a, b shows the velocity distribution Fig. 11, in comparing design point 12, at the optimum case,
and streamline for design point 12. As seen, due to the the outlet width decreases (width of the manifold increases),
maximum width of the inlet and outlet, the width of the and the height of the heatsink increases. Based on Fig. 11a,
manifold is minimum, the corner of the manifold in the the separated flow due to Point A can reattach the manifold
inlet side (Point A—pointed in Fig. 10c) leads to complete that leads to a decrease in the size of the dead zone in the
separation (without reattachment). The separated flow could outlet section (Zone D3), the corner vortex size (Zone D4)

13
178 Page 16 of 20 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2023) 45:178

Fig. 10  a Velocity distribution, b streamlines, and c temperature distribution for design point 12

decreases, and the separation point (Point S) is postponed reduces the convective heat transfer, and the maximum wall
(z = 405 μm). temperature occurs at the separation point (point S).
Comparison of the physics flow field in design point 12 The variation of wall temperature of the optimum point
and the optimum point shows that in design point 12, the is according to flow field pattern. As mentioned, the flow is
flow is accelerated in the horizontal cooling path. In contrast, accelerated in the vertical path, and then in the horizontal
it is accelerated in the inlet section at the optimum point. path, the thickness of the thermal and hydraulic boundary
This different behavior affected the convection heat transfer increases, leading to convective heat transfer reduction
rate, as the wall temperature variation in Fig. 12 shows. As and increasing the wall temperature. As seen, due to the
shown in Fig. 12, the trend of wall temperature changes of corner vortex and separation, the wall temperature increases
Line 1 (pointed in Fig. 10c) at design point 12 is different slightly and peak at maximum value at the separation point.
from the optimal point, and generally, the wall temperature Temperature reduction after the separation is not significant
at point 12 is higher. Because it was already stated that all because of the weak corner vortex.
configurations have the same mass flow rate, the larger
cross section of design point 12 reduces the velocity and,
therefore, reduces the heat transfer rate. The decrease in wall 5 Conclusions
temperature in design point 12 after the stagnation point can
be attributed to the reduction in cross-sectional area due to In this study, the effects of manifold configuration and
the presence of the manifold, leading to an increase in local volume fraction of nanoparticles in the working fluid on the
velocity, which can enhance the convective heat transfer. heat transfer rate and pressure loss have been investigated
Then, the flow faces an adverse pressure gradient due to the to optimize the MMC heat sink design. Design points are
corner vortex, and velocity decreases until it separates from selected based on the response surface method (central
the surface. The reduction in velocity before the separation

13
Journal of the Brazilian Society of Mechanical Sciences and Engineering (2023) 45:178 Page 17 of 20 178

Fig. 11  a Velocity distribution, b and streamlines, and c temperature distribution for the derived optimal case

318 objective functions. Variance analysis is used to remove the

DP 12 terms (design parameters) from the model that have a low
Optimum Point
317 impact on the objective functions. The fitting relationships
between the independent variables and response surface
Temperature (K)

316 are then proposed. Sensitivity of pressure drop and thermal

resistance to independent variables have been analyzed.
Finally, the optimization is accomplished by defining the
315
minimum pressure drop and the minimum thermal resistance
as the objective functions, and the optimum design point has
314
been obtained. The results indicated that:

313
0.0000 0.0001 0.0002 0.0003 0.0004 0.0005
• Loutlet is the most influential parameter that significantly
Z (m) affects pressure loss and thermal resistance.
• Loutlet , hf , Linlet × Loutlet , hf × Linlet ,
Fig. 12  Wall temperature distributions for the design point 12 and the Loutlet × Loutlet , Linlet × Linlet , hf × hf affect the pressure
optimum case drop.
• Loutlet × Loutlet , Linlet × Linlet , 𝜑, hf × hf
Loutlet × Loutlet , Linlet × Linlet , 𝜑, hf × hf affect the thermal
composite design). The flow field and heat transfer are resistance.
simulated at each design point, utilizing the finite volume • Sensitivity of thermal resistance and pressure loss to
method. Thermal resistance and pressure loss are considered Linlet , Loutlet and hf is varied, whereas sensitivity of

13
178 Page 18 of 20 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2023) 45:178

thermal resistance to 𝜑 is constant and equal to − 40.21, 7. Tan H, Wu L, Wang M, Yang Z, Du P (2019) Heat transfer
indicating that increasing MWCNT volume fraction improvement in microchannel heat sink by topology design and
optimization for high heat flux chip cooling. Int J Heat Mass
decreases thermal resistance and enhances heat transfer Transf 129:681–689. https://​doi.​org/​10.​1016/J.​IJHEA​TMASS​
rate. TRANS​FER.​2018.​09.​092
• hf = 245𝜇m , Linlet = 91𝜇m , Loutlet = 85𝜇m a n d 8. Harpole GM, Eninger JE (1991) Micro-channel heat
𝜑 = 0.016 are proposed as the optimum point that exchanger optimization. In: 1991 proceedings, seventh
IEEE semiconductor thermal measurement and management
offers the maximum heat transfer rate and minimum symposium, Phoenix, AZ, USA, pp 59–63. https://​ieeex​plore.​
pressure loss. Based on the numerical simulation in the ieee.​org/​docum​ent/​152913
optimum case, the pressure loss is 2677 Pa, and thermal 9. Ryu JH, Choi DH, Kim SJ (2003) Three-dimensional numerical
resistance is 0.8281 K/W. optimization of a manifold microchannel heat sink. Int J Heat
Mass Transf 46:1553–1562. https://​d oi.​o rg/​1 0.​1 016/​S 0017-​
• The width of inlet/outlet sections affects the corner 9310(02)​00443-X
vortex strength, separation location, and reattachment 10. Adio SA, Olalere AE, Olagoke RO, Alo TA, Veeredhi VR,
of the flow on the manifold. Ewim DRE, Olakoyejo OT (2021) Thermal and entropy analysis
of a manifold microchannel heat sink operating on CuO–water
nanofluid. J Braz Soc Mech Sci Eng 43:76. https://​doi.​org/​10.​
1007/​s40430-​020-​02772-x
11. Mohammadali R, Bayareh M, Sheikhzadeh GA (2021) Study of
flow uniformity within convergent microchannels with a circular
Author contributions F.P was involved in literature reviewing, manifold. J Braz Soc Mech Sci Eng 43:74. https://​doi.​org/​10.​
validation, data curation, post-processing, writing, and revising. 1007/​s40430-​020-​02784-7
M.F was involved in simulation, optimization and writing. LP.W 12. Gilmore N, Hassanzadeh-Barforoushi A, Timchenko V,
was involved in conceptualization, methodology, formal analysis, Menictas C (2021) Manifold configurations for uniform flow
and editing. via topology optimisation and flow visualisation. Appl Thermal
Eng 183:116227. https://​d oi.​o rg/​1 0.​1 016/j.​a pplt​h erma​l eng.​
Declarations 2020.​116227
13. Gilmore N, Timchenko V, Menictas C (2021) Manifold
Conflict of interest The authors declare that they have no known microchannel heat sink topology optimisation. Int J Heat Mass
competing financial interests or personal relationships that could have Transfer 170:121025. https://​doi.​org/​10.​1016/j.​ijhea​tmass​trans​
appeared to influence the work reported in this paper. fer.​2021.​121025
14. Kermani E (2008) Manifold microchannel cooling of
photovoltaic cells for high efficiency solar energy conversion
systems, Department of Mechanical Engineering, University of
References Maryland. https://​drum.​lib.​umd.​edu/​handle/​1903/​8974?​show=​
full
1. Yang X-H, Tan S-C, Ding Y-J, Liu J (2017) Flow and thermal 15. Luo Y, Li J, Zhou K, Zhang J, Li W (2019) A numerical study
modeling and optimization of micro/mini-channel heat sink. of subcooled flow boiling in a manifold microchannel heat sink
Appl Therm Eng 117:289–296. https://​doi.​org/​10.​1016/j.​applt​ with varying inlet-to-outlet width ratio. Int J Heat Mass Transf
herma​leng.​2016.​12.​089 139:554–563. https://​doi.​org/​10.​1016/j.​ijhea​tmass​trans​fer.​2019.​
2. Chen Y, Li B, Wang X, Yan Y, Wang Y, Qi F (2019) 05.​030
Investigation of heat transfer and thermal stresses of novel 16. Pourfattah F, Abbasian Arani AA, Babaie MR, Nguyen HM,
thermal management system integrated with vapour chamber Asadi A (2019) On the thermal characteristics of a manifold
for IGBT power module. Thermal Sci Eng Progress 10:73–81. microchannel heat sink subjected to nanofluid using two-phase
https://​doi.​org/​10.​1016/j.​tsep.​2019.​01.​007 flow simulation. Int J Heat Mass Transf. 143:118518. https://​doi.​
3. Morozumi A, Yamada K, Miyasaka T, Sumi S, Seki Y (2003) org/​10.​1016/J.​IJHEA​TMASS​TRANS​FER.​2019.​118518
Reliability of power cycling for IGBT power semiconductor 17. Sarangi S, Bodla KK, Garimella SV, Murthy JY (2014) Manifold
modules. IEEE Trans Ind Appl 39:665–671. https://​doi.​org/​10.​ microchannel heat sink design using optimization under
1109/​TIA.​2003.​810661 uncertainty. Int J Heat Mass Transf 69:92–105. https://​doi.​org/​
4. Wintrich A, Nicolai U, Tursky W (2011) Tobias reimann. 10.​1016/J.​IJHEA​TMASS​TRANS​FER.​2013.​09.​067
Application Manual Power Semiconductors, SEMIKRON 18. Bae D, Mandel R, Ohadi M (2017) Effect of bonding structure and
International GmbH heater design on performance enhancement of FEEDS embedded
5. Raeisian L, Niazmand H, Ebrahimnia-Bajestan E, Werle P manifold-microchannel cooling https://​doi.​org/​10.​1115/​IPACK​
(2019) Thermal management of a distribution transformer: 2017-​74158
an optimization study of the cooling system using CFD and 19. Lin Y, Luo Y, Li W, Cao Y, Tao Z, Shih TI-P (2021) Single-
response surface methodology. Int J Electr Power Energy Syst phase and two-phase flow and heat transfer in microchannel heat
104:443–455. https://​doi.​org/​10.​1016/j.​ijepes.​2018.​07.​043 sink with various manifold arrangements. Int J Heat Mass Transf.
6. Sohel Murshed SM, Nieto de Castro CA (2017) A critical https://​doi.​org/​10.​1016/j.​ijhea​tmass​trans​fer.​2021.​121118
review of traditional and emerging techniques and fluids for 20. Yang M, Cao BY (2019) Numerical study on flow and heat
electronics cooling. Renew Sustain Energ Rev 78:821–833. transfer of a hybrid microchannel cooling scheme using manifold
https://​doi.​org/​10.​1016/j.​rser.​2017.​04.​112 arrangement and secondary channels. Appl Therm Eng.

13
Journal of the Brazilian Society of Mechanical Sciences and Engineering (2023) 45:178 Page 19 of 20 178

159:113896. https://​doi.​org/​10.​1016/j.​applt ​herma​leng.​2019.​ water–alumina nanofluid by non-dominated sorting genetic

113896 algorithm and response surface method. Nanomaterials 10(5):901.
21. Yang M, Cao B-Y (2020) Multi-objective optimization of a https://​doi.​org/​10.​3390/​nano1​00509​01
hybrid microchannel heat sink combining manifold concept with 35. Ho CJ, Hsieh Y-J, Rashidi S, Orooji Y, Yan W-M (2020)
secondary channels. Appl Therm Eng 181:115592. https://d​ oi.o​ rg/​ Thermal-hydraulic analysis for alumina/water nanofluid inside
10.​1016/j.​applt​herma​leng.​2020.​115592 a mini-channel heat sink with latent heat cooling ceiling—an
22. Pan YH, Zhao R, Fan XH, Le Nian Y, Cheng WL (2020) Study experimental study. Int Commun Heat Mass Transfer 112:104477.
on the effect of varying channel aspect ratio on heat transfer https://​doi.​org/​10.​1016/j.​ichea​tmass​trans​fer.​2020.​104477
performance of manifold microchannel heat sink. Int J Heat Mass 36. Razeghi A, Mirzaee I, Abbasalizadeh M, Soltanipour H (2017)
Transf. https://​doi.​org/​10.​1016/j.​ijhea​tmass​trans​fer.​2020.​120461 Al2O3/water nano-fluid forced convective flow in a rectangular
23. Parlak Z, Parlak N, İslamoğlu Y (2021) Optimal design of curved micro-channel: first and second law analysis, single-phase
a converging diverging channel for convective heat transfer and multi-phase approach. J Braz Soc Mech Sci Eng 39:2307–
enhancement in fully developed turbulent air flow. Int J Thermal 2318. https://​doi.​org/​10.​1007/​s40430-​016-​0686-4
Sci 161:106732. https://​doi.​org/​10.​1016/j.​ijthe​r mals​ci.​2020.​ 37. Asadi A, Rafizadeh S, Molana M, Ghasemiasl R, Armaghani
106732 T (2020) Two‐phase study of nanofluids mixed convection and
24. Güngör S, Ceyhan U, Karadeniz ZH (2021) Optimization of heat entropy generation in an I‐shaped porous cavity with triangular
transfer in a grooved pipe model by stochastic algorithms and hot block and different aspect ratios. Math Methods Appl Sci.
DOE based RSM. Int J Thermal Sci 159:106634. https://​doi.​org/​ https://​doi.​org/​10.​1002/​mma.​7006
10.​1016/j.​ijthe​rmals​ci.​2020.​106634 38. Shajahan MI, Stephen C, Michael JJ, Arulprakasajothi
25. Abo-Zahhad EM, Ookawara S, Radwan A, Elkady MF, El-Shazly M, Rathnakumar P, Parthasarathy M (2023) Heat transfer
AH (2020) Optimization of stepwise varying width microchannel investigations of in-line conical strip inserts using MWCNT/
heat sink for high heat flux applications. Case Stud Thermal Eng. water nanofluid under laminar flow condition. Int J Thermal Sci
https://​doi.​org/​10.​1016/j.​csite.​2020.​100587 183:107844. https://​doi.​org/​10.​1016/j.​ijthe​rmals​ci.​2022.​107844
26. Pourfattah F, Sabzpooshani M, Toghraie D, Asadi A (2020) 39. Ajeeb W, Oliveira MSA, Martins N, Murshed SMS (2021)
Correction to: On the optimization of a vertical twisted tape Performance evaluation of convective heat transfer and laminar
arrangement in a channel subjected to MWCNT–water nanofluid flow of non-newtonian MWCNTs in a circular tube. Thermal Sci
by coupling numerical simulation and genetic algorithm. J Therm Eng Progress 25:101029. https://​doi.​org/​10.​1016/j.​tsep.​2021.​
Anal Calorim. https://​doi.​org/​10.​1007/​s10973-​020-​09584-0 101029
27. Ge Y, Lin Y, Tao S, He Q, Chen B, Huang S-M (2021) Shape 40. Sarlak A, Ahmadpour A, Hajmohammadi MR (2019) Thermal
optimization for a tube bank based on the numerical simulation design improvement of a double-layered microchannel heat sink
and multi-objective genetic algorithm. Int J Thermal Sci by using multi-walled carbon nanotube (MWCNT) nanofluids
161:106787. https://​doi.​org/​10.​1016/j.​ijthe​rmals​ci.​2020.​106787 with non-newtonian viscosity. Appl Therm Eng 147:205–215.
28. Liu S, Huang W, Bao Z, Zeng T, Qiao M, Meng J (2021) Analysis, https://​doi.​org/​10.​1016/j.​applt​herma​leng.​2018.​10.​084
prediction and multi-objective optimization of helically coiled 41. Alnaqi AA, Alsarraf J, Al-Rashed AAAA, Afrand M (2021)
tube-in-tube heat exchanger with double cooling source using Thermal-hydraulic analysis and irreversibility of the MWCNTs-
RSM. Int J Thermal Sci 159:106568. https://​doi.​org/​10.​1016/j.​ SiO2/EG-H2O non-Newtonian hybrid nanofluids inside a zigzag
ijthe​rmals​ci.​2020.​106568 micro-channels heat sink. Int Commun Heat Mass Transfer
29. Hassan MA, Kassem MA, Kaood A (2022) Numerical 122:105158. https://​doi.​org/​10.​1016/j.​ichea​tmass​trans​fer.​2021.​
investigation and multi-criteria optimization of the thermal– 105158
hydraulic characteristics of turbulent flow in conical tubes fitted 42. Wang S, Chen HH, Chen CL (2019) Enhanced flow boiling in
with twisted tape insert. J Therm Anal Calorim 147:6847–6868. silicon nanowire-coated manifold microchannels. Appl Therm
https://​doi.​org/​10.​1007/​s10973-​021-​10998-7 Eng 148:1043–1057. https://​doi.​org/​10.​1016/j.​applt​herma​leng.​
30. He Z, Yan Y, Zhang L (2022) Thermal-hydraulic investigation 2018.​11.​125
on micro heat sinks with ribbed pin-fin arrays and single heating 43. Ghelich R, Jahannama MR, Abdizadeh H, Torknik FS, Vaezi
input: parametrical study. J Therm Anal Calorim 147:6489–6505. MR (2019) Central composite design (CCD)-Response surface
https://​doi.​org/​10.​1007/​s10973-​021-​10977-y methodology (RSM) of effective electrospinning parameters on
31. Le HS, Galal AM, Alhamrouni I, Aly AA, Abbas M, Saidi PVP-B-Hf hybrid nanofibrous composites for synthesis of HfB2-
AS, Truong TH, Dahari M, Wae-hayee M (2022) Heat transfer based composite nanofibers. Compos B Eng 166:527–541. https://​
efficiency optimization of a multi-nozzle micro-channel heat sink doi.​org/​10.​1016/j.​compo​sitesb.​2019.​01.​094
utilizing response surface methodology. Case Stud Thermal Eng 44. Hone J (2007) Carbon nanotubes: thermal properties. Dekker
37:102266. https://​doi.​org/​10.​1016/j.​csite.​2022.​102266 Encyclopedia Nanosci Nanotechnol 10:987–1000. https://​doi.​
32. Mahian O, Kolsi L, Amani M, Estellé P, Ahmadi G, Kleinstreuer org/​10.​1081/E-​ENN
C, Marshall JS, Siavashi M, Taylor RA, Niazmand H, Wongwises 45. Brinkman HC (1952) The viscosity of concentrated suspensions
S, Hayat T, Kolanjiyil A, Kasaeian A, Pop I (2019) Recent and solutions. J Chem Phys 20:571. https://​doi.​org/​10.​1063/1.​
advances in modeling and simulation of nanofluid flows-part I: 17004​93
fundamentals and theory. Phys Rep 790:1–48. https://​doi.​org/​10.​ 46. Pourfattah F, Sabzpooshani M (2020) Thermal management of a
1016/J.​PHYSR​EP.​2018.​11.​004 power electronic module employing a novel multi-micro nozzle
33. Debnath P, Pradhan M (2022) A recent state of art review on liquid-based cooling system: a numerical study. Int J Heat Mass
heat transfer enhanced characteristics and material selection of Transfer 147:118928. https://d​ oi.o​ rg/1​ 0.1​ 016/j.i​ jheat​ masst​ ransf​ er.​
SCTHX. Proc Instit Mech Eng Part E J Process Mech Eng. https://​ 2019.​118928
doi.​org/​10.​1177/​09544​08922​11244​63 47. Wu R, Hong T, Cheng Q, Zou H, Fan Y, Luo X (2019) Thermal
34. Ahmadi AA, Arabbeiki M, Ali HM, Goodarzi M, Safaei MR modeling and comparative analysis of jet impingement liquid
(2020) Configuration and optimization of a minichannel using

13
178 Page 20 of 20 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2023) 45:178

cooling for high power electronics. Int J Heat Mass Transf conditioning system based on response surface method and
137:42–51. https://​doi.​org/​10.​1016/j.​ijhea​tmass​trans​fer.​2019.​03.​ NSGA-II. Renew Energ 147:249–264. https://​doi.​org/​10.​1016/j.​
112 renene.​2019.​08.​109
48. Diao YH, Li CZ, Zhang J, Zhao YH, Kang YM (2017) 53. Pasandideh SHR, Niaki STA (2006) Multi-response simulation
Experimental investigation of MWCNT–water nanofluids optimization using genetic algorithm within desirability function
flow and convective heat transfer characteristics in multiport framework. Appl Math Comput 175(1):366–382. https://​doi.​org/​
minichannels with smooth/micro-fin surface. Powder Technol 10.​1016/j.​amc.​2005.​07.​023
305:206–216. https://​doi.​org/​10.​1016/j.​powtec.​2016.​10.​011 54. Derringer G, Suich R (1980) Simultaneous optimization of several
49. Esfe MH, Mahian O, Hajmohammad MH, Wongwises S (2018) response variables. J Qual Technol 12:214–219. https://​doi.​org/​
Design of a heat exchanger working with organic nanofluids 10.​1080/​00224​065.​1980.​11980​968
using multi-objective particle swarm optimization algorithm and 55. Shirvan KM, Ellahi R, Mirzakhanlari S, Mamourian M (2016)
response surface method. Int J Heat Mass Transfer 119:922–930. Enhancement of heat transfer and heat exchanger effectiveness in
https://​doi.​org/​10.​1016/j.​ijhea​tmass​trans​fer.​2017.​12.​009 a double pipe heat exchanger filled with porous media: numerical
50. Shirvan KM, Mirzakhanlari S, Mamourian M, Kalogirou SA simulation and sensitivity analysis of turbulent fluid flow. Appl
(2017) Optimization of effective parameters on solar updraft tower Thermal Eng 109:761–774. https://​doi.​org/​10.​1016/j.​applt​herma​
to achieve potential maximum power output: a sensitivity analysis leng.​2016.​08.​116
and numerical simulation. Appl Energ 195:725–737. https://​doi.​
org/​10.​1016/j.​apene​rgy.​2017.​03.​057 Publisher's Note Springer Nature remains neutral with regard to
51. Ighose BO, Adeleke IA, Damos M, Junaid HA, Okpalaeke KE, jurisdictional claims in published maps and institutional affiliations.
Betiku E (2017) Optimization of biodiesel production from
thevetia peruviana seed oil by adaptive neuro-fuzzy inference Springer Nature or its licensor (e.g. a society or other partner) holds
system coupled with genetic algorithm and response surface exclusive rights to this article under a publishing agreement with the
methodology. Energ Conv Manag 132:231–240. https://​doi.​org/​ author(s) or other rightsholder(s); author self-archiving of the accepted
10.​1016/j.​encon​man.​2016.​11.​030 manuscript version of this article is solely governed by the terms of
52. Xie Y, Hu P, Zhu N, Lei F, Xing L, Xu L (2020) Collaborative such publishing agreement and applicable law.
optimization of ground source heat pump-radiant ceiling air

13