Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

International Communications in Heat and Mass Transfer 44 (2013) 157–164

Contents lists available at SciVerse ScienceDirect

International Communications in Heat and Mass Transfer


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

CFD modeling (comparing single and two-phase approaches) on thermal


performance of Al2o3/water nanofluid in mini-channel heat sink☆
Mostafa Keshavarz Moraveji ⁎, Reza Mohammadi Ardehali
Department of Chemical Engineering, Faculty of Engineering, Arak University, P.O. Box 38156-8-8349, Arak, Iran

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

Available online 5 March 2013 CFD modeling of laminar forced convection on Al2O3 nanofluid with size particles equal to 33 nm and particle
concentrations of 0.5, 1 and 6 wt.% within 130 b Re b 1600 in mini-channel heat sink is executed by four indi-
Keywords: vidual models (single phase, VOF, mixture, Eulerian). Three-dimensional steady-state governing partial dif-
Laminar ferential equations was discretized using finite volume method.
Mini-channel Influences of some important parameters such as nanoparticle concentration and Reynolds number on the
Two-phase models
enhancement of nanofluid heat transfer have been investigated. The difference between the two-phase
Nanofluid
Heat transfer
models results was marginal, and they were more precise by comparison with experimental reference data
CFD than single phase model. Besides with regard to the most precise and less CPU usage and run time, mixture
model was chosen to obtain a correlation based on dimensionless numbers for the Nusselt number and
friction factor estimation.
© 2013 Elsevier Ltd. All rights reserved.

1. Introduction The most common base fluids used throughout the preparation of
nanofluids are water, ethylene glycol and engine oil, etc. Adding
Along with the decreasing of electronic component dimensions nanoparticles to abase fluid even at very low concentration has signif-
and increasing of heat generation by these devices, the problem of re- icant improvements in thermal performance. Chein and Huang
moving heat from them and achieving a successful design for maxi- inspected the performance of the silicon micro channel heat sink by
mum cooling has become drastically important. Then, in order to using Cu–water nanofluid. Their results show that using nanofluid
overcome the challenge of keeping electronic equipment at their has greatly improved the performance of the heat sink [4].
best performance, finding new ways of thermal load managing and Jung did experiments for Al2O3–water nanofluids in rectangular
performing optimization processes is inevitable [1]. microchannels. The particle diameter in their experiments was
Many innovative ideas have been proposed for cooling the elec- 170 nm. With only 1.8% of volume concentration they reported a
tronic devices. Water as the working fluid inside a microchannel 32% increase of the heat transfer coefficient in comparison to single
heat sink was used by Tuckerman and Pease [2]. distilled water [5].
Incapability of conventional fluids such as water in critical heat Also, experiments on nanofluid heat transfer in trapezoidal silicon
flux situations was compensated with the creation of “Nanofluid” micro channels have been performed by Wu. For channels with a hy-
conception. The innovative idea of adding metallic and non-metallic draulic diameter of194.5 μm and an Al2O3–water nanofluid, they
nano powders to a base fluid was proposed first by Choi, showing a reported an increase in the Nusselt number with increasing particle
number of potential advantages, such as increase in heat transfer concentration, Reynolds and Prandtl numbers, while the pressure
and reduction of heat transfer system size [1]. drop increased slightly when compared to pure water [5].
To explore the fundamental physical mechanisms of fluid flow and Ijam and Saidur [6] have studied the cooling of the minichannel
heat transfer in micro tube, many effects, including the size effect, rar- heat sink with turbulent flow regime using SiC–water and TiO2–
efaction effect, surface roughness, viscous effect, axial heat conduc- water nanofluids. Their investigation showed that using nanofluids
tion in the channel wall, surface geometry, the measurement errors, as a working fluid enhances the cooling by 7.25%–12.43% for SiC–
etc. should be taken into account [3]. water and 7.63%–12.77% for TiO2–water.
Many different types of nanoparticles have been used for nano- In the numerical simulation process of the heat transfer and hy-
fluid preparation such as Sic, Tic, Ag, Au, Cu, Al2O3, CuO, and TiO2. drodynamic behavior, two groups of studies are available in the liter-
ature. One approach is to assume that the solid suspended particles
☆ Communicated by Dr. W.J. Minkowycz.
are in thermal equilibrium with the fluid phase and a relative velocity
⁎ Corresponding author. of zero exists between solid and fluid parts implying that the particles
E-mail address: m-moraveji@araku.ac.ir (M.K. Moraveji). move with the same velocity of the base fluid. The authors use the

0735-1933/$ – see front matter © 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.icheatmasstransfer.2013.02.012
158 M.K. Moraveji, R.M. Ardehali / International Communications in Heat and Mass Transfer 44 (2013) 157–164

Different two-phase models also sometimes are compared to


Nomenclature
know the most precise model to be used in calculations. In an inter-
esting research accomplished by Lotfi et al. different approaches in-
A heat transfer surface, m 2
cluding single-phase, mixture and Eulerian two-phase models were
a acceleration, m s −2
compared while studying the forced convection of a nanofluid in a
Cp specific heat at constant pressure, J kg −1 K −1
horizontal tube. It was indicated that the mixture model is more accu-
Cd drag coefficient
rate compared to the other two models [7].
dp nano-particles diameter, m
The objective of the present paper is involved with the single
Dh hydraulic diameter, m
phase (homogeneous) and three different two-phase models (volume
F force, N
of fluid, mixture and Eulerian) to analyze laminar forced convection
Fcol particle–particle interaction force, Pa m −1
flow of a water–Al2O3 nanofluid in a mini-channel heat sink with uni-
Fd drag force, Pa m −1
form and constant heat flux applied at the base interface in Reynolds
Fvm virtual mass force, Pa m −1
number ranging from 133 to 1515. For each model, an individual com-
f friction factor
puter code was developed to satisfy this purpose.
fdrag drag function
The CFD results of each model compared with experimental reference
G particle–particle interaction modulus, Pa
data are reported by Ho [8]. Accuracy of each model is evaluated with ex-
g gravity acceleration, m s −2
perimental and compare with each other. Influences of some important
hv volumetric heat transfer coefficient, W m −3 K −1
parameters such as nanoparticle concentration and Reynolds number
hp liquid-particle heat transfer coefficient, W m −2 K −1
on nanofluid heat transfer enhancement are investigated. Based on the
h heat transfer coefficient, W m −2 K −1
most precise results and less CPU usage and run time two correlations
k thermal conductivity, W m −1 K −1
for prediction of average Nusselt number and friction factor are obtained.
l mean free path, m
Nu average Nusselt number
2. Model configuration
P pressure, Pa
Pr Prandtl number
2.1. Geometric configuration
Re Reynolds number
T temperature, K
In this investigation, 3D computational fluid dynamics (CFD) was
V velocity, m s −1
considered for our modeling. Simulation was performed based on single
phase (homogenous) and three two-phase models (VOF, Eulerian, and
mixture model). For each approach an individual code was developed.
Greek letters The mini-channel heat sink consists of 10 parallel rectangular
α thermal diffusivity, m 2 s −1 mini-channels with a length of 50 mm and a cross-sectional area of
β friction coefficient, kg m −3 s −1 1 mm in width by 1.5 mm in height for each mini-channel. All
η viscosity, Pa s mini-channels are supposed to be identical in heat transfer and
μ fluid dynamic viscosity, kg m −1 s −1 hydrodynamic. Hence, one channel can be picked out as the representa-
ρ fluid density, kg m −3 tive for numerical modeling. The domain of our modeling is shown
φ volume fraction in Fig. 1. Details of dimensions are available in experimental work
reported by C.J. Ho [8]. Steady state laminar flow of Al2O3/water
nanofluid, with nanoparticle size of 33 nm, passes through the
Subscripts mini-channel heat sink.
b bulk
dr drift 2.2. Governing equations
eff effective
f base fluid 2.2.1. Single phase model
m mixture Nanofluid as a single phase was considered. Furthermore, the fluid
nf nanofluid phase was assumed continuous. Governing equation continuity, mo-
p solid particle mentum and energy respectively are given [9]:
w wall Continuity equation:

∇:ðρnf :V m Þ ¼ 0: ð1Þ
justification that although a nanofluid consists of a solid phase, due to
the ultra-small size of particles, it is fluidized and as a result could be Momentum equation:
considered as a single-phase fluid. However, according to the findings
∇:ðρnf :V m :V M Þ ¼ −∇P þ ∇:ðμ nf :∇V m Þ: ð2Þ
of Ding and Wen this assumption may not always remain true for a
nanofluid [7].
Among the literature available on CFD simulation for nanofluids, Thermally isolated

some have focused on considering a two-phase flow for the used


1 mm
nanofluid indicating that the slip velocity between particles and
the fluid might not be zero. In a numerical study by Behzadmehr et
al. for the first time a two-phase mixture model was implemented Flow outlet
1.5 mm

Flow inlet 50 mm

to investigate the behavior of Cu–water nanofluid in a tube and


the results were also compared with the previous works using a Y
single-phase approach. The authors claimed that the simulation is
Wall subjected to constant heat flux q″
done by assuming that base fluid and particles behave separately X Z
possessed results that are more precise compared to the previous
computational modeling [7]. Fig. 1. Numerical domain of one channel.
M.K. Moraveji, R.M. Ardehali / International Communications in Heat and Mass Transfer 44 (2013) 157–164 159

Energy equation: Modeling by assuming that the coupling between phases is strong,
and particles closely follow the flow. The two phases are assumed to
∇:ðρnf :C:V m :T Þ ¼ ∇:ðknf :∇T Þ: ð3Þ be interpenetrating, meaning that each phase has its own velocity
vector field, and within any control volume there is a volume fraction
In this model, suitable definitions of thermo physical property rela- of primary phase and also a volume fraction of the secondary phase.
tions for nanofluid have great significance. The following formulas are Instead of utilizing the governing equations of each phase separately,
used to calculate the thermal and physical properties of the nanofluid the continuity, momentum and fluid energy equations for the mixture
[10]: are employed. Therefore, the steady state governing equations de-
scribing a mixture fluid flow and the heat transfer in mini-channel
ρnf ¼ ð1−φÞρf þ φρCuo ð4Þ are presented as follows [13]:
Continuity equation:
C nf ¼ ð1−φÞC f þ φC Cuo ð5Þ
∇:ðρnf :V m Þ ¼ 0: ð8Þ
where ρnf is density and Cnf is specific heat of the nanofluid.
-Viscosity of the nanofluid is estimated as:
Momentum equation:
 
1
μ nf ¼ μ f ð6Þ ∇:ðρm :V m :V M Þ ¼ −∇P þ ∇:ðμ m :∇V m Þ ð9Þ
ð1 þ φÞ0:25  n 
þ∇: ∑ ϕk ρk V dr;k V dr;k −ρm;i βm g ðT−T i Þ
  k¼1
kp þ ðn−1Þkf −ðn−1Þφ kf −kp
knf ¼   kf ð7Þ
kp þ ðn−1Þkf þ φ kf −kp ρm is the mixture density:

ρm ¼ ð1−φÞρf þ φρp ð10Þ


where φ is the particle volume fraction and the subscripts p, f, nf
correspond to particle, fluid and nanofluid. The spherical particles where φ is the volume fraction of solid or liquid phase.
are assumed for the nanoparticles with n = 3. Fluid energy equation:

2.2.2. Two-phase model n  


There are two general approaches for modeling the flow of solid liq- ∇: ∑ ρk :C pk :ϕk: V k :T ¼ ∇:ðkm :∇T Þ: ð11Þ
k¼1
uid mixtures. For low solid volume fractions, the most suitable approach
is the Lagrangian–Eulerian which analyzes the base fluid by the Eulerian
Volume fraction equation:
assumption and the particle phase by the Lagrangian one. For higher
solid volume fractions, the appropriate approach is the Euleriane–    
Eulerian. For nanofluids, the number of particles in the computational ∇: ϕp ρp V m ¼ −∇: ϕp ρp V dr;p : ð12Þ
domain, even for a very small particle volume fraction, is extremely
large due to the very small size of the particles. This makes it impossible Vm is mass average velocity:
to solve the nanofluid flow problems by the Lagrangian–Eulerian
approach due to limitations of the software abilities, memory and CPU ∑nk¼1 ϕk ρk V k
Vm ¼ : ð13Þ
requirements, etc. In fact, there are different Euleriane–Eulerian models. ρm
The most popular ones are the VOF (volume of fluid), mixture, and
Eulerian [11]. Vdr,k is the drift velocity for the secondary phase k, i.e. the nano-
particles in the present study. This is related to the relative velocity:
2.2.2.1. Mixture model. The essential consideration of the mixture
model is that only one setoff velocity component is solved from the n
ϕk ρk
differential equations for mixture momentum conservation. The ve- V dr;k ¼ V pf − ∑ V : ð14Þ
i¼1 ρm fk
locities of dispersed phases are inferred from the algebraic balance
equations. This reduces the computational effort considerably. In the The slip velocity (relative velocity) is defined as the velocity of a sec-
mixture model, the primary phase influences the secondary phase ondary phase (nanoparticle, p) relative to the velocity of the primary
via drag and turbulence, while the secondary phase in turn influence phase (water, f):
the primary phase via reduction in mean momentum and turbulence.
The mixture model is based on the following assumptions: V pf ¼ V p −V f ð15Þ
- A single pressure is shared by all phases.  
- The secondary dispersed phases are assumed to consist of spherical ρp dp
2 ρp −ρm
particles of uniform particle size being specified during calculations. V pf ¼ a ð16Þ
18μ f f drag ρp
- The interactions between different dispersed phases are neglected.
- The concentrations of the secondary dispersed phases are solved from (
scalar equations taking into account the correction due to phase slip. 1 þ 0:15Re0:687 ; Rep ≤1000
f drag ¼ f ðxÞ ¼ p : ð17Þ
0:0183Rep ; Rep > 1000
The mixture model has the following limitations and requirements:
- There is no possibility for phase change. The acceleration in Eq. (16) is:
- The flow compressibility is not accounted for.
- Pressure boundary condition cannot be specified because the ideal a ¼ g−ðV m :∇ÞV m ð18Þ
gas law cannot be employed.
- Turbulence generation in the secondary phases is not accounted where, g is gravitational acceleration.
for, nor is the turbulence of primary phase directly affected by It is noted that terms relating to the natural convection must be
the presence of secondary phase [12]. neglected, because our work is dealing with forced convection.
160 M.K. Moraveji, R.M. Ardehali / International Communications in Heat and Mass Transfer 44 (2013) 157–164

2.2.2.2. Eulerian model. In the Eulerian model there are different kinds where, hp is the fluid-particle heat transfer coefficient that should be
of coupling between phases. The pressure is shared by all the phases, calculated from empirical correlations. In the present study the fluid-
while separate continuity, momentum, and energy equations are particle heat transfer coefficient is calculated using this equation:
employed for different phases including primary and secondary
phases. The volume of each phase is calculated by integrating its hp dp 0:6 1=3
Nup ¼ ¼ 2 þ 1:1Rep Pr : ð31Þ
volume fraction throughout the domain, while the summation of all kl
the volume fractions is equal to unity [14].
The governing mass, momentum and energy equations for the The effective thermal conductivities for liquid and particle phases
particle and base liquid phases can be written as follows [15]: are estimated as:
  kb;l

∇: ρl φl V l ¼ 0 ð19Þ keff;l ¼ ð32Þ
φl
 
→ kb;p
∇: ρp φp V p ¼ 0 ð20Þ keff;p ¼ ð33Þ
φp

φl þ φp ¼ 1 ð21Þ where:
      qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
→→ → →
∇: ρl φl V l V l ¼ −φl ∇p þ ∇: φl μ l ∇V l þ ∇V l T þ F d þ F vm ð22Þ kb;l ¼ 1− ð1−φl Þ kl ð34Þ

     qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
→ → → → kb;p ¼ ð1−φl ÞðωA þ ½1−ωΓ Þkl ð35Þ
∇: ρp φp V p V p ¼ −φp ∇p þ ∇: φp μ p ∇V p þ ∇V p T −F d
þ F Vm þ F col : ð23Þ and
(   )
2 BðA−1Þ A ðB−1Þ B þ 1
Due to the very small size of the particles, the lift force is neglected Γ¼

ln −
− ð36Þ
1− AB A 1− AB 2 B 1− AB 2
in our study. Only the drag force between the phases is considered.
The drag force between the phases is calculated as:
with
   
→ → ½1−φl  10=9
F d ¼ −β V l −V p : ð24Þ B ¼ 1:25 : ð37Þ
φl

The friction coefficient β is calculated according to the nanoparticle For spherical particles we have:
volume concentration range. For a very dilute two-phase flow with par-
ticle diameter dp, the friction coefficient is: kp −3
A¼ and ω ¼ 7:26  10 : ð38Þ
kl
3 φ ð1−φl Þ → →  −2:65
β ¼ Cd l ρl V l −V p φl : ð25Þ
4 dp 2.2.2.3. VOF model. The VOF model solves a single set of momentum
equations for all the phases and tracks their volume fraction all over
Eq. (25) is valid for two-phase flows with φl > 0.8 and Cd is the the domain of study by solving a continuity equation for the secondary
drag coefficient and its magnitude depends on the particle Reynolds phases. The total summation of the volume fractions for all the phases
number: is equal to unity. Therefore, the magnitude of the primary phase
volume fraction is calculated. With this method all of the physical
8  
< 24 1 þ 0:15Re0:687 properties are calculated by taking a weighted average of different
p Rep b1000
C d ¼ Rep ð26Þ phases based on their volume fraction throughout each control volume.
:
0:44 Rep ≥1000 The single set of momentum equations is solved to find the velocity
components which are shared by all the phases. In the same manner,
where a shared temperature is calculated from a single energy equation.
→ →  Specifically, mass conservation is expressed as [11]:
   
φl ρl V l −V p dp →
Rep ¼ : ð27Þ ∇: φq ρq V q ¼ 0 ð39Þ
μl

n n
Considering the base liquid and the nanoparticle phases as incom-
where ∑ φq ¼ 1 and all properties are calculated like N ¼ ∑ φq N q (n is
pressible fluids, and neglecting the viscous dissipation and radiation, q¼1 q¼1
the energy equations are written as: number of the phases).
  Also the conservation of momentum and energy equations are similar
→    
∇: ρl φl C pl T l V l ¼ ∇: φl keff;l ∇T l −hv T l −T p ð28Þ as Eqs. (2) and (3).

  2.3. Boundary condition definition


→    
∇: ρp φp C pp T p V p ¼ ∇: φp keff;p ∇T p þ hv T l −T p : ð29Þ
Non-linear governing equations have been subjected to the fol-
lowing boundary conditions:
For mono-dispersed spherical particles, hv can be calculated from:
2.3.1. Inlet
6ð1−φl Þ At the entrance of the mini-channel heat sink assembly velocity
hv ¼ hp ð30Þ
dp inlet boundary condition depending on the value of Reynolds number,
M.K. Moraveji, R.M. Ardehali / International Communications in Heat and Mass Transfer 44 (2013) 157–164 161

Fig. 2. Model grid independency test, the average Nusselt number versus Reynolds number for pure water with single phase and two-phase approaches ((a) single phase, (b) mixture
model, (c) Eulerian model, d: VOF model).

that was extracted from reference experimental [8],was applied convective and diffusive terms a first order upwind method was
(z = 0): used. The algebraic system resulting from numerical discretization
was solved by using Gauss–Seidel applied in a line going through all
V x ¼ V y ¼ 0; V z ¼ V 0 T ¼ T 0 ¼ 36:2 C:
o
ð40Þ volumes in the computational domain.
During the iterative process, the absolute residuals were carefully
monitored and convergence criteria for every parameter (mass,
2.3.2. Outlet velocity and energy) was restricted to be lower than10 −5.
At the mini-channel outlet plain, the static pressure with no viscous For more accuracy of the results grid independence examinations
stress has been assumed (z= Lch): were done on pure water individually in each model. For all cases a
structured non-uniform grid has been considered spotting for our
P ¼ atmospheric pressure: ð41Þ modeling domain. It is finer vicinity the bottom wall and entrance
of mini-channel where the temperature and velocity gradients are
2.3.3. Walls large. Several combinations of node numbers in the length, width
The no-slip boundary condition was enforced on all solid walls: and height directions were examined. The grid which has been
employed for the present study consisted of 600 × 18 × 12 nodes
V x ¼ Vy ¼ Vz ¼ 0: ð42Þ (600 grids in length, 18 grids in height, and 12 grids in width direc-
tion). Also fining of grids did not make a significant influence on the
First heat was removed by conduction through the solid and then accuracy of Nusselt number.
dissipated away by force convection of the cooling fluid in the mini- The grid independence examination of the numerical solution is
channel. The bottom surface was uniformly heated with constant heat illustrated in Fig. 2, where the Nusselt number versus Reynolds num-
flux and adiabatic boundary conditions were applied at all other sides ber for the several combinations of nod numbers is displayed.
of the mini-channel:
2.5. Calculation of the Nusselt number and the pumping power
}
−knf :∇T ¼ q ð@Bottom of mini−channelÞ ð43Þ
From the collecting data of temperature, volume flow rate and
−knf :∇T ¼ 0ð@Top; front and rearof mini−channelÞ: ð44Þ pressure drop, the average Nusselt number and pumping power or
FOMitd can be calculated from the following formulas:

2.4. Numerical implementation and grid-independence study 4W ch H ch


Dh ¼ ð45Þ
2ðW ch þ H ch Þ
The set of 3D coupled non-linear differential equations has been  
discretized by the control volume technique. For the treatment of C p :μ
Pr ¼ ð46Þ
pressure–velocity coupling SIMPLE algorithm was adopted. For the k nf
162 M.K. Moraveji, R.M. Ardehali / International Communications in Heat and Mass Transfer 44 (2013) 157–164

Fig. 3. Nu deviation from experimental in different Re for (a) φ = 0.139, (b) φ = 0.278
and (c) φ = 1.734%.
Fig. 4. Average heat transfer coefficient versus nanoparticle volume fraction in three
Reynolds number ((a) Re = 133, (b) Re = 650 and (c) Re = 1520) with four models.

ρnf :U m :Dh .
Re ¼ ð47Þ h itd;m
μ nf
h itd;bf
FOMitd ¼  1 ð53Þ
P 3
2:Dh :ΔP m
P bf
f ¼ ð48Þ
ρnf Lch u2
where Dh, Wch, Hch, Lch, Q_ , Δp, P, h itd , Nu itd and FOMitd are hydraulic
P ¼ Δp:Q_ ð49Þ diameter, mini-channel width, height, length, volume flow rate,
total pressure drop, pumping power, average heat transfer coefficient,
qf ¼ ρ:Q_ :C p ðT out −T in Þ ð50Þ average Nusselt number based on the temperature difference be-
tween the average base wall and the inlet fluid and figure of merit,
respectively.
qf q}
h itd ¼ or h itd ¼ ð51Þ
ðT w −T in ÞAbase ðT w −T in Þ 3. Code validation, results and discussion

Nu itd ¼ h itd :Dh Mentioned geometry configuration and boundary conditions are
ð52Þ
k same as experimental work reported by C.J. Ho [8].
M.K. Moraveji, R.M. Ardehali / International Communications in Heat and Mass Transfer 44 (2013) 157–164 163

In order to indicate the validity and precision of the computer


codes, average Nusselt number Nu itd , based on the temperature dif-
ference between the average base wall and the inlet fluid, and figure
of merit (FOMitd) were extracted from published experimental work
[8]. Two mentioned parameters ( Nu itd and FOMitd) for pure water
and three particle fraction (.5, 1 and 6 wt.% equal to volume fraction,
respectively 0.139, 0.278 and 1.734%) were used to validate our
modeling, hydrodynamically and thermally. Fig. 3 indicates compari-
son between the experimental and calculated Nusselt number versus
Reynolds number for the mentioned volume concentrations by single
phase and three two-phase models. As we can see there is a good
agreement between present modeling results and the experimental
one. In the lower volume concentration (φ = .139 and .278%) the dif-
ferences between single phase model and two phase model aren't
sensible (especially in φ = .139% that are under 10%). With increasing
in φ, modeling results with two phase models acquire over estimate
while single phase prediction remains under estimate. Undoubtedly
Fig. 6. Comparison correlated Nu with simulated one.
deviation from experimental with two phase models is lower than
single phase in all cases but this deviation for single phase approach
is becoming more sensible when volume concentration goes up to
φ = 1% or to a higher Reynolds number. Maximum deviation from Above correlation is compared with the calculating results and is
experimental with two-phase approaches was 16%. depicted in Fig. 6. The deviation of the correlated average Nusselt
The effect of nanoparticle volume fraction on the average heat number from the calculated is between + 3 and − 7%.
transfer coefficient in three Reynolds number (133, 650 and 1520) Fig. 7 shows the deviation of FOMitd from experimental for φ =
by four models is indicated in Fig. 4. 1.734%. It can be observed that the FOMitd of the nanofluid based on
It is apparent that two modeling results reflect either the qualita- the two-phase models agrees well with those of experimental data
tive or the quantitative excellent behavior with the experimental under the same Reynolds number. Also, the three two-phase models
values. The difference between the three two-phase models is very led to nearly the same hydrodynamic results. The maximum devia-
narrow. While taking into account less run time and CPU usage the tion of the two phase models is less than 10%.
mixture model is an excellent selection for our modeling. In order to use the nanofluids for industrial applications, it is nec-
Fig. 5 illustrates the variation of Nusselt number with Reynolds essary to find out the flow characteristic of the nanofluids in addition
number and volume fraction. It is recognizable augmenting Nusselt to the heat transfer evaluation.
number with the increasing Reynolds number and volume fraction. By using the simulation results, friction factor in the mini-channel
It is due to the fact that nanofluids with higher particle concentra- heat sink as a function of the Reynolds number, particle volume frac-
tions have higher static and dynamic thermal conductivities, and tion (φ), was correlated (130 ≤ Re ≤ 1530, 0 ≤ φ ≤ 0.01734):
higher Reynolds numbers lead to higher velocity and temperature
gradients at the mini-channel wall which in turn increase the heat
−0:90158 0:066694
transfer coefficient [16]. f ¼ 34:81713Re ð1 þ φÞ : ð55Þ
Because of the mixture model that has represented the most precise
consequences of all, a correlation for average Nusselt number was devel-
oped based on its results (130≤Re≤ 1530, 6.5 ≤Pr ≤7, 0≤φ≤ 0.01734): Fig. 8 represents the comparison of the correlated friction factor
obtained from the above equation with the modeling ones. The max-
0:391801 19:8223 imum deviation is less than 9%.
Nu ¼ 0:290263ðRe:PrÞ ð1 þ φÞ : ð54Þ

This correlation can be used for laminar flow of al2o3/water nano-


fluid in mentioned mini-channel heat sink.

Fig. 5. Results of variation Nu versus Re in various nanoparticle volume fraction with


mixture model. Fig. 7. Comparison FOMitd with experimental in different Re for φ=1.734% with four models.
164 M.K. Moraveji, R.M. Ardehali / International Communications in Heat and Mass Transfer 44 (2013) 157–164

References
[1] S.A. Fazeli, S.M.H. Hashemi, H. Zirakzadeh, M. Ashjaee, Experimental and numerical
investigation of heat transfer in a miniature heat sink utilizing silica nanofluid,
Superlattices and Microstructures 51 (2012) 247–264.
[2] D.B. Tuckerman, R. Pease, High-performance heat sinking for VLSI, Electron Device
Letters, IEEE 2 (5) (1981) 126–129.
[3] Z. Li, Y.L. He, G.H. Tang, W.Q. Tao, Experimental and numerical studies of liquid
flow and heat transfer in microtubes, International Journal of Heat and Mass
Transfer 50 (2007) 3447–3460.
[4] A. Ijam, R. Saidur, P. Ganesan, Cooling of minichannel heat sink using nanofluids,
International Communications in Heat and Mass Transfer 39 (2012) 1188–1194.
[5] M. Kalteh, A. Abbassi, M. Saffar-Avval, Jens Harting, Eulerian–Eulerian two-phase
numerical simulation of nanofluid laminar forced convection in a microchannel,
International Journal of Heat and Fluid Flow 32 (2011) 107–116.
[6] A. Ijam, R. Saidur, Nanofluid as a coolant for electronic devices (cooling of elec-
tronic devices), Applied Thermal Engineering 32 (2012) 76–82.
[7] A. Kamyar, R. Saidur, M. Hasanuzzaman, Application of computational fluid
dynamics (CFD) for nanofluids, International Journal of Heat and Mass Transfer
55 (2012) 4104–4115.
[8] C.J. Ho, W.C. Chen, An experimental study on thermal performance of Al2O3/water
nanofluid in a minichannel heat sink, Applied Thermal Engineering 50 (2013)
516–522.
[9] M.K. Moraveji, M. Darabi, S.M.H. Haddad, R. Davarnejad, Modeling of convective heat
transfer of a nanofluid in the developing region of tube flow with computational
fluid dynamics, International Communications in Heat and Mass Transfer 38
(2011) 1291–1295.
Fig. 8. Comparison of correlated friction factor with simulated one. [10] M.K. Moraveji, M. Hejazian, Modeling of turbulent forced convective heat transfer
and friction factor in a tube for Fe3o4 magnetic nanofluid with computational
fluid dynamics, International Communications in Heat and Mass Transfer 39
(2012) 1293–1296.
4. Conclusion [11] M. Akbari, N. Galanis, A. Behzadmehr, Comparative assessment of single and
two-phase models for numerical studies of nanofluid turbulent forced convection,
International Journal of Heat and Fluid Flow 37 (2012) 136–146.
A 3D laminar forced convection of Al2O3–water nanofluids inside [12] H.M. El-Batsh, M.A. Doheim, M.A. Doheim, A.F. Hassan, On the application of mixture
the mini-channel heat sink was studied using CFD tool. Four individual model for two-phase flow induced corrosion in a complex pipeline configuration,
computer codes (single phase, VOF, mixture, Eulerian) were developed. Applied Mathematical Modelling 36 (2012) 5686–5699.
[13] R.M. Moghari, A. Akbarinia, M. Shariat, F. Talebi, R. Laur, Two phase mixed convec-
Based on experimental data the validity of each model was investigated.
tion Al2O3–water nanofluid flow in an annulus, International Journal of Multiphase
The following conclusions can be drawn from the present assessment: Flow 37 (2011) 585–595.
[14] M. Akbari, N. Galanis, A. Behzadmehr, Comparative analysis of single and two-
- When we look at the two-phase model results, they are very close phase models for CFD studies of nanofluid heat transfer, International Journal of
to each other by contrast (in both aspect: hydrodynamic and heat Thermal Sciences 50 (2011) 1343–1354.
transfer). [15] M. Kalteh, A. Abbassi, M. Saffar-Avval, A. Frijns, A. Darhuber, J. Harting, Experi-
mental and numerical investigation of nanofluid forced convection inside a
- The two phase models represented a better approximation of the wide microchannel heat sink, Applied Thermal Engineering 36 (2012) 260–268.
experimental data comparing to the single phase model. [16] E. Ebrahimnia-Bajestan, H. Niazmand, W. Duangthongsuk, S. Wongwises, Numer-
- Increase in the Reynolds number (Re) and nanoparticle concentration ical investigation of effective parameters in convective heat transfer of nanofluids
flowing under a laminar flow regime, International Journal of Heat and Mass
led to Nusselt number enhancement. Transfer 54 (2011) 4376–4388.
- Regarding less run time and CPU usage the mixture model is a good
recommendation for this modeling.
- Based on mixture model results verifiable correlations for Nusselt
number and friction factor were obtained.

You might also like