Applied Thermal Engineering 71 (2014) 104e118

Contents lists available at ScienceDirect

Applied Thermal Engineering

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

Correlation studies of hydrodynamics and heat transfer in metal foam

heat exchangers
Shaolin Mao*, Norman Love, Alma Leanos, Gerardo Rodriguez-Melo
Department of Mechanical Engineering, The University of Texas at El Paso, TX 79968, USA

h i g h l i g h t s

Systematically examined correlations for ow and heat transfer in bulk metallic foam heat exchangers.

Established novel correlations for friction factor in metallic foam with applications in ACCs.

Compared analytical and numerical studies of heat transfer and air ow in metallic foam matrix.

2D and 3D computational uid dynamics (CFD) simulations have been conducted for optimization design and analysis in this R&D work.

a r t i c l e i n f o

a b s t r a c t

Article history:
Received 21 February 2014
Accepted 15 June 2014
Available online 25 June 2014

This study presents the correlations of both hydrodynamics and heat transfer in a metal foam heat
exchanger. The present work is focused on the application to dry cooling such as air-cooled condensers
(ACCs). In particular empirical correlations for the permeability, form drag coefcient, friction factor, and
the overall heat transfer coefcient for different samples of metallic foam have been validated and
veried with available experimental data and numerical simulations. The modied correlations used in
this study are established through the validation & verication studies of metal foam heat exchangers. In
order to address the difference, nned tube heat exchangers are used to compare to the metal foam heat
exchangers with the same geometry size and layout. For fully wrapped metal foam heat exchangers, the
prediction using empirical correlation is consistent with computational uid dynamics (CFD) simulations. However, the scenarios become complicated for partially wrapped metal foam heat exchangers.
The numerical results show that there is an optimal choice of the porosity of metal foam in which the
wall heat transfer coefcient and pressure drop reach the design goal. Overall, the heat transfer capability
of metal foam heat exchangers can supersede conventional compact heat exchangers given optimal
scenario.
Published by Elsevier Ltd.

Keywords:
Metal foam
Metallic foam
Dry cooling
Air-cooled condenser
Porous medium
DarcyeForchheimer model

1. Introduction
The stochastic or periodic metal and carbon foam materials are
considered as one of the most promising types of methods that can
enhance heat transfer through increases in surface area, structural
strength, and contains many multifunctional capabilities. For
example, the high ratio of contact area to volume, embedded tortuosity, and outstanding heat transfer capability make the bulk
glass metallic foam one of the best candidates to replace traditional
nned tube heat exchangers with great potential in dry cooling for
industrial power plants [9]. In some cases this type of technology
may partially or completely supersede the current wet cooling

* Corresponding author. Tel.: 1 915 747 5830.

E-mail addresses: smao@utep.edu, slm_wvu@yahoo.com (S. Mao).
http://dx.doi.org/10.1016/j.applthermaleng.2014.06.035
1359-4311/Published by Elsevier Ltd.

tower technologies used in thermal power plant [1,16,31]. There is a

great interest in research and industrial applications of bulk cellular
foams in the areas of nuclear waste storage, impact reduction,
medical devices, environmental protection, and aerospace exploration [6,20]. For more details about thermal-hydraulic transport in
high porosity cellular ceramic and metallic foam materials, a recent
state-of-the-art knowledge review article by Zhao [36] presents a
good overview. There are also huge resources about foam material
properties which can be obtained directly from manufacturers.
In the last decade, heat transfer and pressure drop of ceramic
and metallic foam heat exchangers have been extensively investigated in both microscopic structure and macroscopic thermohydraulic behavior ([21] and references therein). Calmidi and
Mahajan [8] characterized the heat transfer behaviors of aluminum
foams under air forced convection in a wind tunnel. Bhattacharya
et al. [4] experimentally and analytically studied the effective

S. Mao et al. / Applied Thermal Engineering 71 (2014) 104e118

thermal conductivity keff, permeability K, and the inertial coefcient F of high porosity aluminum and carbon foams, in which they
showed that F only depends on the porosity 4, not on the pore
diameter dp and shapes. Boomsma et al. [6] studied the thermal
resistance of water forced convection for open-cell aluminum
foams compact heat exchangers on the heater-foam interface and
obtained two to three times lower thermal resistance than those
comparable commercial products. Modied correlations for the
overall heat transfer were suggested in their work. Carbon foam
heat exchangers have been investigated by Wu et al. [33] and
Gallego and Klett [14] for the uniqueness of its heat transfer
capability. Odabaee and Hooman [28] performed numerical
investigation of metal foam heat exchangers in both overall heat
transfer and the pressure drop for geothermal power plant aircooled condensers. The metal foam design in their simulation has
shown 2 to 6 times higher of heat transfer capability than conventional design with the same operating conditions. Dai et al. [10]
compared metal foam design vs. conventional louver-ns design
and shows a lighter and smaller size with the same thermalhydraulic performance of the heat exchangers. More recently,
Mancin et al. [22,23] experimentally measured heat transfer coefcient and pressure drop for a series of aluminum and copper foam
samples under air forced convection conditions. From the optimization point of view, the pressure loss reduction and the heat
transfer increase are identied as two critical parameters to meet
the goal of geometry and weight of metal foam heat exchangers.
The motivation of this study is two-fold. First the determination
of the criteria to compare metal foam heat exchangers to conventional nned tube heat exchangers is presented. Historically both
metal foams and nned tubes can be treated as porous media,
which is conversely dened by its porosity 4, i.e., the fraction of the
total volume occupied by void and free to the uid ow. In addition
to the porosity 4, the pore size df and dp shape and relative density,
i.e. number of pores per unit length (PPI, inch) are important factors
to inuence the uid ow in metal foams. Most transport experiments conducted on porous medium have been based on beds of
packed particles with more or less uniform spherical shape for
individual particle [13,18]. Because the geometric features of metal
foams show a large difference from the structure of ns, the strict
quantitative comparison is not a trivial task. A literature survey
shows that different research groups have led to different, and even
contradictory, conclusions for the overall heat transfer capability of
metal foam heat exchangers [21]. In order to establish the underlying correlations of ow and heat transfer, as the rst step, this
study will introduce equivalent parameters for the uniqueness of
metal foams. The discussion of strict criteria for the comparison will
help to clarify the confusion in the design and analysis of metal
foam heat exchangers. The second objective is to use the evaluation
criteria established here to validate and verify most available correlation models for the permeability, form drag coefcient, overall
heat transfer and the pressure drop. To explore the second objective, linear and non-linear regression analysis techniques will be
applied for tting widely spread experimental data and simulation
results. As a supplementary design tool, the 2D/3D CFD simulation
of typical metal foam heat exchangers are undertaken to further
examine the modied correlation models [24,25,15,16,35]. The
overall goal is to determine the right correlation of pressure drop
and heat transfer factors which can be directly applied to conceptual design and prototyping of metal foam heat exchangers for
ACCs.

105

century, is a long-standing challenge facing thermal-hydraulics and

multiphase ow simulations due to the complicated structure of
the materials, widely disparate length scales and turbulent ow
time scales involved. In order to overcome the difculties associated with the multiple scales in a porous medium, the concept of
representative elementary volume (r.e.v.) has been used to simplify
the governing equations of momentum and energy transport [2].
The equations of motion are written with respect to the averaged
!
uid velocity u for simplicity, the average symbol is dropped out
hereafter. Basically the average process in the r.e.v. can be conducted through two distinct ways, the averaged uid velocity over

!
the volume of uid of the r.e.v., and the seepage velocity uf over
the total volume of the r.e.v. including both the uid phase and the
solid phase. The relation between the seepage velocity and the
averaged uid velocity is given by

!
!
uf f u

(1)

The governing equations for mass, momentum and energy

conservation are written as [27]



vrf

!
V$ rf uf 0
vt

Vp 

frf cpf

m
!
r 
!
!
uf  CF pf  uf  uf
K
K

vT
000

!
rf cpf uf VT fV$kVT fq_f
vt

(2)

(3)

(4)

where the so-called DarcyeForchheimer model is used to simplify

the momentum equation as shown in Eq. (3), which has been
widely used for engineering design and analysis. The subscript f
denotes uid quantities. The inertial effects of uid ow are
neglected in Eq. (3) while the friction force and pressure drop
balance each other. This methodology found a great success in
modeling uid ow in granular particulate systems such as a bed of
packed beads [7,18]. However, most of the results are for low
porosity 4  0.6 while metal foams of interest have a high porosity
(0.80) and high tortuosity. It is important to note that the DarcyeForchheimer model here is for homogeneous and isotropic
porous medium to calculate the pressure drop which does not
count for anisotropic and transient scenarios in which a general
form has been used by Yakinthos et al. [34] to model recuperative
heat exchangers for aero engines.
In order to compare empirical results and CFD simulations the
heat exchanger equation for thermal balance is also introduced
below. In general, the interfacial heat transfer coefcient around a
tube depends on the ow dynamics in the metal foam as well as the
boundary conditions on the tube such as the wall, geometry cylinder or square shape. Bearing this mind, the analytical solution can
be obtained from the heat exchanger thermal balance equation,
which is a zero dimensional energy balance equation with heat
exchange between hot solid wall and cold uid only. The heat
exchanger governing equations read [19],

1  frs Cps



vTm
hv Tm  Tf 0
vt
:

Tout  Tin mf Cpf q_

000

(5)

(6)

2. Governing equations for porous medium

Flow through a porous medium, which has been studied in a
variety of applications since Darcy's work in the nineteenth

where subscript f denotes quantities of uid while s stands for

quantities of solid materials. Assuming the tube surface temperature and metal foam temperature are the same, Tm, while the uid

106

S. Mao et al. / Applied Thermal Engineering 71 (2014) 104e118

temperature, Tf, is the average of inlet and outlet air temperature,

i.e., Tf (Tin Tout)/2. The heat ux is calculated by



000
q_ hv V Tm  Tf

(7)

The analytical solution of the volume heat transfer coefcient is

obtained from Eqs. (5)e(7),

1
V
A
hv b1  frs Cps @1 
2m_ f cpf

(8)

In Eq. (8), the constant b is directly related to the transition

solution of heat exchanger thermal balance equation. The correlation relation and CFD results are then compared to the analytic
solutions for typical ACCs scenarios in Section 3 of this paper.
A typical size and layout of metal foams for air-cooled condensers (ACCs) are shown in Fig. 1. It is a single row open-cell metal
foam used for compact heat exchanger with aluminum or copper
for both foams and tubes. The current study considers heat transfer
through the metal foam-tube interface for single phase only;
however, the uid ow in the tube is considered as two-phase with
condensation from vapor to liquid water in the thermal cycle of
power generation. Most empirical correlations of the permeability,
inertial coefcient and effective thermal conductivity available for
regular pore size and distribution (e.g. packed beads) cannot be
directly applied to open-cell metallic/ceramic foams with the
porosity 4  0.8 for applications of heat exchangers [29]. For the
nned tube bundle air-cooled condensers, Hooman and Gurgenci
[16] suggested the following correlation for the form drag coefcient obtained from the CFD modeling of porous medium,

CF 0:55  9:8871  f  f  0:323  0:8443

(9)

However, it is unclear whether this correlation can be adapted

to metal foam air-cooled condensers in general. Experimental and
numerical studies have shown that the form drag coefcient depends on porosity not directly on the pore size and shape (e.g.,
Calmidi and Mahajan [8]). Numerical modeling combined with
experimentation is used to verify empirical and/or semi-empirical
correlations in recent years (e.g. Hugo et al. [17]). Even though
there are lots of data for other applications, in metal foam heat
exchangers for ACC applications, the case of open-cell metal foams
has yet limited success when using numerical approaches due to
the scarcity of data from accurate experimental measurements.
A sequence of test cases about metal foam samples used for heat
exchangers has been selected from available data to compare to
above correlations for permeability and inertial coefcients for heat
exchangers. The best practice of validation will be conducted in the
regime of turbulent ow for most scenarios of the air side in ACCs,
and inlet and outlet boundary conditions will be tuned here to
address the optimal values of interfacial overall heat transfer and
acceptable pressure drop through the porous matrix. Without
diluting our attention, the modeling and simulation are assuming
that the system is under locally thermal equilibrium (LTE) state and
the air ow through the metal foam is uniform and without
boundary effects so that the mass ow rate is obtained by the
production of the front area and the inlet air velocity.
3. Correlation results for metal foam heat exchangers
3.1. Characteristic parameters of metal foam heat exchangers
In order to address metal foam geometric characteristics 30
metal foam samples were taken from available literature and have
been chosen to compare to the relation between the permeability

Fig. 1. A single row metal foam wrapped tube (left) as well as two bulk metallic foam samples (middle and right); A typical pore topology is also shown on the bottom of the central
panel, in which the edge length of an open cell is dp, and the pore ber thickness is df.

S. Mao et al. / Applied Thermal Engineering 71 (2014) 104e118

and porosity. Two types of correlation formulations are studied

here. The rst type is based on CarmaneKozeny model as shown by
Eq. (10) which is widely used to compute the permeability for tube
bundle porous medium for conventional heat exchangers (Tables 1
and 2), one focuses on overall performance (Table 1), the other
highlights individual pore topology and size parameters (Table 2). A
length scale dt, the diameter of the tubes, is needed in the model. In
addition, the constant a is a tting value to be consistent with the
geometry and the structure of the porous medium.

d2t f3

(10)

a1  f2

In tube bank heat exchangers the value for a is typically 100

(e.g., Bejan and Morega, Hooman and Gurgenci, Mao, Cheng et al.
[3,16,24]. Recently, Tadrist et al. [32] applied CarmaneKozeny
model to study the permeability of metal foam heat exchangers and
found a widely disparate value of a 100e865 in Eq. (10). Instead
of the diameter of tube of nned tube structure, the ber thickness
of metal foam, df is used in Eq. (11)

d2f f3

(11)

a1  f2

Fig. 2 shows the distribution of two constants used in the CarmaneKozeny model for metal foam heat exchangers. The factor a is
related to the permeability while factor b is related to form drag
coefcient. The term b is a constant empirical value depending on
material and geometry of metal foams, the empirical formula by
Tadrist et al. [32] is applied here (Eq. (13)). We see that the constant
spreads in a variety of margin from 100 to 400 for a and four times
of magnitude difference for parameter b for two materials in Fig. 2.
In addition, aluminum and copper illustrate distinct sensitivity to
the permeability. This suggests that CarmaneKozeny model may
not be accurate to describe metal foam type of porous medium and
the application of the model should be careful for different
materials.
The second type of correlation directly relies on the pore shape,
size and the relative density rather than that based on empirical
constant as shown in Eqs. (10) and (11). From the engineering
application point of view, it is straightforward to study the microscopic structure of metal foams, which can shed light on the relation between the porosity f and the permeability K. The following
correlations have been used to study metal foams by [12]:

K
f2

2
36cc  1
d

107

Table 2
Mechanical characteristics of metal foam samples used for Eqs. (10) and (11).
df

dp

60 mm
N/A
0.25 mm
0.50 mm
0.35 mm
0.24 mm

30 mm
N/A
3.13 mm
4.02 mm
2.58 mm
1.98 mm

0.71
0.9486
0.9272
0.9726
0.9005
0.952

3:61  105 m2
1:2  107 m2
1:2  107 m2
2:7  107 m2
0:9  107 m2
0:562  107 m2

0.148
0.097
0.097
0.097
0.088
0.0976

permeability than the empirical assumption in Eq. (10) given the

pore geometry of metal foams. Eqs. (10) and (11) do not consider
underlying microscopic structure, which might not be reasonable
for metal foams.
3.2. Correlation of momentum exchange of metal foam heat
exchangers
The structure of the metal foam heat exchanger has two rows of
tubes with parallel layout, slightly different from the layout as
shown in Fig. 1, the diameter of the tube is 30 mm, and the metal
foam wrapping thickness is 30 mm, no gaps between metal foam
wrapping are allowed between two neighboring tubes. In other
words, the metal foam lls all interstices of the tube bundle where
the distance between two neighboring tubes is 60 mm. The total
number of tubes used in this compact heat exchanger is 12 with the
geometry size of 1 m  0.9 m  0.25 m. Despite having the same

(12)

Where the diameter d or the element width dp of the open-cell

metal foam pores, based on the cubic representative unit cells or
CRUCs concept, and the tortuosity c are two unknowns to be
determined before using Eq. (12) or (15) to calculate the permeability or the pressure drop. Overall, Eq. (12) is a better prediction of

Table 1
Metal foam sample material characteristic parameters [23,26].
Material
Sample
Sample
Sample
Sample
Sample
Sample
Sample

1:
2:
3:
4:
5:
6:
7:

Al
Al
Al
Cu
Cu
Cu
Al

Porosity

Fiber diameter (m)

PPI

0.956
0.93
0.896
0.933
0.905
0.936
0.9272

0.000445
0.000324
0.000484
0.0005
0.000403
0.000244
0.00025

10
40
10
5
10
40
10

1.82
0.634
2.65
0.97
1.21
4.5
1.2

0.102
0.086
0.106
0.051
0.056
0.221
0.097

Fig. 2. Correlation factors a and b used for permeability (e.g. CarmaneKozeny model)
and the inertial coefcient (Eq. (13)) in metal foam heat exchangers for aluminum and
copper.

108

S. Mao et al. / Applied Thermal Engineering 71 (2014) 104e118

layout of geometry and material (Cu), the pressure drop illustrates a

big disparate distribution between Sample 4 and Sample 6
(Table 1). Two samples have close porosities, however, the
permeability and form drag coefcients show big difference. The
major difference between Sample 2 (Al) and Sample 4 (Cu) is the
relative density, even though the porosities are close each other. It
is not clear whether the material or the structure play the critical
role to lead a big disparate of pressure drop considering the fact
that both permeability and form drag coefcient cannot explain the
phenomenon. Given the same porosity, or slight change in porosity,
the pressure drop seems constant for different samples at the same
uid ow inlet conditions. Data show signicant change for distinct
relative density and pore geometry of metal foams while the
porosity is very close each other for Sample 2 and Sample 7, both
are Al.
Using these samples we compare four models for the prediction
of pressure drop, Dp, in metal foams, which is quantied p
using
the
form drag coefcient CF or the inertial coefcient F CF = K in the
DarcyeForchheimer model. Alternatively the friction factor has
been used to compute the pressure drop in metal foams. The rst
three are from archived publications and the fourth model is proposed here based on CFD simulation and regression analysis of
legacy data.
The rst model is used by Tadrist et al. [32], which are an
empirical formula for the inertial coefcient,

CF
b1  f

F p
; b 0:65  2:6
f3 df
K

(13)

The subscript f denotes pore ber (Fig. 1) instead of the uid

quantities. The inertia coefcient F is distinguished from the friction factor f discussed in the following. The correlation presented
here provides a concise expression with clear physical meaning, but
needs extensive validation to determine the value of tting constant b for metal foam structures numerically and/or experimentally. After combining Eqs. (11) and (13), the form drag coefcient is
recast,

b 1
CF p 3=2
af

(14)

 


DP
L

p
K

ru2

fk

1
m
0:105 p 0:105
Rek
ru K

(17)

where fk is the familiar friction factor f used p

inFluid Mechanics
textbooks except the characteristic length
K : However, the
empirical constant 0.105 in Eq. (17) has its limitation and cannot be
extended to a variety of scenarios of metal foam heat exchangers
directly.
Fig. 3 shows the comparison among the predictions from four
correlations of the form drag coefcient, which cover a variety of
porosity. It is seen that the correction from Paek et al. [30] is
insensitive to porosity due to an empirical constant. The form drag
coefcient monotonically decreases with the porosity of metal
foam for two models by Tadrist et al. [32] and Du Plessis et al. [12].
Because the correlation in Hooman and Gurgenci [16] was established from CFD simulation based on nned tube heat exchangers,
it exhibits a maximum at a value of K z 0.65. This shows that it is
probably true to strict validate correlation from nned tube heat
exchangers for metal foam topological design.
Fig. 4 shows the change of pressure drop of metal foam heat
exchangers at different Reynolds number (velocities of air ow)
through an ACC unit without considering phase change of the air
side for these cases. Fig. 4 also shows two distinct group data
included in this comparison. The rst group, the very left three
cases for the same metal foam sample. The only change is the air
ow rate. It is seen as a non-linear relation between the pressure
drop and the air velocity, which is close to a quadratic curve. The
second group, four cases shown in the middle of the gure, compares how different materials and metal foam geometries inuence
the pressure drop. For the same air ow rate, the change in
magnitude of the pressure drop can be as much as twice as large
among the four cases, again showing the importance of metal foam
geometry for system optimization design.
Following the notation in nned tube heat exchangers, the hydraulic diameter is dened as four times the ow passage volume
divided by the total heat transfer area:

Dh

4LAmin
A

(18)

The second type of pressure drop correlation is from Du Plessis

[12],

CF

p
2:05cc  1 K
d
f2 3  c

(15)

where the diameter or the width of the open-cell metal foam pores,
based on the cubic representative unit cell or CRUC concept, and
the denition of tortuosity c are the same as before. Because both
the length scale and the tortuosity are not empirical parameters, it
can be directly scaling up to large-scale engineering applications.
For a given metal foam product, the width of open-cell pores is not
difcult to obtain. After using Eq. (12) to replace the tortuosity c,
the above formula becomes,

p
1 2:05 cc  1
2:05d
CF

q!
f
3c
p
2 2
6 K 15  9 f Kd

(16)

Eq. (16) provides alternative approach to calculate the form drag

coefcient without knowing the tortuosity directly.
The third type is from Paek et al. [30] for the friction factor
correlation,

Fig. 3. Comparison of the form drag coefcient CF from four models. The Hooman and
Gurgenci [16] correlation results were based on nned tube heat exchangers. The other
three were based on metal foam porous medium.

S. Mao et al. / Applied Thermal Engineering 71 (2014) 104e118

109

where the above friction factor is referred to as Darcy friction

factor, which has a relation with Fanning friction factor
f 4fFanning. Compared to the empirical correlation in Eq. (17), a
general correlation considering the contribution of linear and
nonlinear viscous forces is obtained here. Eq. (20) implies that
both permeability K and the inertia coefcient F or form drag
coefcient CF evolve only with the solid matrix, here the pore
diameter, independent of the ow eld. Furthermore, if the
permeability and form drag coefcient are constant, the friction
factor f is only a function of velocity or Reynolds number. The
friction factor introduced by Paek et al. [30]; fK, has the following
relation with f,

2[
fk f p
K

(21)

Now if we substitute CarmaneKozeny model into Eq. (20) to

replace the permeability, we have the friction factor below,
Fig. 4. Pressure drop versus uid ow velocity for seven metal foam samples (out of
the 30 samples) used in the correlation study.

where L is the ow length of the metal foam matrix, the thickness

of the heat exchanger along the air ow direction, LAmin is the
minimum free ow passage volume and A is the total heat transfer
area. Without considering the metal foam, the tube diameter of
heat exchangers can be chosen as the characteristic length [ Dh;
while the pore diameter can be used as the characteristic length for
metal foam heat exchangers [ DP. The Reynolds number is dened
based on the characteristic length Re rair[uair/mair. From the
denition of friction factor f, we have

Dp
f
 rair juair juair
Dx
2[

(19)

where uair is the averaged velocity of uid u instead of the seepage

velocity of uid uf. The generic correlation for the friction factor can
be obtained using DarcyeForchheimer model, Eq. (3),

f f2 CF

2[
K

1
2

2[2 1
K Re

(20)

f ab

1
Re

(22)

where parameters a and b are constants, independent of the ow

eld. The tting constants are a 1.9, b 125.0 (Fig. 6). Thispwould

recover the relation in Eq. (17), considering the fact that K is a

length scale with unit of m. In general, the friction factor can be
recast as

a
b

Re Rec

(23)

Therefore the correlation has three tting parameters, a, b and c,

constant at a given ow regime. For low Reynolds number ow, we
can use the following coefcients through regression analysis

a 25:6; b 0:50; and c 0:04

(24)

Figs. 5 and 6 illustrate the comparison between the general

correlation Eq. (20) and empirical formulas in Eqs. (21)e(23). The
classic results by Ergun for packed beads were included in Fig. 5 to
illustrate the difference between the general correlation for metal
foam and that obtained from granular materials. The tting parameters in Eq. (22) overall show a good agreement with available
experimental data for both low and high Reynolds number for
metal foam heat exchangers. It also conrms that the ow friction
factor is only a function of the porosity of porous media, not directly
related to the microscopic shape such as lament diameter, size,
and the relative density PPI.

3.3. Correlation for the interfacial heat transfer coefcient of metal

foam heat exchangers

Fig. 5. Darcy friction factor used for nned tube heat exchanger and comparison to
packed bed results [13]. The involved characteristic length is the diameter of the tube
dh.

Heat transfer in forced ow in metal foam heat exchangers is

even more complicated than nned tube heat exchangers due to
the fact that a signicant thermal dispersion, the so-called hydrodynamic mixing of interstitial uid, exists at the pore scale. The
correlation of heat transfer for bared tubes may not be used directly
in metal foam heat exchangers due to the thermal dispersion effect.
Another challenge for heat transfer correlation is the experimental
difculties of accurate measurement in porous media environment.
From an application point of view, simple correlations could be
obtained through analog to bared tubes. For an internal ow,
considering a channel packed with a porous medium for fully
developed ow, the relative heat transfer augmentation effect is
written as follows [27],

110

S. Mao et al. / Applied Thermal Engineering 71 (2014) 104e118

Fig. 6. Comparison of Darcy friction factor among experimental data and correlation results for metal foam heat exchangers. The characteristic length used here is the pore
thickness dp.

hsurf with porous media

km
z
hsurf without porous media kf

(25a)

where km (1  f)ks fkf is the effective thermal conductivity,

and kf is uid thermal conductivity. However, it is not clear whether
the heat transfer augmentation for an external ow in porous
media observes a similar relation or not.
The macroscopic correlation of the volumetric heat transfer
coefcient hv (W/m3 K) or the local interfacial heat transfer coefcient hs (W/m2 K) is presented as Nusselt number versus Reynolds
number and Prandtl number as follows

Nus c0 c1 Ren Pr 1=3

(25b)

where c0 and c1 are two tting constants for a range of Reynolds

number, which covers the limit value c0 for a laminar ow in porous
media, while the power n 0.3e0.8 counts for different situations.
For intermediate or high Reynolds number ow, the constant c0 can
be negligible, Eq. (25b) becomes the classic formula [5],

Nus 0:023Re0:8 Pr 1=3

Nus 0:418Re0:53 Pr 1=3

(27)

where Eq. (27) is suitable for ow 30  Re  200. The denition of

Reynolds number and Prandtl number are same as before in Eq.
(26).
In order to validate the general correlation in Eq. (25) and
address the differences among different heat transfer coefcient
predictions, a total of 23 cases of metal foam samples (out of the 30
samples) have been used in heat exchangers.
Fig. 7 shows the change of the interfacial heat transfer coefcient hs with the air ow velocity while other ow conditions
remain the same for all metal foam samples. It is seen that, when
Reynolds number Redf increases, the interfacial heat transfer coefcient increases for the same metal foam sample monotonically.
However, it illustrates a distinct distribution of the interfacial heat
transfer coefcient for the same Reynolds number, e.g. sample #2
and #3 are less than half value of the hs obtained in sample #5 and
#6. On the other hand, the interfacial heat transfer coefcient increases drastically with high air ow rate at a nearly linear relation
for sample #2 and #3. From the optimal design point of view, there

(25c)

Several specic correlations have been established in past

decade for typical metal foam samples with given porosities so far.
For example, the following metal foam-tube interfacial heat
transfer relation was correlated by Calmidi and Mahajan based on
wind-tunnel experiments [8],

Nus

hs df
uf df
mC
0:37
; Pr ; CT 0:52
CT Re0:5
; Redf
df Pr
kf
kf
fn

(26)

where the subscript s denotes the quantity on the interface between the metal foam and the tubes. The parameter hs is evaluated
at the interface between a tube and the surrounding wrapped
metal foam. The constant c is the specic heat of air, CT is a correlation constant, the velocity uf again denotes the Darcy velocity or
the seepage velocity, and uf/f is the intrinsic velocity of uid ow
in porous medium. The value of df is determined by the diameter of
pore ber (Fig. 1), m and n are the kinetic and kinematic viscosity of
air, respectively. Recently, Mancin et al. [23] introduced the
following heat transfer estimate from experimental and modeling
results,

Fig. 7. The heat transfer coefcient at the metal foam-tube interface versus uid ow
velocity for seven metal foam samples in the correlation study.

S. Mao et al. / Applied Thermal Engineering 71 (2014) 104e118

is a trade-off between the improvement of overall heat transfer

effect and the acceptance of total pressure drop through the metal
foam matrix. In order to address this relation, the metal foam heat
exchanger, conventional type nned tube heat exchanger, and
simple tube heat exchanger without ns are compared for two
major thermal-hydraulic features. Finally, we see that the interfacial heat transfer coefcient reported in the literature show a wide
range of variability which results in a large certainty of quantication of metal foam heat transfer capability.
Fig. 8 includes 21 from the total 30 samples by removing
redundant tests for pressure drop and heat transfer considering the
optimal design of heat exchanger performance curve. By changing
the air side ow rate, better heat transfer is obtained with the price
of increase of pressure drop. One thing that should be avoided is
set-up of the operating point above the threshold of performance
curve in Fig. 8 (the performance curve was not shown here). The
rate of change in heat transfer is far smaller than the rate of change
in pressure drop if the operating point above the threshold. It is
seen that data for metal foam can be separated into three groups:
the four points on the top right corner; ve to six points in the
middle and the left clustered points. Samples #1 & #2 in Figs. 4 and
7 represent the rst group, Samples #3 & #4 (Figs. 4 and 7) for the
middle and Samples #5-#7 (Figs. 4 and 7) for the last group. It is
important to note that the correlation between pressure drop and
the heat transfer not only depends on the ow rate and distribution
of pressure, temperature, but also on the metal foam materials and
geometric parameters.
The correlation of interfacial heat transfer coefcient was
adapted for metal foams as well as classic results of Eq. (25c) [5] in
Fig. 9 for comparison. In order to focus on uid physics, the Prandtl
number is chosen as a constant, i.e. Pr 0.72 for all cases here. The
Reynolds number of air ow again used the pore thickness df as the
characteristic length. It is seen that the classic correlation Eq. (25c)
cannot be directly applied to metal foam heat exchangers due to the
Reynolds number is far above the range in the metal foam. The
correlations between Calmidi et al. [8] and Mancin et al. [23] show a
close result even though different tting parameters have been
used by two research groups. When Reynolds number increases,
the three models are close to each other. At the low limit, the
prediction shows large differences using the classic model. As a
nal comment, the 0D model or heat exchanger thermal balance
equation leads to over-prediction of results compared to the other
three correlations, in particular, at the high Reynolds number ow.

Fig. 8. Comparison of heat transfer and pressure drop change for different type of heat
exchangers (metal foam, nned tubes and bared tubes) with the equivalent geometric
size and the layout of tube bundles.

111

Fig. 9. Nusselt number versus Reynolds number for different correlation models as
well as the analytical solution.

For this reason, two dimensional and three-dimensional CFD simulations are needed to validate empirical or semi-empirical models.
4. Numerical results for metal foam heat exchangers
In this section commercial CFD tool ANSYS FLUENT has been
applied to simulate two scenarios at the stage of conceptual design
of metal foam heat exchanger for ACCs. The rst scenario is partially
wrapped metal foam around each tube while the second scenario
will be fully wrapped metal foam around each tube in the core area
of the heat exchangers. Both metal foam and tubes are aluminum.
The geometry size of the metal foam heat exchanger unit is
1.0 m0.2 m0.55 m. There are total 112 tubes in the unit with
30 mm for the inner diameter D and 1.6 mm thickness for each
tube. Fig. 10 illustrates the rst scenario as well as the computational set-up. The staggered layout of tubes is considered here to
enhance ow mixing. The pitch L0 between two neighboring tubes
is 60 mm, the wrapper depth of aluminum metal foam is
h 10 mm, The tube center distance from the edge of the core area
L1 is 50 mm. The geometry, layout and size for the second scenario
are the same as the rst scenario except fully wrapped metal foam
in the core size of heat exchangers. Three samples of aluminum
metal foam (Table 3), total of six cases, are simulated. In each case,
the air ow is from the left side to the right side of the computational domain, while a mass ow rate boundary condition is
imposed at the inlet and an outow boundary condition at the
outlet, respectively. The top and bottom are periodic boundary
conditions to save computing time. All six cases are focusing on the
steady-state elds and the distribution of pressure drop and heat
transfer coefcient will be obtained for steady ows only.
The following assumptions are considered for all cases. First, the
air ow is assumed uniform along the front area, the use of
isotropic porous medium in principle will neglect the nonuniform
air ow. The inlet air velocity for the baseline is 5.5 m/s which is set
as the reference value. Second, the DarcyeForchheimer model is
applied to calculate pressure drop through trial and error approach
to determine the form drag coefcient CF in Eq. (3). Because Table 3
has already provided the inertial coefcient for each metal foam
sample, the trial and error iteration that is needed in most heat
exchanger simulation is skipped by direct using the quantity provided here. Third, a methodology based on non-dimensional

112

S. Mao et al. / Applied Thermal Engineering 71 (2014) 104e118

Fig. 10. CFD simulation setup for the rst scenario, partially wrapped metal foam is considered (h is the depth of the wrapper around each tube).

Table 3
Permeability and inertia coefcient of three samples used in the simulations.
Property

Symbol

Sample 1

Sample 2

Sample 3

Porosity
Permeability
Inertia Coefcient

0:90
6:6  108
389

0:91
4:79  109
1088

0:95
7:2  108
1107

K(m)2
b(m1)

Nusselt number is used to predict the local heat transfer coefcient

[24]. Finally, the CFD simulation focuses on the air side ow and
heat transfer without considering phase change in the vapor side of
ACCs, for simplicity and without loss of generality, a constant

temperature on the tube wall is assumed for all tubes,

w
Tw
in Tout 348 K. In other words, we only consider the phase
change that vapor condenses to liquid water without considering
any supercritical heat in the tube. The ambient air temperature is
assuming a constant, 298 K while the ambient pressure is 1 atm. For
turbulent ow simulation in porous medium, the standard k 
model is used with recommended parameters in metal foam matrix, even though other turbulent models can be used [11]. Finally,
each metal foam heat exchanger is simulated for different vair
in ,
Re 4.0  104e5.0  105 to compare with the baseline velocity
v 5.5 m/s.

Fig. 11. The baseline mesh (top, 170421 cells in 2D) and ner mesh (bottom, 362572 cells in 2D) for the rst scenario simulation as shown in Fig. 10.

S. Mao et al. / Applied Thermal Engineering 71 (2014) 104e118

113

Fig. 12. Comparison of x direction velocity vectors for the rst scenario (partially wrapped metal foam): (a) Sample 1 and (b) Sample 3.

In addition to the momentum equation for CFD simulation, two

additional terms are used to represent the effects of porous medium on air ows, the effective thermal conductivity keff (1  f)
ks fkf is applied in the energy equation with ks and kf represents
thermal conductivity of metal foam and the air, respectively.
Different porosity will lead to slightly different effective thermal
conductivity for the same materials. After using an appropriate
temperature for the heat exchanger tubes, the heat ux between
the heat exchanger and the dry air ow are computed locally as

q_ hlocal T  Tw Asf

(28)

where Asf is the heat exchanger surface area per volume (m2/m3).
Finally, the pore size of the metal foam samples varies from 5 to
20 PPI (number of pores per inch) so the mesh size in both 2D and
3D CFD simulation will be consistent with this value. The minimum
mesh size varies from 0.05 mm to 0.2 mm for the 6 test cases. Nonuniform body-tted stretching meshes are generated using commercial tools. Smaller mesh sizes are applied in the region of metal
foam-tube interfaces and in the region of metal foam-space interfaces. In order to test mesh independence, three meshes A
(coarse), B (intermediate), and C (ne) have been generated. The
numerical results are almost unchanged between mesh B and mesh
C. For this reason, the intermediate mesh B will be used as the
baseline in all six test cases. Fig. 11 shows two mesh proles (B and
C).
Fig. 12 illustrates the velocity vectors for the rst scenario
(Sample 1 and 3 in Table 3), the only difference among three metal
foam samples are the air ow in the gap between wrapped tubes.
The lower porosity in solid matrix, the more air will be squeezed to
the gap. This is understandable considering that low porosity lead
to less ow passage inside the solid matrix. However, the effect on
heat transfer is more complicated than the distribution of ow
volume due to the fact that thermal dispersion is anisotropic in the
domain. This partial wrapping will allow for the thermal design to

take into account more options while more difcult to predict using
conventional empirical correlations. A quantication comparison
among velocity change is shown in Fig. 13, in which velocity
magnitude was computed along the line that cuts through the
centers of two bottom cylinders. Before and after the tube bank the
prole is similar, however, a big change exhibits between the two
tubes among three cases. Fig. 14 shows the steady-state static
temperature and pressure contours for the rst scenario (Sample 1
in Table 3). Because of the staggered layout used in the design, the
distribution of temperature and pressure are not symmetric in the y
direction (perpendicular to the air ow direction). The advantage is
a possible enhancement of ow mixing, so-called thermal dispersion, and heat transfer between solid tube/matrix and uid. In order to estimate the overall heat transfer and the pressure drop, the
area-weighted average surface heat transfer coefcient and total
pressure drop are output for three samples. As a comparison, Fig. 15
shows the static pressure contours for fully wrapped metal foam
heat exchanger scenario (Sample 1 and 3 in Table 3). It is seen that
larger temperature gradient for partially wrapped metal foam
(Fig. 14) across the tube/metal foam matrix, which may cause
higher ow and heat transfer mixing. On the other hand, the total
surface area per unit volume (m2/m3) is much higher in Fig. 16 than
that in Fig. 14. From the optimization point of view, there may be
critical value of the wrapped depth for metal foam matrix.
Fig. 17 shows the change of heat transfer coefcient hf versus
porosity f (right panel) and for pressure drop (left panel) versus
porosity f. It is seen that the porosity reduced from 0.95 to 0.91, the
pressure drop increased from 236.05 Pa to 358.97 Pa (52.07% gain),
while the surface heat transfer coefcient increased 550.42 W/m2 K
to 825.32 W/m2 K (49.9% gain). However, change of porosity from
0.91 to 0.90, the gain of heat transfer is 7:7% and 2:4% of pressure
drop, respectively. The cause of the different trend is due to the fact
that metal foam is partially wrapped around each tube in the
domain, a challenge facing the thermal-hydraulic design of metal
foam heat exchangers.

114

S. Mao et al. / Applied Thermal Engineering 71 (2014) 104e118

Fig. 13. The velocity magnitude prole along the line which cuts the centers of the two cylinders, (a) f 0.90, (b) f 0.91 and (c) f 0.95.

In order to quantify the impact of porosity on the ow and heat

transfer in metal foam heat exchangers, Fig. 18 shows the prole of
static pressure and static temperature along the air ow direction
(horizontal direction) for fully wrapped metal foam. In all three
samples (Table 3), the pressure proles are widely spread even
though the porosity slightly changes from 0.90 to 0.91. However,
the temperature changes are less sensitive, only a large variant of
static temperature occurred when the porosity increased from 0.91
to 0.95. An interesting conclusion here is that the results are

consistent with the 0D model discussed before. It is fair to say that

the sensitivity of pressure drop to the porosity is higher than the
interfacial heat transfer coefcient for metal foam heat exchangers,
either partially or fully wrapped structure.
5. Conclusion
Extensive parameters observed in this study have been tested
for metal foam heat exchangers in order to optimize the overall

S. Mao et al. / Applied Thermal Engineering 71 (2014) 104e118

115

Fig. 14. Steady state temperature distribution (top) and static pressure contours (bottom) for the rst scenario (Sample 1 in Table 3).

heat transfer capability and the pressure drop in the air-side of aircooled condensers (ACCs). A variety of metal foam samples have
been selected from archived literature in recent years to compare to
2D/3D CFD simulations using the porous medium model. The form

drag coefcient CF was found to be insensitive to the metal foam

pore size and its shape but only a function of the porosity f of the
metal foam sample. The CarmaneKozeny model that has been
widely applied in nned tube heat exchangers design and analysis

Fig. 15. Steady state pressure distribution for the second scenario (fully wrapped metal foam heat exchanger): (a) Sample 1 and (b) Sample 3.

116

S. Mao et al. / Applied Thermal Engineering 71 (2014) 104e118

Fig. 16. Static temperature distribution for the second scenario (fully wrapped metal foam heat exchanger): (a) Sample 1 and (b) Sample 3.

needs further validation before being directly applied for metal

foam heat exchanger design and analysis. The interfacial heat
transfer coefcient obtained from CFD simulation is in good
agreement with the empirical correlation as given by Calmidi and
Mahajan [8] for low Reynolds numbers, however, it shows
discrepancy for turbulent ow with high Reynolds numbers. It is
seen a similar trend between the analytic solution of surface heat
transfer coefcient to the area-weighted average heat transfer coefcient from CFD simulations.
The overall objective of this paper was to determine a correlation between pressure drop and heat transfer which can be used to
assist in the design on metal foam heat exchangers for ACCs. For
this purpose a methodology to compare metal foams to nned heat
exchangers was developed and these evaluation criteria have been

made available to those interested in the design and prototyping of

metal foam heat exchangers. New empirical correlations have been
suggested in this study from the parametric studies to better t in
the available experimental data from the published literature and
numerical results from CFD simulations based on the porous medium model. These include the Darcy friction factor Eqs. (22) and
(23) obtained by least squares regression analysis, and the surface
heat transfer coefcient through the consistence comparison between Eqs. (25) and (26) and an analytical solution. It is seen that
classical packed beads and other granular systems may not be accurate for metal foam structures. There is a trade-off between the
increase of overall heat transfer and the acceptance of pressure
drop for a given ow eld. From the optimal design point of view,
the metal foam thickness, wrapped depth around tubes, and other

Fig. 17. Area-weighted average pressure drop (left panel) and surface heat transfer coefcient (right panel) versus porosity for the rst scenario.

S. Mao et al. / Applied Thermal Engineering 71 (2014) 104e118

117

Fig. 18. Pressure drop at different position along air ow direction (top panel); total temperature change along air ow direction (bottom panel) for the fully wrapped metal foam
scenario. Case 1: f 0.90, Case 2: f 0.91 and Case 3: f 0.95.

parameters could cause signicant impact on both heat transfer

and pressure drop.
Finally, the numerical modeling of the related metal foam heat
exchangers suggests that a higher overall heat transfer capacity of
2e3 times compared to conventional nned tube heat exchanger
with the same number of tubes and geometry size and layout can
be obtained. The pressure drop from metal foam is in the acceptable
range of the optimizing design to use in the thermal power plant for
dry-cooling applications. Furthermore simulation studies are
needed to accurately address the thermal optimization of the
system.
References
[1] R. Aull, Cooling Tower Technology Advances, NSF-EPRI Joint Workshop on
Advancing Power Plant Water Conserving Cooling Technologies, ASME 2012
IMECE, Houston, Texas, Nov 9e15, 2012.
[2] J. Bear, Dynamics of Fluids in Porous Media, Elsevier, New York, 1972.
[3] A. Bejan, A.M. Morega, Optimal arrays of pin ns and plate ns in laminar
forced-convection, ASME J. Heat Transf. 115 (1) (1993) 75e81.
[4] A. Bhattacharya, V.V. Calmidi, R.L. Mahajan, Thermophysical properties of high
porosity metal foams, Int. J. Heat Mass Transf. 45 (2002) 1017e1031.
[5] R.B. Bird, W.E. Stewart, E.N. Lightfoot, Transport Phenomena, second ed., John
Wiley & Sons, 2002.

[6] K. Boomsma, D. Poulikakos, F. Zwick, Metal foams as compact high performance heat exchangers, Mech. Mater. 35 (2003) 1161e1176.
[7] L. Bottura, C. Marinucci, A porous medium analogy for the helium ow in
CICCs, Int. J. Heat Mass Transf. 51 (2008) 2494e2505.
[8] V.V. Calmidi, R.L. Mahajan, Forced convection in high porosity metal foams,
ASME J. Heat Transf. 122 (2000) 557e565.
[9] A.E. Conradie, D.G. Kroger, Performance evaluation of dry-cooling systems for
power plant applications, Appl. Therm. Eng. 16 (3) (1996) 219e232.
[10] A. Dai, K. Nawaz, Y. Park, Q. Chen, A.M. Jacobi, A comparison of metal-foam
heat exchangers to compact multilouver designs for air-side heat transfer
applications, Heat Transf. Eng. 33 (1) (2012) 21e30.
[11] M.J.S. de Lemos, Turbulence in Porous Media: Modeling and Applications,
second ed., Elsevier, 2012.
[12] P. Du Plessis, A. Montillet, J. Comiti, J. Legrand, Pressure drop prediction for
ow through high porosity metallic foams, Chem. Eng. Sci. 49 (21) (1994)
3545e3553.
[13] S. Ergun, Fluid ow through packed columns, Chem. Eng. Prog. 48 (1952)
89e94.
[14] N.C. Gallego, J.W. Klett, Carbon foams for thermal management, Carbon 41
(2003) 1461e1466.
[15] O. Gerbaux, F. Buyens, V.V. Mourzenko, A. Memponteil, A. Vabre, J.F. Thovert,
P.M. Adler, Transport properties of real metallic foams, J. Colloid Interface Sci.
342 (2010) 155e165.
[16] K. Hooman, H. Gurgenci, Porous medium modeling of air-cooled condensers,
Transp. Porous Med. 84 (2010) 257e273.
[17] J.M. Hugo, E. Brun, F. Topin, Metal foam effective transport properties, in:
A. Ahsan (Ed.), Evaporation, Condensation and Heat Transfer, InTech, 2011.
ISBN 978-5307-583-9.

118

S. Mao et al. / Applied Thermal Engineering 71 (2014) 104e118

[18] M. Kaviany, Principles of Heat Transfer in Porous Media, second ed., SpringerVerlag, New York, 1995.
[19] S. Kaka, H. Liu, A. Pramuanjaroenkij, Heat Exchangers: Selection, Rating, and
Thermal Design, third ed., CRC Press, 2012.
[20] T. Lu, Ultralight porous metals: from fundaments to applications, ACTA Mech.
Sin. 18 (5) (2002) 457e479.
[21] S. Mahjoob, K. Vafai, A synthesis of uid and thermal transport models for
metal foam heat exchangers, Int. J. Heat Mass Transf. 51 (2008) 3701e3711.
[22] S. Mancin, C. Zilio, A. Cavallini, L. Rossetto, Pressure drop during air ow in
aluminum foams, Int. J. Heat Mass Transf. 53 (2010) 3121e3130.
[23] S. Mancin, C. Zilio, A. Diani, L. Rossetto, Air forced convection through metal
foams: experimental results and modeling, Int. J. Heat Mass Transf. 62 (2013)
112e123.
[24] S. Mao, C. Cheng, X.C. Li, E.E. Michaelides, Thermal/structural analysis of
radiator for heavy-duty truck, Appl. Therm. Eng. 30 (2010a) 1438e1446.
[25] S. Mao, Z.G. Feng, E.E. Michaelides, Off-highway heavy-duty truck underhood
thermal analysis, Appl. Therm. Eng. 30 (2010b) 1726e1733.
[26] K. Nawaz, J. Bock, Z. Dai, A. Jacobi, Experimental studies to evaluate the use of
metal foams in highly compact air-cooling heat exchangers, in: International
Refrigeration and Air Conditional Conference, Purdue University, 2010. Paper
1150.
[27] D.A. Nield, A. Bejan, Convection in Porous Media, fourth ed., Springer, New
York, 2013.
[28] M. Odabaee, K. Hooman, Application of metal foams in air-cooled condensers
for geothermal power plants: an optimization study, Int. Commun. Heat Mass
Transf. 38 (2011) 838e843.

[29] B. Ozmat, B. Leyda, B. Benson, Thermal application of open cell metal foams,
Mater. Manuf. Process. 19 (5) (2004) 839e862.
[30] J.W. Paek, B.H. Kang, S.Y. Kim, J.M. Hyun, Effective thermal conductivity and
permeability of aluminum foam materials, Int. J. Thermophys. 21 (2) (2000)
453e464.
[31] J. Shi, S. Bushart, EPRI Technology Innovation Water Conservation Program
Overview, NSF-EPRI Joint Workshop on Advancing Power Plant Water
Conserving Cooling Technologies, ASME 2012 IMECE, Houston, Texas, Nov
9e15, 2012.
[32] L. Tadrist, M. Miscevic, O. Rahli, F. Topin, About the use of brous materials
in compact heat exchangers, Exp. Therm. Fluid Sci. 28 (2e3) (2004)
193e199.
[33] Z. Wu, C. Caliot, G. Flamant, Z. Wang, Numerical simulation of convective heat
transfer between air ow and ceramic foams to optimize volumetric solar air
receiver performances, Int. J. Heat Mass Transf. 54 (2011) 1527e1537.
[34] K. Yakinthos, D. Missirlis, A. Sideridis, Z. Vlahostergios, O. Seite, A. Goulas,
Modeling operation of system of recuperative heat exchanger for aero engine
with combined use of porosity model and thermo-mechanical model, Eng.
Appl. Comput. Fluid Mech. 6 (4) (2012) 608e621.
[35] B.J. Yang, S. Mao, O. Altin, Z.G. Feng, E.E. Michaelides, Condensation analysis of
exhaust gas recirculation (EGR) system for heavy-duty trucks, ASME J. Therm.
Sci. Eng. Appl. 3 (4) (2011) 0410007.
[36] C.Y. Zhao, Review on thermal transport in high porosity cellular metal foams
with open cells, Int. J. Heat Mass Transf. 55 (2012) 3618e3632.