International Journal of Heat and Mass Transfer 132 (2019) 443–461

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer

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

Fluid flow, heat transfer and entropy generation analyses of turbulent

forced convection through isotropic porous media using RANS models
Mehrdad Torabi a, Mohsen Torabi b,c,⇑, Mohammad Eftekhari Yazdi d, G.P. Peterson b
a
Young Researchers and Elite Club, Central Tehran Branch, Islamic Azad University, Tehran, Iran
b
The George W. Woodruff School of Mechanical Engineering, Georgia Institute of Technology, Atlanta, GA 30332, USA
c
School of Engineering, University of California, Merced, CA 95348, USA
d
Department of Mechanical Engineering, Islamic Azad University, Central Tehran Branch, Tehran, Iran

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

Article history: Turbulent fluid flow, heat transfer and entropy generation through isotropic porous media were investi-
Received 31 July 2018 gated utilizing two RANS models, i.e., RNG k-e and SST k-x. Circular, longitudinal elliptical and transverse
Received in revised form 30 November 2018 elliptical cross-sectional configurations were analyzed with symmetrical boundary conditions for the
Accepted 3 December 2018
upper and lower lines and periodic boundary conditions for the back and front lines. Evaluation of the
Available online 10 December 2018
temperature contours indicated two low-temperature regions behind the solid rods of the transverse
elliptical cross-sectional configuration exist. For a given specific normalized pressure gradient, dimen-
Keywords:
sionless turbulence kinetic energy, Nusselt number, and heat transfer efficiency, the RNG k-e model
Porous media
Turbulence
resulted in more reliable results for the dimensionless turbulence kinetic energy at low Reynolds num-
Entropy generation bers when compared with the SST k-x model. Using the thermal analysis of the three cross-sectional con-
Heat transfer figurations investigated, the longitudinal elliptical resulted in higher heat transfer efficiencies when
Cross-sectional configuration compared with the circular and transverse elliptical cross-sectional configurations. When the influence
of the turbulence effects were included in the entropy generation rate, a second law of thermodynamics
analysis indicated that the longitudinal elliptical cross-sectional configuration resulted in a lower entropy
generation rate when compared with the circular and transverse elliptical cross-sectional configurations.
Combining the results of the first and second law analyses for the three cross-sectional configurations, it
is apparent that isotropic porous media consisting of the longitudinal elliptical cross-sectional configura-
tion can simultaneously result in a high heat transfer efficiency and low entropy generation rate.
Ó 2018 Elsevier Ltd. All rights reserved.

1. Introduction results in additional thermophysical terms in the classical govern-

ing equations such as Darcy drag, Forchheimer drag, thermal dis-
Turbulent fluid flow and heat transfer through porous media persion, and interfacial heat transfer [3]. The VAT approach
have been the subject of a number of investigations over the past utilizes volume averaged governing equations, including the
several years, however several fundamental questions remain. To Darcy-Forchheimer-Brinkman momentum equation and volume
answer some of the more fundamental questions, it is necessary averaged energy equations. With respect to the porous media
to conduct additional theoretical and experimental investigations, investigations previously conducted using the VAT approach, some
particularly those related to practical applications [1]. To simulate scholars have examined the porous media in the presence of differ-
turbulent fluid flow and heat transfer through porous media, pre- ing thermophysical conditions, such as turbulent fluid flow and
vious investigations have employed several different macroscopic heat transfer through partially filled porous channels [4] and phys-
and microscopic approaches that capture the influence of the var- ical structures such as a composite annulus [5]. Other investigators
ious physical parameters. One of the well-known macroscopic have employed the VAT approach in order to examine practical
approaches identified and investigated by Whitaker [2], is the vol- applications of the porous media, such as shell-and-tube heat
ume averaging theory (VAT) [2]. Utilization of the VAT approach exchangers [6], transcritical hydrocarbon fuel flow [7] and river
flows with submerged vegetated floodplains [8]. As an alternative
method to the VAT methodology, pore scale modeling (PSM) of
⇑ Corresponding author. porous media provides precise information on the velocity and
E-mail addresses: torabimce@yahoo.com (M. Torabi), Mohsen.Torabi@my. temperature distributions as derived from the classical governing
cityu.edu.hk, mtorabi@ucmerced.edu (M. Torabi), Bud.Peterson@gatech.edu
(G.P. Peterson).
equations [9]. Given the significant computational resources

https://doi.org/10.1016/j.ijheatmasstransfer.2018.12.020
0017-9310/Ó 2018 Elsevier Ltd. All rights reserved.
444 M. Torabi et al. / International Journal of Heat and Mass Transfer 132 (2019) 443–461

Nomenclature

a elliptical rod short radius, m Tw wall temperature, K

A surface area, m2

u time averaged flow direction velocity, m
s1
Aexp : exposed surface area, m2 v
time averaged flow perpendicular direction velocity,
b elliptical rod long radius, m m
s1
1
Cp specific heat, J
kg
K1 u0 flow direction velocity fluctuation, m
s1
D exposed surface diameter, m v0 flow perpendicular direction velocity fluctuation, m
s1
F1 blending function in Eq. (15) uD Darcy velocity, m
s1
F2 damping function in Eq. (30)
B
u bulk velocity, m
s1
h convection heat transfer coefficient, W
m2
K1 x flow direction, m
H height of unit cell, m y flow perpendicular direction, m
k turbulence kinetic energy, m2
s2
hki f surface averaged turbulence kinetic energy, m2
s2 Greek Symbols
kf fluid conductivity, W
m1
K1 l fluid dynamic viscosity, kg
m1
s1
n local coordinate normal to the wall, m lturb: turbulent viscosity, kg
m1
s1
Ns dimensionless local entropy generation rate leff : effective dynamic viscosity, kg
m1
s1
Nt dimensionless surface averaged entropy generation rate ak inverse turbulent Prandtl number in Eq. (5)
Nu Nusselt number ae inverse turbulent Prandtl number in Eq. (6)


D
p time averaged pressure, N
m2 DT ml logarithmic mean temperature difference, K

i f
p surface averaged p
, N
m2 Dp
pressure difference, N
m2
P turbulence kinetic energy production for the RNG k-e x specific dissipation rate, s1
model, kg
m1
s3 xw wall specific dissipation rate, s1


P turbulence kinetic energy production for the SST k-x / porosity
model, kg
m1
s3 e turbulence dissipation rate, m2
s3
Prturb: turbulent Prandtl number m fluid kinematic viscosity, m2
s1
Q heat transfer rate, W m
^ effective dynamic viscosity to fluid dynamic viscosity
Re Reynolds number ratio
S strain rate magnitude, s1 q fluid density, kg
m3
S_ gen local entropy generation rate, W
m3
K1 rk turbulent Prandtl number in Eq. (14)
T
time averaged temperature, K re turbulent Prandtl number in Eq. (15)
T0 temperature fluctuation, K c heat transfer efficiency
T
B bulk temperature, K

require for this type of analysis [10], the PSM analysis of turbulent tor were developed. Galuppo and de Lemos [17] numerically ana-
fluid flow and heat transfer has only been performed for relatively lyzed turbulent fluid flow and heat transfer in backward-facing
simplified porous structures, such as packed beds [11] and metal steps utilizing porous inserts. The effects of porosity, the Darcy
foams [12]. number, and the thickness of the porous insert on the streamlines,
The numerical analysis of turbulent fluid flow and heat transfer dimensionless turbulence kinetic energy contours, skin friction
through porous media using the VAT approach has been previously coefficient, and Nu were examined and discussed. The conven-
discussed [13–15]. In an interesting investigation by Möbner and tional VAT approach utilizing a high Reynolds k-e turbulence
Radespiel [14], a new turbulence model for fluid flow and heat model was employed. Thermal boundary conditions for the top
transfer through porous media was presented. The generated and bottom walls of the channel were assumed to be adiabatic
parameters in this new turbulence model revealed good agreement and constant heat flux, respectively. It was found that the turbu-
when compared with direct numerical simulation (DNS) results. lence characteristics of flow is damped with the insertion of the
The results included the effects of the turbulent jump coefficient porous material and the thickness of the porous insert was shown
and the jump coefficient of turbulent dissipation rates, on the tur- to have a significant impact on the overall thermophysical proper-
bulence dissipation rate and modified Reynolds stresses [14]. A ties of the system.
numerical investigation of turbulent fluid flow and heat transfer Alternatively, some investigations have utilized the PSM
in heat exchangers filled with porous materials was performed approach in order to simulate turbulent fluid flow and heat trans-
by Shirvan et al. [15]. The response surface methodology procedure fer through isotropic porous media [18–20]. Yang et al. [19] inves-
was utilized in order to calculate correlations of the heat exchanger tigated three-dimensional turbulent flow and thermal fields for
effectiveness and Nusselt number (Nu). Conclusions revealed that packed beds consisting of spheres and ellipsoids in different pack-
the Reynolds number (Re) and thickness of the dimensionless por- ing arrangements. In order to include the turbulence effects in gov-
ous substrate had a significant influence on the Nu and the heat erning equations, the renormalization group (RNG) k-e model was
exchanger effectiveness [15]. In an interesting investigation by utilized. For similar physical conditions, the sphere packing and
Saberinejad and Keshavarz [16], oscillating turbulent fluid flow flat ellipsoidal packing arrangements demonstrated the highest
and heat transfer through porous media embedded in a circular and lowest overall heat transfer efficiencies, respectively. In a com-
tube were analyzed. Utilizing the VAT approach and assuming a bined numerical and experimental investigation, turbulent fluid
local thermal equilibrium (LTE) model for temperature distribution flow and heat transfer through flat and slotted tube bundles using
in the porous media, the unsteady momentum and energy govern- a low Reynolds turbulence model was analyzed by Kong et al. [20].
ing equations were numerically solved. In order to investigate heat Comprehensive results including the pressure drop, heat flow rate,
transfer coefficient and pressure loss in thermal-fluid systems with Nu, performance evaluation criterion, and system performance cri-
porous media, such as the cooling processes in porous heat terion were all evaluated and discussed. With respect to the perfor-
exchangers, two correlations for the Nu and maximum friction fac- mance evaluation and system performance criterion investigated,
 M. Torabi et al. / International Journal of Heat and Mass Transfer 132 (2019) 443–461 445

equal longitudinal and transverse tube pitches revealed superior lence modeling, heat transfer discussion, and boundary conditions
results from both the theoretical and practical engineering per- for the isotropic porous media are all provided. Section 3 discusses
spectives. Further, in some investigations more advanced turbu- the entropy generation formulation in turbulent fluid flow and
lence simulation approaches have been utilized [21]. Recently, heat transfer. Section 4 addresses the use of an applied numerical
Kundu et al. [22] examined various turbulence models for isotropic method and the resulting grid independence. In Section 5 the code
porous media consisting of circular and square rods. The standard validation investigation and results such as dimensionless pressure
k-e model, the low Reynolds k-e model and large eddy simulation gradient, dimensionless turbulence kinetic energy, heat transfer
(LES) model, were all considered. In order to model the porous efficiency and dimensionless surface averaged entropy generation
structures accurately, the periodic boundary conditions for three- rate are explored and evaluated. Finally, Section 6 presents some
dimensional circular and square cylinders were utilized. The low concluding remarks regarding the results obtained.
Reynolds k-e model revealed results that were much closer to the
results of the LES model when compared with the standard k-e 2. Problem statement
model [22].
In a comprehensive investigation by Uth et al. [23], the DNS If forced convection of air at 300 K with constant thermophysi-
method was utilized to explore the possibility of macroscopic tur- W  
cal properties, i.e., kf ¼ 0:0242 m
K , l ¼ 1:7894  105 m
s
kg
,
bulence in isotropic porous media that consisted of two-    
dimensional and three-dimensional matrices. Two different cp ¼ 1004:63 kg
K , q ¼ 1:225 m3 , flowing through a heat exchan-
J kg

numerical methods, namely the finite volume and lattice Boltz-

ger consisting of solid rods at 200 K, as shown in Fig. 1a is consid-
mann methods, were employed. To reduce the computational time,
ered, the thermal interaction can be analyzed [27]. Because the
periodic boundary conditions were implemented in the physical
solid rods are repeated in the fluid flow direction, a single unit cell
domain for each individual simulation. The investigation revealed
can be used, utilizing periodic boundary conditions in the fluid
that the size of the turbulent eddies are limited to the pore size
flow direction, as illustrated in Fig. 1b [22]. In addition, in order
of the isotropic porous media, i.e., the pore scale prevalence
to examine different structures of porous media, the cross-
hypothesis is valid [23].
sectional configuration can be modified to longitudinal elliptical
With respect to the above-mentioned discussion, it is apparent
and transverse elliptical, as illustrated in Fig. 2. The height of the
that there is only very limited information that addresses the eval-
unit cell (H) was set to H = 0.1 m as shown in Figs. 1 and 2. The cir-
uation of the different RANS models in turbulent fluid flow and
cular rod diameters were D ¼ 0:03 and D ¼ 0:05 m for / ¼ 0:8 and
heat transfer through isotropic porous media using the PSM
/ ¼ 0:6, respectively [28]. The elliptical cross-sectional configura-
approach. Although a limited number of investigations have been
tions were produced using the porosity equation of / ¼ 1  pHab 2
conducted using different turbulence models [22,24], the heat
transfer and entropy generation through isotropic porous media and a straightforward relationship between the two radii was
has not yet been fully analyzed. Therefore, in the current investiga- defined as ba ¼ 53 [29]. In addition, the porosity values 0.7 and 0.85
tion turbulent fluid flow, heat transfer and entropy generation uti- for the longitudinal elliptical cross-sectional configuration and
lizing the shear stress transport (SST) k-x model and RNG k-e 0.7 and 0.8 for the transverse elliptical cross-sectional configura-
model through isotropic porous media consisting of three cross- tion were used.
sectional configurations, namely circular, longitudinal elliptical Analyses of the thermophysical properties in isotropic porous
and transverse elliptical, are examined, evaluated and discussed. media have been previously provided in laminar regime [30,31],
Further, even though turbulent fluid flow through isotropic porous the current investigation focuses on turbulent fluid flow, heat
media has been analyzed from a first law of thermodynamics per- transfer and entropy generation through similar porous systems.
spective [15,16,20], the second law analysis of turbulent fluid flow To compensate for the effects of turbulence in porous media, differ-
through isotropic porous media has only rarely been discussed ent values for the Re have been suggested [32]. Following an exper-
[25]. An extensive literature search indicated that the numerical imental investigation performed by Kececioglu and Jiang [33], in
investigation by Chihk and Allouache [26], is perhaps the only which chaotic regimes including turbulence initiated at Re < 100,
numerical second law analysis of turbulent fluid flow through par- the Re varied between 200 and 200,000. Different Reynolds num-
tially and/or fully filled porous media. In order to investigate and bers and cross-sectional configurations were considered in order
understand the effects of different thermophysical properties,
including porous layer thickness, permeability and effective ther-
mal conductivity, on the entropy generation rate, the VAT
approach and a modified k-e model was utilized. The investigation
considered the turbulent entropy generation rate and turbulent
Bejan number (Be), and the minimum entropy generation rate in
the porous annulus was provided. Given a high effective thermal
conductivity for porous materials, the optimum thickness of the
porous layer was determined in order to minimize the entropy
generation rate. Even given the utilization of the VAT approach in
the current investigation [26], there is still a lack of information
regarding the second law analysis of turbulent fluid flow and heat
transfer through isotropic porous media using the PSM approach.
Thus, in order to achieve the cross-sectional configuration that
minimizes the entropy generation rate, i.e., minimization of exergy
destruction, the current investigation also examines turbulent fluid
flow and heat transfer through isotropic porous media from a sec-
ond law of thermodynamics perspective.
The current investigation has been divided into five sections. In Fig. 1. Geometric model: (a) heat exchanger bundles with circular cross-sectional
Section 2 the physical considerations, governing equations, turbu- configuration, (b) unit cell with periodic and symmetry boundaries.
446 M. Torabi et al. / International Journal of Heat and Mass Transfer 132 (2019) 443–461

  !
@k @k   @2k @2k
q u
þ v
¼ P-e þ ak leff : þ ð5Þ
@x @y @x2 @y2

  !
@e @e e e2   @2e @2e
q u
þ v
¼ C 1e P  C 2e þ ae leff : þ ð6Þ
@x @y k k @x2 @y2

In the RNG k-e model the effective dynamic viscosity (leff : ) can
be described as:
leff : ¼ l þ lturb: ð7Þ

The leff : is calculated as:

 2 
qk m^
d pffiffiffiffiffiffi ¼ 1:72 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi dm
^ ð8Þ
el m^3  1 þ C m
where m
^ is defined as:
leff :
m^ ¼ ð9Þ
Fig. 2. Geometric models: (a) longitudinal elliptical cross-sectional configuration,
l
The turbulence kinetic energy production for the RNG k-e model
(b) transverse elliptical cross-sectional configuration.

(P) in Eqs. (5) and (6) can be written as:

to evaluate the RNG k-e and SST k-x models with various thermo-  2  2   !


@u @ v

@ v
2
@u
physical conditions. The Re was calculated using the Darcy velocity P ¼ lturb: 2 þ2 þ þ ð10Þ
(uD ) and H, namely Re ¼ uDmH [1].
@x @y @y @x
The following assumptions were made throughout the current
investigation: The C 2e parameter in Eq. (6) can be described as:
Two-dimensional, steady-state turbulent, fluid flow and heat  
C l g3 1  gg
transfer through the thermal porous systems were assumed C 2e ¼ C 2e þ
0
ð11Þ
[28,34]. 1 þ bg3
Considering heat transfer analysis in high Reynolds numbers,
where g is:
forced convection was assumed to be dominant and natural con-
vection was therefore neglected [9]. Further, with respect to negli- k
g¼S ð12Þ
gible emissivity of the materials, the radiation heat transfer was e
neglected. vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
The heat generation in the fluid and solid media was neglected. u  2  2   !
u @u
@ v
@u
@ v
2
Considering these assumptions, the time averaged continuity, S¼t 2 þ2 þ þ ð13Þ
@x @y @y @x
momentum and energy governing equations can be written as:

@ v

@u The SST k-x model employed in the current investigation utilizes
þ ¼0 ð1Þ
@x @y the standard k-x model, close to solid boundaries and the standard
! k-e model away from solid boundaries [37]. In order to effectively
 


@u

@u

@p  @2u

@2u


q u þv


¼  þ l þ lturb: þ ð2Þ
@x @y @x @x2 @y2
Table 1
  !
Constants for the RNG k-e model and
@ v
@ v


@p  @ 2 v
@ 2 v

q u
þ v
¼  þ l þ lturb: þ ð3Þ the SST k-x model.
@x @y @y @x2 @y2
Parameters Values

   2
! C 1e 1:42
@T @ T
C p lturb: @ T @ 2 T

qC p u
þ v
¼ kf þ þ ð4Þ C 2e 1:68
@x @y Prturb: @x2 @y2 g0 4:38
b 0:012
Here, the constant turbulent Prandtl number can be defined as Cl 0:0845
Prturb: ¼ 0:85 [35]. Cm 100
Rk 6
To simulate turbulent viscosity (lturb: ) in Eqs. (2)-(4) the RNG Rx 2:95
k-e and SST k-x models were employed. Utilization of the RNG a0 1=9
k-e model increases the accuracy of the calculation of the lturb: a0 0:024
[36]. This can be explained as follows. Since the RNG k-e model rk;1 1:176
rk;2 1
controls the dissipation of the turbulence dissipation rate with a
rx;1 2
variable parameter, this model decreases the lturb: and prevents rx;2 1:168
overestimation of the turbulence kinetic energy in complex flows a1 0:31
such as rapidly strained flows, the approach utilized leads to an bi;1 0:075
accurate calculation of the lturb: when compared with the standard bi;2 0:0828
k-e model [36]. The turbulence kinetic energy and turbulence dis- b1 0:09
Rb 8
sipation rate governing equations for the RNG k-e model can be
j 0:41
expressed as:
 M. Torabi et al. / International Journal of Heat and Mass Transfer 132 (2019) 443–461 447

Table 2 convert the standard k-x model to the standard k-e model, a blend-
Grid independence study for the circular cross-sectional configuration in / ¼ 0:6 and ing function (F 1 ) can be used. The turbulence kinetic energy and
Re ¼ 200 from the RNG k-e model.
specific dissipation rate governing equations for the SST k-x model
Grid hki f Error are written as [37]:
u2D
   " #
7000 0.190148 %0.37 @k @k lturb: @ 2 k @ 2 k
q u þv




¼ P  b kx þ l þ

þ ð14Þ
10,000
13,000
0.19086
0.191065
%0.1
%1.17
@x @y rk @x2 @y2
17,000 0.19129 %0.0099
19,000 0.191271

Table 3
Meshes used for three cross-sectional configurations.

Circular cross-sectional Longitudinal elliptical cross- Transverse elliptical cross-

configuration sectional configuration sectional configuration
/ ¼ 0:6 / ¼ 0:8 / ¼ 0:7 / ¼ 0:85 / ¼ 0:7 / ¼ 0:8

RNG k-e model 17,000 23,435 26,231 23,275 41,590 22,669

SST k-x model 41,882 51,996 26,231 23,294 83,069 22,669

Re = 200 Re = 2000 Re = 20000

φ = 0 .8

RNG
k −ε

φ = 0 .6

φ = 0 .8

SST
k −ω

φ = 0 .6

Fig. 3. Pressure (mN2 ) contours from the RNG k-e and SST k-x models for the circular cross-sectional configuration.
448 M. Torabi et al. / International Journal of Heat and Mass Transfer 132 (2019) 443–461

 
@x @x qa
The a in Eq. (15) can be expressed as [37]:
q u
þ v
¼ P  qbi x2 !
@x @y lturb:
 " # a a0 þ Re t

lturb: @ 2 x @ 2 x a ¼ 1 Rx
ð19Þ
þ lþ þ 2 þ 2ð1  F 1 Þ a 1 þ Rxt
Re

rx @x2 @y
 
1 @k @ x @k @ x a0 þ ReRkt
 þ ð15Þ a ¼ ð20Þ
xrx;2 @x @x @y @y 1 þ Ret
R k

The turbulence kinetic energy production for the SST k-x model (P)
in Eqs. (14) and (15) can be described as: a1 ¼ F 1 a1;1 þ ð1  F 1 Þa1;2 ð21Þ

¼ min ðP; 10qb kxÞ ð16Þ
P bi;1 j2
a1;1 ¼   pffiffiffiffiffiffiffi ð22Þ
2  4 3 b1 rx;1 b1
615 þ Rb 7
4 Ret

b ¼ b1 4  4 5 ð17Þ bi;2 j2

1 þ Re t a1;2 ¼   pffiffiffiffiffiffiffi ð23Þ
Rb b1 rx;2 b1
 
k
Ret ¼ ð18Þ F 1 ¼ tanh U41 ð24Þ
mx

Re = 200 Re = 2000 Re = 20000

φ = 0 .85

RNG
k −ε

φ = 0 .7

φ = 0 .85

SST
k −ω

φ = 0 .7

Fig. 4. Pressure (mN2 ) contours from the RNG k-e and SST k-x models for the longitudinal elliptical cross-sectional configuration.
 M. Torabi et al. / International Journal of Heat and Mass Transfer 132 (2019) 443–461 449

"pffiffiffi ! #
k 500l 4qk qk
U1 ¼ min max ; ; ð25Þ lturb: ¼ h i ð30Þ
0:09xy qy2 x rx;2 Dþx y2 x max a1 ; aSF1 x2
   
1 1 @k @ x @k @ x
Dþx ¼ max 2q þ ; 1010 ð26Þ F 2 ¼ tanh U22 ð31Þ
rx;2 x @x @x @y @y
pffiffiffi " #
The bi in Eq. (15) can be expressed as: k 500l
U2 ¼ max 2 ; ð32Þ
bi ¼ F 1 bi;1 þ ð1  F 1 Þbi;2 ð27Þ 0:09xy qy2 x

The rk and rx in Eqs. (14) and (15) are defined respectively as: Values of constants for the RNG k-e model and the SST k-x model
1 are shown in Table 1.
rk ¼ F1 ð1F 1 Þ
ð28Þ Given the dimensionless analysis of thermophysical parameters
rk;1 þ rk;2 in the current investigation, the dimensionless turbulence kinetic
f
energy (hki ) is discussed. The surface averaged turbulence kinetic
1 u2
rx ¼ ð29Þ
D
F1
þ ð1F 1Þ energy (hki f ) can be expressed as:
rx;1 rx;2
Z
1
Moreover, the lturb: in the SST k-x model can be defined as: hki f ¼ kdA ð33Þ
A

Re = 200 Re = 2000 Re = 20000

φ = 0 .8

RNG
k −ε

φ = 0 .7

φ = 0 .8

SST
k −ω

φ = 0 .7

Fig. 5. Pressure (mN2 ) contours from the RNG k-e and SST k-x models for the transverse elliptical cross-sectional configuration.
450 M. Torabi et al. / International Journal of Heat and Mass Transfer 132 (2019) 443–461

A comprehensive literature review revealed that different defini- Nu

c¼

ð41Þ
 dhdx
pi f 2
tions of the Nu have been used in previous numerical investigations H
l uD
of isotropic porous media [19,27]. Here, the Nu definition provided

f
by Kong et al. [20] is utilized: The dimensionless pressure gradient ( dhdx
pi 2
H
luD ) reveals the effects
hH of the friction drag and pressure drag in porous media [29].
Nu ¼ ð34Þ The standard symmetry boundary conditions for the upper and
kf
lower longitudinal lines in Figs. 1b and 2 can be defined as:
where h is the convective heat transfer coefficient and can be
defined as: H H
@ v
@k @ e @ x @ T

@u
y¼ y¼ ¼ ¼ ¼ ¼ ¼ ¼0 ð42Þ
2 2 @y @y @y @y @y @y
Q
h¼ ð35Þ
DT ml Aexp: On the surface of the solid rods the no-slip boundary conditions are
applied:
 
Q ¼ qcp uD ðH  DÞ T
Bx¼H  T
Bx¼H ð36Þ
¼ v
¼ 0 T
¼ T w
u ð43Þ

B
uD ¼ /u ð37Þ The turbulence kinetic energy on the solid rods can be expressed as:

Z @k
1 ¼0 ð44Þ

B ¼
u
dA
u ð38Þ @n
A
Standard wall functions were implemented for the RNG k-e model
R
T
u

dy [1] and the wall boundary condition for the specific dissipation rate
T
B ¼ R ð39Þ is expressed as:


udy
    ðu Þ2
T w  T
Bx¼H  T w  T
Bx¼H xw ¼ xþ ð45Þ
DT ml ¼   ð40Þ m
T w T
B
ln T w T
x¼H The xþ for laminar sublayer and logarithmic region is defined as,
Bx¼H
respectively:
The exposed surface area (Aexp: ) in Eq. (35) was obtained using 1
the perimeter of the solid rods, as illustrated in Figs. 1b and 2. Also, xþ ¼ ð46Þ
bi ðyþ Þ2
the exposed surface diameter (D) in Eq. (36) was defined as D ¼ 2a
and D ¼ 2b for the longitudinal elliptical and transverse elliptical þ
1 duturb:
cross-sectional configurations, respectively. xþ ¼ pffiffiffiffiffiffi

ffi þ ð47Þ
Even though thermal analysis of the isotropic porous media can b1 dy
be performed using the Nu, utilization of a pressure gradient More details regarding Eqs. (46) and (47) can be found in Ref. [37].
results in a more comprehensive thermal assessment. Thus, in Following fundamental investigations of laminar flow [34] and
order to include the effects of both the heat transfer and pressure turbulent flow [29,38] through isotropic porous media, periodic
gradient, simultaneously, the heat transfer efficiency can be boundary conditions for the back and front lines of the unit cells
defined as [19]: can be expressed as:

Fig. 6. The dimensionless pressure gradient from the RNG k-e and SST k-x models for the circular, longitudinal elliptical and transverse elliptical cross-sectional
configurations.
 M. Torabi et al. / International Journal of Heat and Mass Transfer 132 (2019) 443–461 451

x¼H ¼ u
u
x¼H v
x¼H ¼ v
x¼H
x¼H ¼ p
p


x¼H þ Dp analyses of turbulent fluid flow and heat transfer, the local entropy
ð48Þ
kx¼H ¼ kx¼H ex¼H ¼ ex¼H xx¼H ¼ xx¼H generation rate in the current study can be expressed as [40,41]:
"  2   2 # "    2   #
kf @ T
@ T
l @u
2 @ v
@u
@ v
2
Alternatively, in order to establish a periodic temperature profile in S_ gen ¼
2 þ þ
2 þ2 þ þ
the back and front lines of the unit cell, a dimensionless tempera- T @x @y T @x @y @y @x
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
ture boundary condition can be defined as [34,39]: time v eraged HTI time av eraged FFI
2 3 "  
 2  0 2  0 2  0 2 #
T
x¼H  T w T
x¼H  T w kf 4 @T 0 @T 5 l @u0
2
@v @u @ v 0
¼ ð49Þ þ
2 þ þ
2 þ2 þ þ
T
Bx¼H  T w T
Bx¼H  T w T @x @y T @x @y @y @x
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
fluctuating HTI fluctuating FFI

3. Entropy generation ð50Þ

The time averaged HTI and time averaged FFI can be obtained using
For the turbulent fluid flow and heat transfer under investiga- the governing equations and boundary conditions discussed in Sec-
tion the entropy generation is due to a combination of heat transfer tion 2. Alternatively, the fluctuating HTI and FFI can be written,
irreversibility (HTI) and fluid friction irreversibility (FFI). Consider- respectively as [41]:
ing the effects of laminar and turbulent thermophysical parame- 2 3 " 2  2 #
 2  0 2
ters on the irreversibility, the contributions of the time averaged/ kf 4 @T 0 @T 5 lturb: C p @ T
@ T

þ ¼ þ ð51Þ
fluctuating HTI and time averaged/fluctuating FFI should be taken T
2 @x @y Prturb: T
2 @x @y
into account. Using an approach similar to previous second law

RNG k − ε SST k − ω

Circular

Longitudinal
elliptical

Transverse
elliptical

Fig. 7. Velocity magnitude (ms) vectors from the RNG k-e and SST k-x models for the circular ð/ ¼ 0:8Þ, the longitudinal elliptical (/ ¼ 0:85) and the transverse elliptical
(/ ¼ 0:8) cross-sectional configurations at Re ¼ 200.
452 M. Torabi et al. / International Journal of Heat and Mass Transfer 132 (2019) 443–461

"    0 2  0 2 #
l @v @u @ v 0 qe second order upwind scheme for momentum, energy, turbulence
2
@u0
2 þ2 þ þ ¼
ð52Þ kinetic energy, turbulence dissipation rate, and specific dissipation
T
@x @y @y @x T
rate along with the SIMPLE algorithm to solve the pressure-
Considering the utilization of the dimensionless parameters in the velocity coupling were implemented. The convergence criteria for
investigation of the thermal porous systems from a second law of residuals of the continuity, momentum, turbulence kinetic energy,
thermodynamics perspective, the dimensionless local entropy gen- turbulence dissipation rate, and specific dissipation rate governing
eration rate can be expressed as [42]: equations were set to 106 and to 109 for the energy equation
residuals. Numerical simulations were performed using a com-
S_ gen H2
Ns ¼ ð53Þ puter with 2 GB of RAM and a 1.6 GHz CPU, which avoids investi-
kf gation of several unit cells. In the grid independence study, Nu
Further, the dimensionless surface averaged entropy generation and dimensionless pressure gradient in different meshes were
rate can be defined as [42]: examined and revealed insignificant variations. Alternatively, due
Z to the variation in the turbulence kinetic energy observed in the
1 different meshes, the dimensionless turbulence kinetic energy
Nt ¼ Ns dA ð54Þ
A A was selected as a suitable parameter to evaluate and achieve grid
independent results. The values of the dimensionless turbulence
kinetic energy for the circular cross-sectional configuration when
4. Numerical method
Re ¼ 200 and / ¼ 0:6 are illustrated in Table 2. The error between
the third and fourth grid is lower than 1%. Thus, it is apparent that
In order to solve the governing equations, the finite volume
a grid, consisting of 17,000 nodes is sufficient to provide grid
method is utilized. To discretize the governing equations, the

Re = 200 Re = 2000 Re = 20000

φ = 0 .8

RNG
k −ε

φ = 0 .6

φ = 0 .8

SST
k −ω

φ = 0 .6

Fig. 8. Turbulence kinetic energy (ms2 ) contours from the RNG k-e and SST k-x models for the circular cross-sectional configuration.
2
 M. Torabi et al. / International Journal of Heat and Mass Transfer 132 (2019) 443–461 453

independent results. More information regarding the meshes used heat transfer through the isotropic porous media were considered.
for different cross-sectional configurations is illustrated in Table 3. Code validation investigations for the Nu showed good agreement
with analytical, numerical and experimental data available in Ref.
[31]. Further, to ensure accuracy of the applied numerical proce-
5. Results and discussion dure, the dimensionless pressure gradient for the circular, longitu-
dinal elliptical and transverse elliptical cross-sectional
5.1. Validation configurations was compared with results provided by Pedras
and de Lemos [28,29].
A literature review on turbulent fluid flow and heat transfer
through isotropic porous media revealed that relatively few exper-
imental investigations have addressed the thermophysical condi- 5.2. Pressure and velocity
tions considered in the current investigation. Most recently,
Nazari et al. [43] conducted an experimental analysis of turbulent Figs. 3–5 illustrate the pressure contours simulated by the RNG
fluid flow and heat transfer through channels filled with stainless k-e and SST k-x models for the circular, longitudinal elliptical and
steel spheres. Entropy generation, the Nu and pressure drop were transverse elliptical cross-sectional configurations, respectively. As
discussed for different values of internal heat generation, sphere the Re increases, variation of pressure in the unit cells increases for
diameters, the Re, and volumetric flow rates. It was found that all three cross-sectional configurations. A careful comparison
the entropy generation number increases with sphere diameter between the low and high porosity media reveals an inverse rela-
for Reynolds numbers more than 1800. With respect to the lack tionship between the porosity and pressure gradient for the three
of experimental analyses in the investigated thermophysical con- cross-sectional configurations examined. Considering the pressure
ditions, the experimental results related to laminar fluid flow and contours illustrated from two turbulence models, an unstable

Re = 200 Re = 2000 Re = 20000

φ = 0 .85

RNG
k −ε

φ = 0 .7

φ = 0 .85

SST
k −ω

φ = 0 .7

Fig. 9. Turbulence kinetic energy (ms2 ) contours from the RNG k-e and SST k-x models for the longitudinal elliptical cross-sectional configuration.
2
454 M. Torabi et al. / International Journal of Heat and Mass Transfer 132 (2019) 443–461

trend for the SST k-x model was observed. It should be noted that when compared with the circular and transverse elliptical cross-
the symmetric pressure contours for three cross-sectional configu- sectional configurations. It could be argued that the longitudinal
rations were illustrated in Ref. [29]. Given the symmetric physical elliptical cross-sectional configuration looks like a wing-shaped
conditions, the pressure contours provided by the RNG k-e and SST cross-section [38], which can be considered as a reason for the
k-x models in Figs. 3 and 5 were compared. As shown, the SST k-x low dimensionless pressure gradient. The dimensionless pressure
model illustrates asymmetric pressure contours when compared gradient of the transverse elliptical cross-sectional configuration
with the RNG k-e model. from Ref. [29] falls between the presented results for the trans-
Fig. 6 illustrates the dimensionless pressure gradients for the verse elliptical cross-sectional configuration with / ¼ 0:8 and
circular, longitudinal elliptical and transverse elliptical cross- / ¼ 0:7.
sectional configurations. As expected, the dimensionless pressure Fig. 7 illustrates the velocity magnitude vectors for the circular,
gradient increases with Re and decreases with porosity. The RNG longitudinal elliptical and transverse elliptical cross-sectional con-
k-e and SST k-x models under-predict the dimensionless pressure figurations at Re ¼ 200. With respect to the Re defined in Section 2,
gradient for the circular cross-sectional configuration when com- the velocity may be more or less than values of the Darcy velocity.
pared with Refs. [28,29], and show monotonically increasing trend As there is less surface area for fluid to flow around solid rods com-
with Re. It should be noted that the low Rek-emodel utilized in pared with regions before the solid rods, higher velocity values are
Refs. [28,29] over-predicts the dimensionless pressure gradient observed around the solid rods which maintains constant mass
[44]. Fig. 6 provides a good agreement for the longitudinal elliptical flow rate in the unit cells. A common feature between three
cross-sectional configuration when compared with Ref. [29]. A cross-sectional configurations is high-velocity regions at the inlet
comparison between different cross-sectional configurations indi- and outlet, and separation regions on the surface of the solid rods.
cates that the longitudinal elliptical cross-sectional configuration Given the steady-state assumption used in the current investiga-
with / ¼ 0:8 results in the lowest dimensionless pressure gradient tion, a wake flow is produced at the back of the central solid rods

Re = 200 Re = 2000 Re = 20000

φ = 0 .8

RNG
k −ε

φ = 0 .7

φ = 0 .8

SST
k −ω

φ = 0 .7

Fig. 10. Turbulence kinetic energy (ms2 ) contours from the RNG k-e and SST k-x models for the transverse elliptical cross-sectional configuration.
2
 M. Torabi et al. / International Journal of Heat and Mass Transfer 132 (2019) 443–461 455

Fig. 11. The dimensionless turbulence kinetic energy from the RNG k-e and SST k-x models for the circular, longitudinal elliptical and transverse elliptical cross-sectional
configurations.

Re = 200 Re = 2000 Re = 20000

φ = 0 .8

RNG
k −ε

φ = 0 .6

φ = 0 .8

SST
k −ω

φ = 0 .6

Fig. 12. Temperature (K) contours from the RNG k-e and SST k-x models for the circular cross-sectional configuration.
456 M. Torabi et al. / International Journal of Heat and Mass Transfer 132 (2019) 443–461

in the different cross-sectional configurations. The wake flow at increases, more turbulence kinetic energy is produced for three
the back of the central solid rod in the longitudinal elliptical cross-sectional configurations. Fig. 9 shows higher turbulence
cross-sectional configuration is smaller than that observed in the kinetic energy at higher Reynolds numbers for the longitudinal
circular and transverse elliptical cross-sectional configurations. elliptical cross-sectional configuration illustrated by both turbu-
The latter expression could provide an explanation for the low lence models. It is shown that a low turbulence kinetic energy
dimensionless pressure gradient observed in the longitudinal ellip- region behind the central rod decreases for high Reynolds numbers
tical cross-sectional configuration illustrated in Fig. 6. To verify such as Re ¼ 20; 000. A similar explanation can be applied for the
periodic boundary conditions applied for velocity vectors, the circular and transverse elliptical cross-sectional configurations in
velocity magnitude vectors were examined at the right and left Figs. 8 and 10, respectively. Considering symmetric turbulence
lines. It was revealed that this parameter shows similar values at kinetic energy contours obtained by different turbulence modeling
the left and right lines of the unit cells. approaches [22,29,44], performance of the RNG k-e and SST k-x
models can be compared. The RNG k-e model illustrates the sym-
5.3. Turbulence kinetic energy metric turbulence kinetic energy contours for the circular cross-
sectional configuration. Alternatively, the SST k-x model shows
Figs. 8–10 illustrate the turbulence kinetic energy contours for asymmetric turbulence kinetic energy contours with / ¼ 0:8 in
the circular, longitudinal elliptical and transverse elliptical cross- Fig. 8. For the longitudinal elliptical cross-sectional configuration
sectional configurations, respectively. Similar to previous investi- in Fig. 9 both the RNG k-e and SST k-x models provide the sym-
gations [22,28,45], as porosity for all three cross-sectional configu- metric turbulence kinetic energy contours [29]. Fig. 10 illustrates
rations decreases, i.e., the space for fluid to flow through the the asymmetric turbulence kinetic energy contours obtained from
isotropic porous media is reduced, the RNG k-e and SST k-x models the SST k-x model for the transverse elliptical cross-sectional
show higher turbulence kinetic energy. Alternatively, as the Re configuration with / ¼ 0:7. Again, while the RNG k-e model is

Re = 200 Re = 2000 Re = 20000

φ = 0 .85

RNG
k −ε

φ = 0 .7

φ = 0 .85

SST
k −ω

φ = 0 .7

Fig. 13. Temperature (K) contours from the RNG k-e and SST k-x models for the longitudinal elliptical cross-sectional configuration.
 M. Torabi et al. / International Journal of Heat and Mass Transfer 132 (2019) 443–461 457

recommended for highly strained flows, the SST k-x model was model provides more stable results for different Reynolds numbers
developed to improve the performance of the k-x model in bound- for the longitudinal elliptical and circular cross-sectional configu-
ary layer problems [37]. rations. Again, the SST k-x model reveals lower dimensionless tur-
Fig. 11 illustrates the dimensionless turbulence kinetic energy bulence kinetic energy than the RNG k-e model for the transverse
for the circular, longitudinal elliptical and transverse elliptical elliptical cross-sectional configuration at Re ¼ 200, except for the
cross-sectional configurations. As expected, decreasing porosity, transverse elliptical in / ¼ 0:7. While the RNG k-e turbulence
i.e., larger solid rods in the isotropic porous media, increases the model takes both high and low Re effects into account, due to
dimensionless turbulence kinetic energy. This can be explained the lack of low Re effects in the turbulence kinetic energy and
as follows. As the porosity decreases, the turbulence kinetic energy specific dissipation rate governing equations in the SST k-x model,
productions defined in Eqs. (10) and (16) increase [44]. Also, the the dimensionless turbulence kinetic energy at the low Reynolds
turbulence kinetic energy increases the lturb: , see Eqs. (8) and numbers was under-predicted.
(30). Considering the turbulence kinetic energy production terms
given by the product of the velocity gradient and the lturb: , the tur- 5.4. Temperature and Nusselt number
bulence kinetic energy subsequently increases. Fig. 11 compares
the performance of the RNG k-e and SST k-x models to predict Fig. 12 illustrates the temperature contours for the circular
the dimensionless turbulence kinetic energy for the three cross- cross-sectional configuration. Enhancement of thermal field
sectional configurations under investigation. For the circular and around the exposed surface of the central cylinders and lower tem-
longitudinal elliptical cross-sectional configurations at Re ¼ 200, perature behind the central cylinders are apparent. Two low-
the SST k-x model under-predicts the dimensionless turbulence temperature regions are formed behind the central cylinder at
kinetic energy compared with the RNG k-e model. The RNG k-e Re ¼ 200. As the Re increases, the low-temperature regions

Re = 200 Re = 2000 Re = 20000

φ = 0 .8

RNG
k −ε

φ = 0 .7

φ = 0 .8

SST
k −ω

φ = 0 .7

Fig. 14. Temperature (K) contours from the RNG k-e and SST k-x models for the transverse elliptical cross-sectional configuration.
458 M. Torabi et al. / International Journal of Heat and Mass Transfer 132 (2019) 443–461

decrease and thermal boundary layer is limited to the central kinetic energy from the SST k-x model is higher than the RNG
cylinder. Both turbulence models indicate that as the porosity k-e model for high Reynolds numbers, see Fig. 9. The higher turbu-
decreases, the temperature variation in the unit cells increases. lence kinetic energy produces higher lturb: in Eq. (30) and according
Fig. 12 compares the performance of the RNG k-e and the SST to Eq. (4), brings about higher temperatures. Fig. 13 also indicates
k-x models in order to simulate temperature contours in the iso- that the RNG k-e and SST k-x models reveal the symmetric tem-
tropic porous media. As discussed in previous investigations perature contours of the longitudinal elliptical cross-sectional con-
[1,34], symmetric temperature contours should be observed in figuration. Fig. 14 illustrates the temperature contours for the
the unit cells of the isotropic porous media. Fig. 12 illustrates sym- transverse elliptical cross-sectional configuration. The low-
metric temperature contours for the circular cross-sectional con- temperature regions behind the central rod in the transverse ellip-
figurations and indicates that the SST k-x model results in a tical cross-sectional configuration are larger when compared with
discrepancy between the top and bottom parts of temperature con- the circular and longitudinal elliptical cross-sectional configura-
tours. Temperature contours for the longitudinal elliptical cross- tions. Regarding the performance of the RNG k-e and SST k-x mod-
sectional configuration are illustrated in Fig. 13. The low- els utilized, Fig. 14 indicates that temperature contours in the
temperature regions are apparent behind the central solid rod of isotropic porous media consisting of the transverse elliptical
the longitudinal elliptical cross-sectional configuration. As the Re cross-sectional configuration with / ¼ 0:7 may not be simulated
increases, the low-temperature region separates into two parts, using the SST k-x model.
which is clearly apparent for the SST k-x model. For Re ¼ 20; 000 Fig. 15 illustrates the effect of the Re on the Nu using the RNG
the SST k-x model reveals higher temperature values compared k-e and SST k-x models. According to the RNG k-e model for the
with the RNG k-e model. It is also apparent that the turbulence circular cross-sectional configuration in Re ¼ 200; 000 and

Fig. 15. The Nu from the RNG k-e and SST k-x models for the circular, longitudinal elliptical and transverse elliptical cross-sectional configurations.

Fig. 16. The heat transfer efficiency from the RNG k-e and SST k-x models for the circular, longitudinal elliptical and transverse elliptical cross-sectional configurations.
 M. Torabi et al. / International Journal of Heat and Mass Transfer 132 (2019) 443–461 459

/ ¼ 0:6, the Nu ¼ 780. Considering the same turbulence model and configurations, a higher porosity provides more exposed surface
the Re for the longitudinal elliptical and transverse elliptical cross- area for the thermal interaction between the fluid flow and solid
sectional configurations with / ¼ 0:7, the Nu ¼ 628 and rods. Alternatively, as the porosity decreases for the transverse
Nu ¼ 1183, respectively. It is apparent that the transverse elliptical elliptical cross-sectional configuration, the exposed surface area
cross-sectional configuration results in the highest Nu when com- increases. This results in more thermal interaction between the
pared with the circular and longitudinal elliptical cross-sectional fluid flow and solid rods. Fig. 15 reveals that the RNG k-e model
configurations. Fig. 15 shows the effects of the porosity on the provides higher Nu compared with the SST k-x model for the
Nu. As the porosity decreases, two different effects on the Nu are circular cross-sectional configuration in Re ¼ 200; 000 and the lon-
observed. For the transverse elliptical cross-sectional configuration gitudinal elliptical cross-sectional configuration in Re ¼ 2; 000 and
the Nu increases as the porosity decreases. Alternatively, for the Re ¼ 20; 000. Fig. 15 indicates that the RNG k-e model calculates
circular and longitudinal elliptical cross-sectional configurations, lower Nu compared with the SST k-x model for the transverse
the Nu increases with increasing the porosity. It should be noted elliptical cross-sectional configuration with / ¼ 0:7 and in
that for the circular and longitudinal elliptical cross-sectional different Reynolds numbers, except Re ¼ 20; 000.

C i r cu l a r Longitudinal elliptical Transverse elliptical

φ = 0 .8 φ = 0 .85 φ = 0 .8

Re = 20000 φ = 0 .6 φ = 0 .7 φ = 0 .7

φ = 0 .8 φ = 0 .85 φ = 0 .8

Re = 200000
φ = 0 .6 φ = 0 .7 φ = 0 .7

3 :K ) contours from the RNG k-e model for the circular, longitudinal elliptical and transverse elliptical cross-sectional
Fig. 17. The local entropy generation rate (mW
configurations.
460 M. Torabi et al. / International Journal of Heat and Mass Transfer 132 (2019) 443–461

5.5. Heat transfer efficiency and the FFI of the system. This consequently decreases the local
entropy generation rate as the porosity increases.
Fig. 16 illustrates the heat transfer efficiency for the different Fig. 18 illustrates the variation of the N t versus the Re for the
cross-sectional configurations. According to the RNG k-e and SST circular, longitudinal elliptical and transverse elliptical cross-
k-x models, the longitudinal elliptical cross-sectional configura- sectional configurations. As discussed previously, the N t increases
tion with / ¼ 0:85 demonstrates the highest heat transfer effi- with Re for the three cross-sectional configurations. Considering
ciency. The circular cross-sectional configuration with / ¼ 0:8 is low pressure gradient and high Nu resulted from exposed surface
second, followed by the longitudinal elliptical cross-sectional con- of the solid rods, the longitudinal elliptical cross-sectional configu-
figuration with / ¼ 0:7. Although the heat transfer efficiency ration with / ¼ 0:85 reveals the lowest N t when compared with
decreases with the Re for most of the cases investigated, the lowest other cross-sectional configurations. The transverse elliptical
rate of reduction was observed for the transverse elliptical cross- cross-sectional configuration with / ¼ 0:7 demonstrates the high-
sectional configuration with / ¼ 0:7 and / ¼ 0:8. The high porosity est N t , i.e., the highest exergy destruction. Thus, the longitudinal
media brings about a higher heat transfer efficiency for all three elliptical cross-sectional configuration with / ¼ 0:85 reveals the
cross-sectional configurations. Considering the high Nu for the superior performance from the second law of thermodynamics
transverse elliptical cross-sectional configuration discussed in Sec- perspective within all of the cross-sectional configurations
tion 5.4, the high dimensionless pressure gradient of the transverse investigated.
elliptical reduces the heat transfer efficiency. Again, it should be
noted that inclusion of the pressure gradient in the investigation
6. Conclusions
of heat transfer in the porous thermal systems results in more
comprehensive results.
The current numerical investigation applied the RNG k-e and
SST k-x models to turbulent fluid flow, heat transfer and entropy
5.6. Local and total entropy generation generation through isotropic porous media composed of circular,
longitudinal elliptical and transverse elliptical cross-sectional con-
Fig. 17 illustrates the local entropy generation rate contours for figurations. Fundamental parameters such as dimensionless pres-
the three cross-sectional configurations using the RNG k-e model sure gradient, dimensionless turbulence kinetic energy, the Nu,
in Re ¼ 20; 000 and Re ¼ 200; 000. As expected, the local entropy and heat transfer efficiency, have all been investigated for different
generation rate is higher in the Re ¼ 200; 000 compared with the Reynolds numbers and porosities. The second law analysis was dis-
Re ¼ 20; 000 for all three cross-sectional configurations investi- cussed for all three cross-sectional configurations. Considering
gated. This is due to the fact that, as the Re increases, the HTI symmetric contours of thermophysical properties achieved in
and FFI both increase, and consequently the local entropy genera- other investigations, the RNG k-e model revealed a more successful
tion rate, which has direct relation with these two features of the illustration of contours when compared with the SST k-x model.
system (see Eq. (50)) increases. The inlet fluid surrounds the low The RNG k-e model revealed a more stable trend compared with
temperature solid rods and the high local entropy generation rates, the SST k-x model in order to provide the dimensionless turbu-
due to the significant heat transfer from the exposed surface of the lence kinetic energy in low Reynolds numbers. Investigation of
solid rods are observed. Alternatively, the local entropy generation the thermal characteristics of the isotropic porous media revealed
rate is lower at the back of the solid rods, where the low- a direct relationship between the Nu and porosity for the circular
temperature regions, and low heat transfer from the fluid to solid, and longitudinal elliptical cross-sectional configurations and a
were found for the three cross-sectional configurations in Figs. 12– reverse relation between the Nu and porosity for the transverse
14. A reverse relationship between the porosity and the local elliptical cross-sectional configuration. It was shown that the
entropy generation rate is revealed. It could be argued that an longitudinal elliptical cross-sectional configuration results in a
increase in the value of the porosity of the system, shifts the porous higher heat transfer efficiency when compared with the circular
media to that of the clear one, and hence it reduces both the HTI and transverse elliptical cross-sectional configurations. The

Fig. 18. The N t from the RNG k-e model for the circular, longitudinal elliptical and transverse elliptical cross-sectional configurations.
 M. Torabi et al. / International Journal of Heat and Mass Transfer 132 (2019) 443–461 461

longitudinal elliptical cross-sectional configuration also provided [21] T.P. de Carvalho, H.P. Morvan, D.M. Hargreaves, H. Oun, A. Kennedy, Pore-scale
numerical investigation of pressure drop behaviour across open-cell metal
the lowest entropy generation. The current investigation provides
foams, Trans. Porous Med. 117 (2017) 311–336.
fundamental information regarding the application of RANS turbu- [22] P. Kundu, V. Kumar, Y. Hoarau, I.M. Mishra, Numerical simulation and analysis
lence models in isotropic porous media and the cross-sectional of fluid flow hydrodynamics through a structured array of circular cylinders
configuration that maximizes the heat transfer and minimizes forming porous medium, App. Math. Model 40 (2016) 9848–9871.
[23] M.F. Uth, Y. Jin, A.V. Kuznetsov, H. Herwig, A direct numerical simulation study
the entropy generation, simultaneously. Thus, the results will be on the possibility of macroscopic turbulence in porous media: Effects of
of value for engineers who work in connection with cooling elec- different solid matrix geometries, solid boundaries, and two porosity scales,
tronic packages, heat exchangers and other radiator systems. Phys. Fluids 28 (2016) 065101.
[24] J. Fu, Y. Tang, J. Li, Y. Ma, W. Chen, H. Li, Four kinds of the two-equation
turbulence model’s research on flow field simulation performance of DPF’s
Conflict of interest porous media and swirl-type regeneration burner, App. Therm. Eng. 93 (2016)
397–404.
[25] M. Torabi, N. Karimi, G.P. Peterson, S. Yee, Challenges and progress on the
No conflict of interest is declared. modelling of entropy generation in porous media: A review, Int. J. Heat Mass
Transf. 114 (2017) 31–46.
[26] S. Chikh, N. Allouache, Optimal performance of an annular heat exchanger
References
with a porous insert for a turbulent flow, App. Therm. Eng. 104 (2016) 222–
230.
[1] M.J.S. de Lemos, Turbulence in Porous Media: Modeling and Applications, [27] T.I. Mulcahey, M.G. Pathak, S.M. Ghiaasiaan, The effect of flow pulsation on
Elsevier, Oxford, 2012. drag and heat transfer in an array of heated square cylinders, Int. J. Therm. Sci.
[2] S. Whitaker, Advances in theory of fluid motion in porous media, Ind. Eng. 64 (2013) 105–120.
Chem. 61 (1969) 14–28. [28] M.H.J. Pedras, M.J.S. de Lemos, Simulation of turbulent flow in porous media
[3] D.A. Nield, A. Bejan, Convection in Porous Media, Springer, Switzerland, 2017. using a spatially periodic array and a low Re two-equation closure, Num. Heat
[4] M.E. Nimvari, N.F. Jouybari, Investigation of turbulence effects within porous Transf., Part A: App. 39 (2001) 35–59.
layer on the thermal performance of a partially filled pipe, Int. J. Therm. Sci. [29] M.H.J. Pedras, M.J.S. de Lemos, Computation of turbulent flow in porous media
118 (2017) 374–385. using a low-Reynolds k -e model and an infinite array of transversally
[5] M.J.S. de Lemos, C.B. Masciarelli, Turbulent natural convection in a composite displaced elliptic rods, Num. Heat Transf., Part A: App. 43 (2003) 585–602.
annulus using a novel numerical scheme and the thermal nonequilibrium [30] M. Torabi, G.P. Peterson, M. Torabi, N. Karimi, A thermodynamic analysis of
hypothesis, Num. Heat Transf., Part A: App. 71 (2017) 837–854. forced convection through porous media using pore scale modeling, Int. J. Heat
[6] S. Esmaeili Rad, H. Afshin, B. Farhanieh, Heat transfer enhancement in shell- Mass Transf. 99 (2016) 303–316.
and-tube heat exchangers using porous media, Heat Transf. Eng. 36 (2014) [31] M. Torabi, M. Torabi, G.P. Peterson, Heat transfer and entropy generation
262–277. analyses of forced convection through porous media using pore scale
[7] Y. Jiang, Y. Feng, S. Zhang, J. Qin, W. Bao, Numerical heat transfer analysis of modeling, J. Heat Transf. 139 (2017) 012601.
transcritical hydrocarbon fuel flow in a tube partially filled with porous media, [32] N.A. Horton, D. Pokrajac, Onset of turbulence in a regular porous medium: An
Open Phys. 14 (2016) 659–667. experimental study, Phys. Fluids 21 (2009) 045104.
[8] M. Brito, J. Fernandes, J.B. Leal, Porous media approach for RANS simulation of [33] I. Kececioglu, Y. Jiang, Flow through porous media of packed spheres saturated
compound open-channel flows with submerged vegetated floodplains, with water, J. Fluids Eng. 116 (1994) 164–170.
Environ. Fluid Mech. 16 (2016) 1247–1266. [34] F. Kuwahara, M. Shirota, A. Nakayama, A numerical study of interfacial
[9] K. Vafai, Handbook of Porous Media, CRC Press, New York, 2015. convective heat transfer coefficient in two-energy equation model for
[10] F.E. Teruel, Entrance effect on the interfacial heat transfer and the thermal convection in porous media, Int. J. Heat Mass Transf. 44 (2001) 1153–1159.
dispersion in laminar flows through porous media, Int. J. Therm. Sci. 104 [35] W.M. Kays, Turbulent Prandtl number-Where are we?, J Heat Transf. 116
(2016) 172–185. (1994) 284–295.
[11] W. Peng, M. Xu, X. Huai, Z. Liu, CFD study on local fluid-to-wall heat transfer in [36] E. Martinez-Espinosa, W. Vicente, M. Salinas-Vazquez, I. Carvajal-Mariscal,
packed beds and field synergy analysis, J. Therm. Sci. 25 (2016) 161–170. Numerical analysis of turbulent flow in a small helically segmented finned
[12] M. Zafari, M. Panjepour, M. Meratian, M.D. Emami, CFD simulation of forced tube bank, Heat Transf. Eng. 38 (2016) 47–62.
convective heat transfer by Tetrakaidecahedron model in metal foams, J. [37] F.R. Menter, Two-equation eddy-viscosity turbulence models for engineering
Porous Media 19 (2016) 1–11. applications, AIAA J. 32 (1994) 1598–1605.
[13] N.F. Jouybari, T.S. Lundstrom, J.G.I. Hellstrom, M. Maerefat, M.E. Nimvari, [38] A. Horvat, M. Leskovar, B. Mavko, Comparison of heat transfer conditions in
Numerical computation of macroscopic turbulent quantities in a porous tube bundle cross-flow for different tube shapes, Int. J. Heat Mass Transf. 49
medium: An extension to a macroscopic turbulence model, J. Porous Media 19 (2006) 1027–1038.
(2016) 497–513. [39] M.B. Saito, M.J.S. de Lemos, A correlation for interfacial heat transfer coefficient
[14] M. Mößner, R. Radespiel, Modelling of turbulent flow over porous media using for turbulent flow over an array of square rods, J. Heat Transf. 128 (2006) 444–
a volume averaging approach and a Reynolds stress model, Comput. Fluids 108 452.
(2015) 25–42. [40] T. Wenterodt, C. Redecker, H. Herwig, Second law analysis for sustainable heat
[15] K.M. Shirvan, R. Ellahi, S. Mirzakhanlari, M. Mamourian, Enhancement of heat and energy transfer: The entropic potential concept, App. Energy 139 (2015)
transfer and heat exchanger effectiveness in a double pipe heat exchanger 376–383.
filled with porous media: Numerical simulation and sensitivity analysis of [41] F. Kock, H. Herwig, Local entropy production in turbulent shear flows: A high-
turbulent fluid flow, App. Therm. Eng. 109 (2016) 761–774. Reynolds number model with wall functions, Int. J. Heat Mass Transf. 47
[16] H. Saberinejad, A. Keshavarz, Reciprocating turbulent flow heat transfer (2004) 2205–2215.
enhancement within a porous medium embedded in a circular tube, App. [42] M. Torabi, M. Torabi, S.M. Ghiaasiaan, G.P. Peterson, The effect of Al2O3-water
Therm. Eng. 102 (2016) 1355–1365. nanofluid on the heat transfer and entropy generation of laminar forced
[17] W.C. Galuppo, M.J.S. de Lemos, Turbulent heat transfer past a sudden convection through isotropic porous media, Int. J. Heat Mass Transf. 111
expansion with a porous insert using a nonlinear model, Num. Heat Transf., (2017) 804–816.
Part A: App. 71 (2017) 290–310. [43] M. Nazari, D. Jalali Vahid, R.K. Saray, Y. Mahmoudi, Experimental investigation
[18] N.F. Jouybari, M. Maerefat, M.E. Nimvari, A pore scale study on turbulent of heat transfer and second law analysis in a pebble bed channel with internal
combustion in porous media, Heat Mass Transf. 52 (2015) 269–280. heat generation, Int. J. Heat Mass Transf. 114 (2017) 688–702.
[19] J. Yang, Q. Wang, M. Zeng, A. Nakayama, Computational study of forced [44] F. Kuwahara, T. Yamane, A. Nakayama, Large eddy simulation of turbulent flow
convective heat transfer in structured packed beds with spherical or in porous media, Int. Commun. Heat Mass Transf. 33 (2006) 411–418.
ellipsoidal particles, Chem. Eng. Sci. 65 (2010) 726–738. [45] J. Yang, M. Zhou, S.Y. Li, S.S. Bu, Q.W. Wang, Three-dimensional numerical
[20] Y.Q. Kong, L.J. Yang, X.Z. Du, Y.P. Yang, Air-side flow and heat transfer analysis of turbulent flow in porous media formed by periodic arrays of cubic,
characteristics of flat and slotted finned tube bundles with various tube spherical, or ellipsoidal particles, J. Fluids Eng. 136 (2014) 011102.
pitches, Int. J. Heat Mass Transf. 99 (2016) 357–371.