International Journal of Heat and Mass Transfer 113 (2017) 1031–1042

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer

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

Melting behaviors of PCM in porous metal foam characterized by fractal

geometry
Zilong Deng b, Xiangdong Liu c, Chengbin Zhang b, Yongping Huang b, Yongping Chen a,b,c,⇑
a
Jiangsu Key Laboratory of Micro and Nano Heat Fluid Flow Technology and Energy Application, School of Environmental Science and Engineering, Suzhou University of Science
and Technology, Suzhou, Jiangsu 215009, PR China
b
Key Laboratory of Energy Thermal Conversion and Control of Ministry of Education, School of Energy and Environment, Southeast University, Nanjing, Jiangsu 210096, PR China
c
School of Hydraulic, Energy and Power Engineering, Yangzhou University, Yangzhou, Jiangsu 225127, PR China

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

Article history: The fractal Brownian motion is introduced to describe the pore distribution of porous metal foam. By the
Received 24 October 2016 fractal description, a model of melting heat transfer for the phase change material (PCM) is developed and
Received in revised form 7 May 2017 applied to investigate the melting behaviors in porous metal foam with a particular focus on the role of
Accepted 28 May 2017
pore distribution. The dynamic response of temperature and the evolution of melting front are presented.
The effects of porosity and fractal dimension on the melting heat transfer are all examined and investi-
gated. In addition, an experiment of melting heat transfer is performed to verify the present model. The
Keywords:
results indicate that the porous metal foam is of significance for the enhancement of melting heat trans-
Melting
Phase change
fer. When compared with PCM alone, the melting process in porous metal foam possesses the larger
Metal foam melting rate, the faster evolution of melting front and the higher liquid fraction. Unlike the PCM alone,
Fractal the melting front is no longer continuous and many independent solid-liquid interfaces are formed inside
pores owing to the interstitial heat transfer. Interestingly, the melting phase change is also affected by the
fractal dimension even though the porosity is identical. A porous metal foam with smaller fractal dimen-
sion is beneficial for the melting heat transfer.
Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction to fundamentally understand the melting behaviors of PCM in por-

The melting phase change is of great interest in a broad range of
engineering applications, including the thermal energy storage
[1,2], electronics cooling [3,4] and solar energy utilization [5,6],
building heating [7]. In the context of real application, the chal-
lenge consists in how to improve the heat conduction performance
of phase change material (PCM) during the melting process owing
to the disadvantage of low thermal conductivity [8,9]. One effec-
tive way to enhance the melting heat transfer is the utilization of
porous metal foam [10]. Unlike the melting of PCM alone, the pres-
ence of porous matrix accelerate the melting rates and may change
the evolution of the melting front [11]. The pore distribution,
although often not properly known, could highly affect the melting
behaviors in porous metal foam. In the context, it is of significance
⇑ Corresponding author at: Jiangsu Key Laboratory of Micro and Nano Heat Fluid
Flow Technology and Energy Application, School of Environmental Science and
Engineering, Suzhou University of Science and Technology, Suzhou 215009, PR
China.
E-mail address: ypchen@mail.usts.edu.cn (Y. Chen).
ous metal foam, especially the role of pore distribution.
Aiming to enhance the phase change heat transfer of PCM, there
have been extensive literatures on the subject of solidification and
melting heat transfer in metal foam using the experimental obser-
vation [2,12,13] and numerical simulation [12,14–17]. Buonomo
et al. [18] focused on a numerical study on latent thermal storage
application via metal foam filled with nano-PCM, and pointed out
that the metal foam greatly increases the charging or discharging
rate while the nanoparticles only moderately adjusts the storing
and releasing thermal processes. The similar performance of
nano-PCM is also observed by Colla et al. [19], when they con-
ducted a numerical and experimental investigation with a focus
on the application of Nano-PCM for electronics cooling. Alshaer
et al. [20–22] systematically investigated the thermal performance
of the carbon foam filled with PCM and different composite in ther-
mal management of electronic systems. In their numerical and
experimental studies, they chose two carbon foam materials
(CF20 and KL1-250) with different thermal conductivities. Two
kinds of composites (Nano carbon tubes and multi wall carbon
nanotubes) are added to PCM (RT65) as a thermal conductivity
enhancer. Their researches pointed out that the improvement in

http://dx.doi.org/10.1016/j.ijheatmasstransfer.2017.05.126
0017-9310/Ó 2017 Elsevier Ltd. All rights reserved.
1032 Z. Deng et al. / International Journal of Heat and Mass Transfer 113 (2017) 1031–1042
Nomenclature
Amush a constant
cp specific heat capacity
fH(t) function
H Hurst exponent
h enthalpy
L length (m)
Ng grid number
n normal direction
p liquid pressure
S source term
t time
u, v velocity
W width
x, y x, y-directions
Greek symbols
b liquid fraction
C indicate the interface
d thermal expansion coefficient
hav dimensionless temperature
thermal conductivity of porous media or filling PCM always
increases the effectiveness of thermal management module. Nada
and Alshaer [23] also found that increasing carbon foam porosity
leads to an increase of the module temperature and a delay of
the approaching steady state. It is worth to note that infiltration
of PCM into highly conductive porous media usually goes with
the formation of voids as micron-size air bubbles within pores,
which contributes to bad thermal performance. Sedeh and Kho-
dadadi [24,25] numerically studied this process and investigated
the effect of voids on solidification of PCM. They pointed out that
the presence of voids within the pores causes a noticeable increase
in solidification time, with a sharper increase for cases without
thermo capillary convection.
In previous studies, it is generally accepted that the solidifica-
tion and melting in porous metal foam is dependent on the poros-
ity, Rayleigh number, thermal properties of the PCM and porous
matrix. The special attention is paid to the importance of porous
metal foam in the enhancement of phase change, in which the
enhancement is typically interpreted in term of the increase in
thermal conductivity [26]. Note that the interstitial heat transfer
between the PCM and the porous matrix is the critical core process
during the melting phase change in metal foam [11]. However, in
the available investigations, the simple assumption of thermal
equilibrium between the PCM and the porous matrix is typically
applied in the melting model of metal foam [14–16]. As a result,
the interstitial heat transfer is not well understood. Up to now, it
is still unclear how the pore distribution affects the melting heat
transfer in metal foam.
The melting phase change in porous metal foam is a complex
unsteady heat transfer process. In this process, the coupled heat
transfer mechanisms, including the heat conduction of porous
matrix, the heat conduction and phase change of PCM and the nat-
ural convection of liquid PCM, are involved. In addition, the melt-
ing phase change are associated with the movement of melting
front. To reveal the mechanism of melting in porous metal foam,
it is necessary to quantify the dimension and spatial distribution
of pores in metal foam (i.e. pore distribution). It is demonstrated
that the pore distribution in porous structure performs random,
multiscale and statistical self-similar properties [27,28]. However,
in the available theoretical investigations, the matrix and PCM is
typically regarded as a homogeneous mixture with averaged prop-
k thermal conductivity
l viscosity
n a computational constant
q density
r2 variance
s time
Subscripts
Exp experiment
l liquid state
p phase change material
Num numerical results
s solid state
Abbreviations
FBM fractal Brownian motion
PCM phase change material
erties [14–16], indicating that only the porosity is considered in the
theoretical model of melting phase change in metal foam. This
implies that the available theoretical model doesn’t consider the
exact nature of random, multiscale and statistical self-similar for
the pore distribution of metal foam. The pore distribution of por-
ous matrix is an important determinant of heat conduction across
the porous metal foam, which has an important effect on the melt-
ing phase change. In this particular situation, it is urgent to develop
a proper characterization of the pore distribution of porous metal
foam taking into consideration random, multiscale and statistical
self-similar properties.
Inspired by the successful utilization of fractal geometry for the
exploration of flow properties in porous media [29], the fractal
Brownian motion is introduced in this paper to describe the pore
distribution of metal foam. By the fractal description, a theoretical
model of melting phase change is developed and applied to study
the melting behaviors in porous metal foam, in an effort to eluci-
date the role of pore distribution on melting heat transfer. The
effects of initial temperature difference, porosity and fractal
dimension on the melting heat transfer are all examined and inves-
tigated. Furthermore, the present model is experimentally verified
by the dynamic temperature variation of PCM and porous matrix
during the melting process obtained by the current study.
2. Fractal description of porous metal foam
The porous media no matter what is existed in nature (e.g.
stone and soil) or is made artificially (e.g. metal foams) all exhibit
a certain degree of disorder [27]. It has been proved that the struc-
ture of porous media shows multi-scale and statistically self-
similar fractal properties. As an available source of mathematical
tool, fractal geometry is capable of characterization of random,
multiscale and statistical self-similar properties. It has been docu-
mented that the fractal Brownian motion (FBM) is successfully
applied to characterize the pore distribution of porous media
[27,30]. Considering this, on the basis of the statistics of FBM, a
fractal description of the pore distribution is presented in this
paper so as to reconstruct the geometry structure of porous metal.
The one-dimensional fractal Brownian motion is defined as a
random process fH(t):[0, 1] 2 R, that satisfies: (i). fH(t) is a contin-
 Z. Deng et al. / International Journal of Heat and Mass Transfer 113 (2017) 1031–1042 1033

uous function and fH(0) = 0; (ii). For any t  0 and h > 0, the incre- mined by the scalar function fH(r). Fig. 1(a) illustrates the scheme
ment fH(t + h)  fH(t)  N(0, h2H) with the probability distribution of the elementary generation steps of a sing-cell fractal porous
Z   structure. As shown in Fig. 1(a), the number on the node of square
1 H f
u2
Prob ðf H ðt þ hÞ  f H ðtÞ 6 f Þ ¼ pffiffiffiffiffiffiffi h exp 2H
du ð1Þ lattice indicates the iteration number j. At the beginning, the
2p 1 2h square is considered to be homogeneous and the function specify-
where H is the Hurst exponent ranging from 0 to 1. For H = 0.5, one ing the pore size is regarded as a constant, e.g., fH(r) = 0. The gener-
recovers the regular Brownian motion, while the fractal Brownian ation steps for the pore distribution of fractal structure is listed as
motion is for H – 0.5. Similar to the regular Brownian motion, the follows:
increment fH(t + h)  fH(t) following FBM statistics is also a station-
ary value. As shown in Eq. (1), the increment’s probability distribu- (i) Assign random number from the Gaussian distribution N(0,
tion is independent of the variable t r20) to the four corner points of square denoted with number
1.
1 2H 2H (ii) Then, the center point of the square is linearly interpolated
Covðf H ðtÞ; f H ðt þ hÞ  f H ðtÞÞ ¼ ½ðt þ hÞ  t 2H  h  ð2Þ
2 with the four corner values. A random number df1 generat-
Eq. (2) is the covariance of fH(t) and its increment fH(t + h)  fH(t). It ing from N(0, r21) is add to the center point. In this way,
is clear that the correlation coefficient equals 0, only when H = 0.5. the pore size at the center of square as denoted by number
In another word, for a random process fH(t) following FBM statistics, 2, is calculated by
there is no independent increment. Consequently, the FBM is statis-
tical self-similar by the parameter H. In terms of the mean and vari- 1X 4
f H;j ðrÞ ¼ f ðrÞ þ df1 ð8Þ
ance of the increments, the statistical property of FBM may be 4 k¼1 k
expressed as
where fk(r) is the pore size located at four corners of the
E½f H ðt þ rhÞ  f H ðtÞ ¼ 0 ð3Þ square. It is clear from Eq. (8) that the increment variance
of fH,j(r) is 2(r20/4 + r21). And the space displacement
2 2 pffiffiffi
E½ðf H ðt þ rhÞ  f H ðtÞÞ  ¼ r 2H E½ðf H ðt þ hÞ  f H ðtÞÞ  ð4Þ r = 1/ 2 for the center point of the square. According to the
Defining the variances scaling rule, the quantity r21 satisfies the relation
 2H
r ¼ E½ðf H ðt þ rhÞ  f H ðtÞÞ2 
2
ð5Þ 1
rh r21 ¼ pffiffiffi r20 ð1  2H2 Þ ð9Þ
2
r2h ¼ E½ðf H ðt þ hÞ  f H ðtÞÞ2  ð6Þ (iii) The next step is to calculate the values of function fH,3(x) at
the midpoints of each side (denoted by number 3) by lin-
Allow Eq. (4) to be written as
early interpolating with two adjacent corner values and add-
r2rh ¼ r2H r2h ð7Þ ing a random number df2 from N(0, r22). Thus, the pore size at
2 2 the midpoint of the four edges can be written as
where rrh and r represent the variances of the increments of fH(t)
h
with lags rh and h, respectively. Eq. (7) expresses the scaling prop- pffiffiffi 1X 2

erties of FBM. It shows that the increments of FBM are statistically f H;j ð xÞ ¼ f ðrÞ þ df2 ð10Þ
2 k¼1 k
self-similar, which has a vital significance on the geometry recon-
struction of fractal porous media. where fk(r) is the pore size of the end-points of each edge.
As a recursive technique, the random midpoint displacement Now, the space displacement r = 1/2 for those midpoints of
algorithm is typically applied for the generation of fractal struc- each side. Contrasted with the space displacement of the cen-
tures [29,31]. In this paper, a 2-dimension fractal porous structure ter point, and applying the scaling rule of FBM, we get the
is characterized based on this algorithm. As stated before, here relationship between r21 and r22 as follows
defining a random process fH(r) following FBM statistics [Eqs. (3)  2H
and (7)], where r is the space displacement. In order to facilitate 1
r22 ¼ pffiffiffi r21 ð11Þ
the quantitative description of pore structure, the pore shape is 2
assumed as a square lattice, and the relevant pore size is deter-

Cell-1 Cell-2 Cell-3

1 3 1

4 4

3 2 3

4 4

1 3 1 Cell-Np+1

(a) elementary generation steps of single-cell fractal (b) multi-cell fractal porous structure
Fig. 1. Fractal description of porous media.
1034 Z. Deng et al. / International Journal of Heat and Mass Transfer 113 (2017) 1031–1042

(iv) After the first run of routine, nine sub-squares are generated
and characterized by the values of function fH,j(x) mentioned
above. The corresponding porous structure is shown in Fig. 1
(a). Then, for each sub-square, repeat steps ii and iii up to
pffiffiffi
level n with applying a scaling factor r = (1/ 2)2H to the vari-
ance of space displacement. Finally, a 2D matrix array fol-
lowing FBM statistic with (2n + 1)  (2n + 1) date points
will be generated. In order to obtain the porous media with
a given range of pore size, the 2D matrix array following
FBM statistics must be rescaled by the means of Kikkinides
et al.’s method [29].
(3) The density of PCM is a function of temperature;
(4) The properties of PCM are linear with temperature in the
mushy zone;
In the current study, the enthalpy-porosity method is used to
model the unsteady heat transfer in metal foam accompanied with
melting. The effect of natural convection in the liquid and mushy
zone is considered in the model, so the change of density in
continuity equation cannot be neglected, as well as the volume
force in momentum equation. Therefore, the governing equations
is written as

It is worth noting that the porous media generated by above @ q @ðquÞ @ðqv Þ
þ þ ¼0 ð12Þ
algorithm contains a limited number of pores. To overcome this @s @x @y
shortcoming, an improvement algorithm is adopted as illustrated
in Fig. 1(b). The new algorithm divides the original square lattice
   
into arrays of smaller sub-lattices, each of which is dealt with @ðquÞ @ðquÞ @ðquÞ @ @u @ @u @p
above FBM procedure for single-cell porous media with a fixed H. þu þv ¼ l þ l  þ Su
@s @x @y @x @x @y @y @x
Note that the grid points shared by neighbored sub-lattices receive
ð13Þ
contributions from adjacent cells, which should be averaged over-
all. With the aid of modified algorithm, one can generate fractal
   
porous media with any degree of correlation, which is more consis- @ðqv Þ @ðqv Þ @ðqv Þ @ @v @ @v @p
þu þv ¼ l þ l  þ Sv
tent with real porous media. @s @x @y @x @x @y @y @y
ð14Þ
3. Melting heat transfer in porous metal foam
   
@ðqhÞ @ðqhuÞ @ðqhv Þ @ @T @ @T
3.1. Model description þ þ ¼ k þ k ð15Þ
@s @x @y @x @x @y @y

In the current investigation, the porous metal foam is character-
ized by the fractal Brownian motion (FBM), as illustrated in Fig. 2
(a). In the plot, the black zone is the porous matrix of porous metal
foam; the discretely-distributed white squares represent the pores,
in which the phase change material (PCM) are embedded. To verify
the superior thermal performance of melting phase change, a
square cavity embedded with PCM also designed [see Fig. 2(b)].
The square cavity has the equal area with fractal porous metal
foam, and its inner space is filled with PCM completely. In the sim-
ulation, the porous matrix of metal foam is aluminum; the lauric
acid (thermal properties in Table 1) is selected as the PCM for ther-
mal storage; the dimensions of the computational domain is of
length L = 9 cm, width W = L = 9 cm.
3.1.1. Governing equations
The heat transfer accompanied with the melting of PCM in
metal foam is illustrated in Fig. 2. For simplicity, the assumptions
are made as follows:
(1) The properties of porous matrix are constant;
(2) There are three possible states in the whole melting process,
i.e., the solid zone, liquid zone and mushy zone;
where p is the liquid pressure, q is the density, l is the viscosity, h is
the enthalpy, T is temperature, k is the thermal conductivity, and s
is the time. u and v are the velocities in x and y directions, respec-
tively. In addition, Su and Sv represent the source term of momen-
tum equation [Eqs. (13) and (14)] along the x, y-directions
respectively, which can be calculated by
ð1  bÞ2
Su ¼ Amush u ð16Þ
b3 þ n
ð1  bÞ2 qref gdðh  href Þ
Sv ¼ 3
Amush v þ ð17Þ
b þn cp
where d is a thermal expansion coefficient, n is merely a computa-
tional constant introduced to avoid division by zero, qref and href are
reference values of density and enthalpy, Amush is a constant. It is
important to note that, for the melting phase change, the
enthalpy-porosity method adopts the liquid fraction b to describe
the mushy state, rather than to track the phase interface explicitly.
In this way, the mushy state is regarded as a porous region with
porosity set equal to the liquid fraction. The value of liquid fraction
ranging from 0 to 1, where 0 denotes the solid state and 1 denotes
liquid state. The mathematical expression of liquid fraction b is

L L
(a) adiabatic
(b) adiabatic
PCM
Solid

L
Tw L Tw PCM
adiabatic
adiabatic

y y
g g
o x o x
adiabatic adiabatic

Fig. 2. Schematic of melting heat transfer: (a) fractal porous metal foam, (b) square cavity.
 Z. Deng et al. / International Journal of Heat and Mass Transfer 113 (2017) 1031–1042 1035

Table 1
Thermal properties of PCM and porous matrix.

Material Density q (kg/m3) Specific heat capacity cp (kJ/kgK) Conductivity k (W/mK) Latent heat Lp (kJ/kg) Melting temperature Tp (°C)
Aluminum 2719 0.871 202.4
Lauric acid (C12H24O2) 869 2.15 (solid phase); 0.15 (solid phase); 178 43
2.203 (liquid phase) 0.22 (liquid phase)

8  
< 0 T < Ts
> @T p;s  @T p;l 
kp;s  ¼ kp;l ð27Þ
b ¼ TTT s
Ts 6 T 6 Tl ð18Þ @n C @n C
> l T s
:
1 T > Tl
where C is used to indicate the interface, n is the normal direction
where subscripts l and s represent the liquid state and solid state of
of heat flux.
PCM. In the simulation, the solidus temperature Ts equals 315 K
while liquidus temperature Tl equals 317 K.
For the PCM in the liquid and mushy zone, the enthalpy h in Eq.
(15) includes the total sensible heat hT and the total latent heat Dh, 3.2. Model validation
which is calculated by
Z The classical Stefan problem arises when studying physical
T
processes related to phase transformation of matter [32]. In order
h¼ cp dT þ Dh ð19Þ
T ref to validate the present model, we numerically simulate the clas-
sical Stefan problem and compared our numerical results with
where cp is the specific heat capacity of PCM, Tref is reference tem- the analytical solution [32]. The corresponding model and the
perature. Thus, after introducing the liquid fraction b into Eq. (19), boundary conditions are given in Fig. 3(a). At the initial moment,
the enthalpy in Eq. (15) can be written as a unified formula as the system temperature equals to the phase change temperature
follow Tsolidus of PCM, and the temperature of left hand boundary equals
Z T a constant value, which is higher than Tsolidus. Then, the melting
h¼ cp dT þ bLp ð20Þ process starts and the solid-liquid interface moves to the right.
T ref
Fig. 3(b) presents the numerical and analytical relationships
where Lp is the latent heat of PCM. In addition, the specific heat between Fourier number and liquid fraction. The numerical solu-
capacity cp and thermal conductivity kp can be calculated by tion agrees well with the analytical one, which proves that the
present model is capable to describe the physical processes
cp ¼ cp;s þ bðcp;l  cp;s Þ ð21Þ
involving phase change.
To verify the present model, the unsteady heat transfer with
kp ¼ kp;s þ bðkp;l  kp;s Þ ð22Þ
melting process in the porous metal object filled with PCM is also
where subscripts p represents PCM. experimentally measured by the present study. An aluminum
block is machined to form a porous metal object [see Fig. 4(a)]
3.1.2. Initial and boundary conditions with the pore dimensions shown in Fig. 4(b). In the experiment,
At the initial time, the whole zone of porous metal foam is in a all pores of porous metal object are filled with phase change
state of thermal equilibrium, so the initial condition is assumed as material of Lauric acid, and the external surfaces of four sides
for the porous metal object are covered with insulation material.
Tðx; y; 0Þ ¼ T 0 ð23Þ
Only one side of porous metal object directly contacts with
In the simulation, the initial temperature is set to be T0 = 300 K. constant temperature heat source that is regulated by the
As shown in Fig. 2, the boundary conditions of metal foam are thermostatic water bath (temperature range of 35 °C  100 °C,
viewed as the adiabatic boundary except the left boundary, where with ±0.1 °C), and another three sides of porous metal foam
the constant temperature boundary is imposed. The boundary con- structure are insulated. All the temperatures are measured by
ditions are set as follows thermocouples (OMEGA T type with ±0.5 °C) and recorded by
   the data-collecting instrument (Agilent 34972A). The temperature
@T  @T  @T 
Tjx¼0 ¼ T w ; ¼ 0; ¼ 0; ¼0 ð24Þ measurement points for the porous matrix and PCM is configured
@x x¼L @y y¼0 @y y¼L
in the porous metal object.
where Tw is the temperature of the heating wall, and the initial tem- Fig. 5 compares the temperature dynamic variations for the
perature difference DT is defined as solid matrix and PCM of porous metal object during the melting
process between the experiment and simulation. The initial tem-
DT ¼ T w  T 0 ð25Þ perature for the whole region of porous metal object is
Without specified, the initial temperature difference is assumed T0 = 22.5 °C. When the experiment begin, the temperature of the
to be DT = 30 K in the following simulation. It is important to note side wall adjacent to the flow channels rises rapidly from the initial
that the boundary and initial conditions of fractal porous metal temperature to the setting temperature Tw = 67 °C. After the star-
foam are the same as the square cavity in the simulation of melting tup stage (about 1.5 min), the temperature of side wall Tw holds
heat transfer. steady at Tw = 67 °C with a temperature fluctuation less than
Melting phase change in metal foam is a conjugate heat transfer 0.7 °C. As shown, the simulated temperature dynamic variations
problem that combines heat conduction of porous matrix and of solid matrix and PCM in porous metal object with pore structure
phase change of PCM. The continuities of temperature and heat shown in Fig. 4(b) agree with the experiment result. Based on this,
flux are assumed at the surface of porous matrix, the present experiment can indirectly verify that the present
model is reasonable and capable of simulating the melting behav-
T s;C ¼ T l;C ð26Þ
iors in porous metal foam.
1036 Z. Deng et al. / International Journal of Heat and Mass Transfer 113 (2017) 1031–1042

(b) 1.0 Numerical Solution

Analytical Solution
0.8

0.6

β
0.4

0.2

0.0
0.0 0.6 1.2 1.8 2.4 3.0
Fo
Fig. 3. Validation of the classical Stefan problem: (a) description of Stefan problem; (b) the numerical and analytical relationships between Fourier number and liquid
fraction.

(a)

30mm Constant Temperature

(b) (c) T
Water Bath
4.5mm
Agilent Data
7.8mm 7.1mm Logger

4.4mm 5.8mm
Solid Matrix
10mm
30mm

3.5mm 6.7mm

7.9mm
Polyamide
PCM
Quartz Glass

Fig. 4. Experimental system for the melting heat transfer in a porous metal structure: (a) porous metal structure, (b) pore distribution and dimensions, (c) schematic of
experimental system.

4. Results and discussion
This paper aims to understand the melting behaviors of PCM in
porous metal foam, especially to explore the role of pore distribu-
tion (fractal dimension and porosity) on the melting process of
PCM. In the current study, the pore structure of porous metal foam
is characterized by FBM. The following results are obtained based
on the numerical model of melting process of PCM in metal foam
via FBM. Since the porous metal foam [see Fig. 1(b)] has complex
geometry structure, an unstructured mesh is applied to arrive at
a numerical solution. The mesh is mainly composed of quadrilat-
eral grid elements, and a few triangular grid elements are also
applied. In addition, a mesh refinement strategy is considered for
the small pores and narrow porous matrix so as to improve the
quality of mesh. The finite-volume method is utilized to numeri-
cally solve the governing equations along with the boundary
conditions.
In the simulation, a grid independence test with several mesh
sizes is conducted to ensure that the numerical analysis is valid.
Fig. 6(a) gives the average temperature of whole porous metal
foam during the melting process with different mesh sizes. For
the present simulations, the time-step is set as Dt = 0.5 s, and the
grid number varies from 89,345 to 442,156. As shown in the
figure, with the increase of grid number, the difference of average
 Z. Deng et al. / International Journal of Heat and Mass Transfer 113 (2017) 1031–1042 1037

(a) 70 Solid Matrix 70 PCM

60 T1, Exp 60
T4, Exp
T2, Exp
50 50 T5, Exp

T/ć

T/ć
T3, Exp
T6, Exp
40 T1, Num 40
T4, Num
T2, Num
30 30 T5, Num
T3, Num
20 20 T6, Num
0.0 1.5 3.0 4.5 6.0 7.5 0.0 1.5 3.0 4.5 6.0 7.5
τ / min τ / min

Fig. 5. Comparison of temperature response between numerical and experimental results: (a) porous matrix, (b) PCM, (c) temperature measurement point.

(a) 48 (b) 48

44 44
Tav / ć
T av / ć

40 40
Ng = 89345 Δ t=0.25s
Ng = 192232 Δ t=0.5s
36 Ng = 246756 36
Δ t=1s
Ng = 329316 Δt=2s
32 Ng = 442156 32

0 200 400 600 800 1000 0 200 400 600 800 1000
τ/s τ/s

Fig. 6. Average temperature of whole porous metal foam during the melting process (e = 0.75, DT = 30 °C): (a) Dt = 0.5 s; (b) Ng = 329,316.

temperature for the whole porous metal foam during the melting
process can be considered negligible when the grid number
Ng > 329,316. Considering the time cost, the grid number is chosen
as Ng = 329,316 in the following simulation.
The melting phase change in porous metal foam is unsteady
heat transfer process. Considering this, a time-step independence
test is also conducted to ensure the validity of the present simula-
tion. Fig. 6(b) shows the average temperature of whole porous
metal foam during the melting process with different time-step.
In this simulation, the grid number is set as Ng = 329,316, and
the time step is assumed to be Dt = 2  2i s (i = 0, 1, 2, 3). It is seen
that, with the decrease of time-step, the difference of average tem-
perature of whole porous metal foam in the melting process
between adjacent time-step tends to be smaller, especially when
time-step Dt < 0.5 s. Therefore, the time-step is set as Dt = 0.5 s
for all the following simulations.
4.1. Melting behaviors
4.1.1. Dynamic response of temperature
In metal foam, the presence of porous matrix has an important
effect on the melting heat transfer. Since the thermal conductivity
of porous matrix is far larger than that of PCM, the melting heat
transfer inside the metal foam is enhanced when compared with
the PCM alone in square cavity. In order to give a visualization of
enhancement heat transfer, Fig. 7 compares the temperature pro-
files of the whole region during the melting process between the
porous metal foam and square cavity. As shown, the heat transfer
rate in metal foam is obviously faster than that in square cavity,
and the temperature distribution in porous metal foam is more
uniform than that in square cavity. That is attributed to the fast
heat flow channel arising from the porous matrix, which is benefi-
cial for the PCM melting in the horizontal and vertical. In the case
of initial temperature difference DT = 30 °C, it takes less than 250 s
for the heat flow through the porous metal foam, however, after
2500 s more than 50% distance is still left for the heat flow through
the square cavity filled with PCM.
To get a deeper insight into the melting heat transfer, the tem-
perature dynamic response of PCM in fractal porous metal foam
and square cavity are presented in Fig. 8. Note that the point T1,
T2 and T3 are at the same locations for the metal foam and square
cavity. As expected, the PCM temperature in porous metal foam is
higher than corresponding square cavity at the same time for the
identical position, since the presence of porous matrix improves
the thermal performance of the whole region. In addition, the tem-
perature of PCM is almost unvaried during the melting phase
1038 Z. Deng et al. / International Journal of Heat and Mass Transfer 113 (2017) 1031–1042
τ = 5s τ = 250s
Fractal
Cavity
Fig. 7. Temperature distribution in fractal porous metal foam and square cavity during the melting process (DT = 30 °C).
55
45
T/ć
Cavity, Fractal
T1: ,
35
T2: ,
T3: ,
25
0 10 20 30
τ / min
Fig. 8. Dynamic response of temperature for PCM in metal foam and square cavity during the melting.
change, and the duration time of melting phase change for PCM
varies with position. The farther the position away from the heat-
ing wall, the longer the duration time is. It is interesting that the
temperatures of PCM in square cavity fluctuates within time, when
their temperatures exceed the melting point. This is attributed to
the occurrence of natural convection, which leads to heat redistri-
bution inside square cavity. It is also indicated that, the tempera-
ture difference of PCM in porous metal foam is far less than that
of square cavity.
4.1.2. Evolution of melting front
The evolution of melting front is an important indicator for the
performance of melting heat transfer. Fig. 9 displays the solid-
liquid distribution of PCM in porous metal foam and square cavity
during the melting process. As expected, the melting front evolu-
tion of PCM in porous metal foam is undoubtedly faster than the
corresponding square cavity. The explanation of this phenomenon
is that: (1) Owing to high thermal conductivity, a fast heat flow
channel is formed arising from the porous matrix for the melting
of PCM in both the horizontal and vertical direction; (2) the inter-
stitial heat transfer is of significance in the melting phase change of
the dispersedly distributed PCM in pores.
It is also shown in the figure that the melting front in square
cavity is initially an approximately straight line, since the main
heat transfer mode during the early period is thermal conduction.
As time goes on, the liquid fraction of PCM increases, the melting
τ = 1250s τ = 2500s
40 50
front varies from straight to curve due to the effect of natural con-
vection of liquid PCM. Unlike the PCM alone in square cavity, the
melting front no longer continuous during the melting process.
The evolution of melting front is more complex inside the fractal
metal foam. Due to the effect of close-celled structure, there are
many independent local solid-liquid phase interfaces inside pores
for fractal metal foam.
Fig. 10 presents the time-dependence of the liquid fraction of
PCM, b, in both the fractal porous metal foam and square cavity
during the melting process. It is seen that, the increase rate of liq-
uid fraction of PCM in porous metal foam during the melting pro-
cess is larger than that of square cavity. This means that the
melting heat transfer rate in porous foam is faster than the corre-
sponding square cavity, and the porous matrix improves the ther-
mal performance of PCM. In addition, the momentum toward
increasing support for the liquid fraction of phase change material
inside the fractal porous metal foam tends to flatten out because of
the reduction in the temperature difference between porous
matrix and PCM as the time proceeds.
4.2. Influence factor analysis
The melting phase change is a complex unsteady heat transfer
process in porous metal foam. To analyze the role of pore structure
on melting behaviors in fractal porous metal foam, the evaluation
parameters, including the melting time and the dimensionless
 Z. Deng et al. / International Journal of Heat and Mass Transfer 113 (2017) 1031–1042 1039

τ = 5s τ = 250s τ = 1250s τ = 2500s

Fractal

Cavity

Fig. 9. Evolution of solid and liquid phase of PCM during the melting process (DT = 30 °C).

1.0 Fractal (ε = 0.75, D =1.8)

PCM in fractal metal foam with different initial temperature differ-
Cavity ence. As expected, the melting time decreases monotonously with
0.8 the initial temperature difference. As see from the figure, the melt-
ing time decrease sharply when DT < 50 °C, however, the decrease
0.6 trend is smooth once DT > 50 °C. This implies that the initial tem-
perature difference of 50 °C is a good choice for the melting phase
β

0.4 change energy storage.

Fig. 11(b) plots the effect of the initial temperature difference
0.2 on the time-dependence of dimensionless domain temperature.
As seen from the figure, the initial temperature difference plays
0.0 an important role in the evolution of thermal equilibrium between
the domain of porous metal foam and the heating wall. The time
0 10 20 30 40 50 required to reach the final thermal equilibrium state is shorter
τ / min for the larger initial temperature difference. The final thermal equi-
librium state arrives at 32 min for the case of DT = 70 °C, while the
Fig. 10. Time-dependence of liquid fraction of PCM in fractal porous metal foam PCM in porous metal foam with the initial temperature difference
and square cavity.
DT = 30 °C does not melt completely. In addition, as time goes on,
the dimensionless domain temperature of porous media during
domain temperature are introduced to quantify the melting heat melting process sequentially experiences the draft increase stage
transfer. In this study, the melting time, sc, is defined as the dura- (marked as stage I), the gentle stage (marked as stage II), the small
tion time for the PCM are completely melted when the heat flow is increase stage (marked as stage III) and finally achieves a constant
imposed to the porous metal foam. The dimensionless domain value. The heat transfer mechanism in stage I and III is heat con-
temperature, hav, is defined as the dimensionless temperature duction dominated while the melting phase change is dominated
averaged over the domain of porous metal foam in stage II.

T av  T 0
hav ¼
Tw  T0
where Tav is the average temperature of the whole region. hav
increases as the increase of average temperature for the domain.
Once the whole domain of porous metal foam is of the same tem-
perature of the heating wall (i.e. the final equilibrium state), the
maximum value of dimensionless domain temperature is reached,
i.e. hav = 1.
4.2.1. Effect of initial temperature difference
The initial temperature difference directly determines the heat
input into the porous metal foam. The large initial temperature dif-
ference is capable of accelerating the heat transfer between the
PCM and the porous matrix. The melting time is shorter for a larger
initial temperature difference. Fig. 11(a) shows the melting time of
ð28Þ 4.2.2. Effect of porosity
Porosity is one of the most important parameters in determin-
ing the structure of the porous metal foam. Since the pores of metal
foam are filled with PCM, the larger porosity means a higher share
of PCM in porous metal foam. As a result, the thermal conductivity
of porous metal foam with higher porosity is lower than that with
lower porosity. The variation of thermal conductivity undoubtedly
affect the melting behaviors of PCM in metal foam. Fig. 12(a)
describes the melting time of PCM in metal foam with different
porosity for a given fractal dimension. As seen from the figure,
the melting time increases monotonously with the porosity. The
increase of porosity to the same extent weakens the heat transfer
performance of porous metal foam. In addition, there is a rapid
increase of melting time for porosity e < 0.6, while the increase
trend become slow for larger porosity.
1040 Z. Deng et al. / International Journal of Heat and Mass Transfer 113 (2017) 1031–1042

(a) 50 (b)
1.0

40 0.8

τc / min
0.6

θav
30
0.4
ΔT = 30ć
20 0.2 ΔT = 50ć
ΔT = 70ć
0.0
10
30 40 50 60 70 0 8 16 24 32
ΔT /ć τ / min

Fig. 11. The effect of initial temperature difference on melting behavior (e = 0.75, D = 1.8): (a) melting time, (b) dimensionless domain temperature.

(a) 50 (b) 1.0

40 0.8
τc / min

0.6
30

θav 0.4 ε = 0.6

20
ε = 0.7
0.2
ε = 0.75
10
0.0
0.4 0.5 0.6 0.7 0.8 0 6 12 18 24 30
ε τ / min

Fig. 12. Effect of porosity on the melting behaviors (D = 1.8): (a) melting time, (b) dimensionless domain temperature.

(a) (b) 1.0

18 D = 1.9
D = 1.5
0.8
17 D = 1.1
0.6
τc / min

θav

16
0.4
15
0.2
14
0.0
1.0 1.2 1.4 1.6 1.8 2.0 0 5 10 15 20 25
D
τ / min

Fig. 13. Effect of fractal dimension on melting behaviors (DT = 50 °C, e = 0.7): (a) melting time, (b) dimensionless domain temperature.

Fig. 12(b) illustrates the evolution of dimensionless domain
temperature for the fractal porous metal foam. It is seen that there
is a larger increase of dimensionless domain temperature for the
porous metal foam with a lower porosity. This phenomenon indi-
cates that the decrease of porosity accelerates the arrival of the
final thermal equilibrium. The phase change heat transfer is
improved with the decrease of porosity, but the heat storage
capacity is weakened in porous metal foam. Therefore, to balance
the melting heat transfer enhancement with the heat storage
capacity, a reasonable selection of porosity is very important for
the engineering application of fractal porous media.
4.2.3. Effect of fractal dimension
Only the porosity is not enough to characterize the detailed nat-
ure of pore structure for a given porous metal foam. For a given
porosity, the layout of the porous matrix (i.e. the feature of pore
distribution) can be different. The degree of disorder of the pore
distribution, including the dimension and spatial distribution of
pores, can be quantified by the fractal dimension. In other words,
the fractal dimension determines the heterogeneity of the result-
ing pore structure of the fractal porous metal foam. The larger
the fractal dimension is, the larger degree of irregularity the pore
distribution is.
As stated above, the presence of porous matrix contributes to the
enhancement of melting heat transfer of PCM. For a given porosity,
the fractal dimension (i.e. feature of the pore distribution) is
expected to affect the heat transfer behaviors in metal foam accom-
panied with melting phase change. Fig. 13(a) presents the effect of
fractal dimension on melting time of PCM in fractal metal foam. As
shown, the melting time increases with fractal dimension in a non-
linear fashion, and the increase of melting time is slow for D < 1.5
while there is an evident increase trend for D > 1.5. This implies that
increase in disorder degree of the pore distribution leads to the
decline of overall heat conduction performance of porous matrix
 Z. Deng et al. / International Journal of Heat and Mass Transfer 113 (2017) 1031–1042
and hence weakens melting heat transfer. Therefore, the porous
metal foam with smaller fractal dimension is beneficial for the
melting heat transfer. This is also to say, the energy storage system
that stores latent heat by the melting of PCM can be improved by
using porous metal foam with smaller fractal dimension.
To further identify the role of fractal dimension on melting
heat transfer performance of porous metal foam, Fig. 13(b) plots
the evolution of the dimensionless domain temperature in por-
ous metal foam with different fractal dimension during the melt-
ing process. It can be seen from the figure that, there is a smaller
dimensionless domain temperature of porous metal foam for a
larger fractal dimension during the melting heat transfer. The
increase in fractal dimension leads to a smaller heat input into
the porous metal foam with PCM, which contributes to a slower
melting rate in porous metal foam. In this case, the arrival of
thermal equilibrium between the PCM and the porous matrix
is delayed.
5. Conclusion
An unsteady model of solid-liquid phase change is developed
and applied to investigate the melting behaviors in porous metal
foam with particular focus on the role of pore distribution. In the
model, the pore distribution of porous metal foam is described
by the fractal Brownian motion, which is capable of characteriza-
tion of random, multiscale and statistical self-similar properties.
The dynamic response of temperature and the evolution of melt-
ing front in fractal porous metal are presented and compared to
the corresponding square cavity. The effects of initial tempera-
ture difference, porosity and fractal dimension on the melting
heat transfer are all examined and investigated. Furthermore,
the present model is verified by the experiment of melting phase
change in porous metal. The major conclusions are listed as
follows:
(1) The porous metal foam has a significant impact on the
enhancement of melting heat transfer. When compared with
PCM alone, the melting process in porous metal foam
possesses the larger melting heat transfer rate, the faster
evolution of melting front and the higher liquid fraction of
PCM.
(2) The interstitial heat transfer between the PCM and the
porous matrix is of significance for the melting behaviors
in porous metal foam. Unlike the PCM alone in square cavity,
the melting front is no longer continuous during the melting
process. In porous metal foam, there are many independent
local solid-liquid phase interfaces inside pores.
(3) The initial temperature difference and the porosity are two
important parameters affecting the melting heat transfer.
In porous metal foam, an increase in initial temperature dif-
ference or a decrease in porosity lead to the acceleration of
melting phase change.
(4) Even though the porosity is identical for the metal foam, the
melting phase change is also affected by the fractal dimen-
sion. The increase in fractal dimension leads to a larger
degree of disorder of the pore distribution. In porous metal
foam with larger fractal dimension, the melting time is
longer and the time-dependence of dimensionless domain
temperature is smaller for the melting process.
Conflict of interest
The authors declared that there is no conflict of interest.
1041
Acknowledgements
The authors gratefully acknowledge the support provided by
National Natural Science Foundation of China (Nos. 11190015
and 51406175) and the Natural Science Foundation of Jiangsu
Province (BK20140488).
References
[1] F.L. Tan, S.F. Hosseinizadeh, J.M. Khodadadi, L.W. Fan, Experimental and
computational study of constrained melting of phase change materials (PCM)
inside a spherical capsule, Int. J. Heat Mass Transfer 52 (15–16) (2009) 3464–
3472.
[2] M. Martinelli, F. Bentivoglio, A. Caron-Soupart, R. Couturier, J.F. Fourmigue, P.
Marty, Experimental study of a phase change thermal energy storage with
copper foam, Appl. Therm. Eng. 101 (2016) 247–261.
[3] S. Krishnan, J.Y. Murthy, S.V. Garimella, Analysis of solid-liquid phase change
under pulsed heating, J. Heat Transfer 129 (2007) 395–400.
[4] T.J. Lu, Thermal management of high power electronics with phase change
cooling, Int. J. Heat Mass Transfer 43 (13) (2000) 2245–2256.
[5] A. Koca, H.F. Oztop, T. Koyun, Y. Varol, Energy and exergy analysis of a latent
heat storage system with phase change material for a solar collector, Renew.
Energy 33 (2008) 567–574.
[6] S.C. Lin, H.H. Al-Kayiem, Evaluation of copper nanoparticles – paraffin wax
compositions for solar thermal energy storage, Sol. Energy 132 (2016) 267–
278.
[7] R. Baetens, B.P. Jelle, A. Gustavsen, Phase change materials for building
applications: a state-of-the-art review, Energy Build. 42 (9) (2010) 1361–1368.
[8] A.A.R. Darzi, M. Jourabian, M. Farhadi, Melting and solidification of PCM
enhanced by radial conductive fins and nanoparticles in cylindrical annulus,
Energy Convers. Manage. 118 (2016) 253–263.
[9] A. Sharma, V.V. Tyagi, C.R. Chen, D. Buddhi, Review on thermal energy storage
with phase change materials and applications, Renew. Sustain. Energy Rev. 13
(2009) 318–345.
[10] J. Chen, D. Yang, J. Jiang, A. Ma, D. Song, Research progress of phase change
materials (PCMs) embedded with metal foam (a review), Procedia Mater. Sci. 4
(2014) 389–394.
[11] Z. Yang, S.V. Garimella, Melting of phase change materials with volume change
in metal foams, J. Heat Transfer 132 (2010) 062301.
[12] Z.Q. Chen, D.Y. Gao, J. Shi, Experimental and numerical study on melting of
phase change materials in metal foams at pore scale, Int. J. Heat Mass Transfer
72 (2014) 646–655.
[13] S. Mancin, A. Diani, L. Doretti, K. Hooman, L. Rossetto, Experimental analysis of
phase change phenomenon of paraffin waxes embedded in copper foams, Int.
J. Therm. Sci. 90 (2015) 79–89.
[14] Y. Zhao, C.Y. Zhao, Z.G. Xu, H.J. Xu, Modeling metal foam enhanced phase
change heat transfer in thermal energy storage by using phase field method,
Int. J. Heat Mass Transfer 99 (2016) 170–181.
[15] Y. Tian, C.Y. Zhao, A numerical investigation of heat transfer in phase change
materials (PCMs) embedded in porous metals, Energy 36 (2011) 5539–5546.
[16] Z.Y. Liu, Y.P. Yao, H.Y. Wu, Numerical modeling for solid-liquid phase change
phenomena in porous media: shell-and-tube type latent heat thermal energy
storage, Appl. Energy 112 (2013) 1222–1232.
[17] C. Zhang, L. Wu, Y. Chen, Study on solidification of phase change material in
fractal porous metal foam, Fractals 23 (01) (2015) 1540003.
[18] B. Buonomo, D. Ercole, O. Manca, S. Nardini, Thermal behaviors of latent
thermal energy storage system with PCM and aluminum foam, Int. J. Heat
Tech. 34 (2016) S359–S364.
[19] L. Colla, L. Fedele, S. Mancin, B. Buonomo, D. Ercole, O. Manca, Nano-PCMs for
passive electronic cooling applications, J. Phys: Conf. Ser. 655 (1) (2015)
012030.
[20] W.G. Alshaer, S.A. Nada, M.A. Rady, E.P. Del Barrio, A. Sommier, Thermal
management of electronic devices using carbon foam and PCM/nano-
composite, Int. J. Therm. Sci. 89 (2015) 79–86.
[21] W.G. Alshaer, S.A. Nada, M.A. Rady, C. Le Bot, E.P. Del Barrio, Numerical
investigations of using carbon foam/PCM/Nano carbon tubes composites in
thermal management of electronic equipment, Energy Convers. Manage. 89
(2015) 873–884.
[22] W.G. Alshaer, M.A. Rady, S.A. Nada, E. Palomo Del Barrio, A. Sommier, An
experimental investigation of using carbon foam–PCM–MWCNTs composite
materials for thermal management of electronic devices under pulsed power
modes, Heat Mass Transfer 53 (2) (2017) 569–579.
[23] S.A. Nada, W.G. Alshaer, Comprehensive parametric study of using carbon
foam structures saturated with PCMs in thermal management of electronic
systems, Energ. Convers. Manage. 105 (2015) 93–102.
[24] M.M. Sedeh, J.M. Khodadadi, Interface behavior and void formation during
infiltration of liquids into porous structures, Int. J. Multiphase Flow 57 (2013)
49–65.
[25] M.M. Sedeh, J.M. Khodadadi, Solidification of phase change materials
infiltrated in porous media in presence of voids, J. Heat Trans-Trans. ASME
136 (11) (2014).
1042 Z. Deng et al. / International Journal of Heat and Mass Transfer 113 (2017) 1031–1042
[26] E. Fleming, S.Y. Wen, L. Shi, A.K. da Silva, Experimental and theoretical analysis
of an aluminum foam enhanced phase change thermal storage unit, Int. J. Heat
Mass Transfer 82 (2015) 273–281.
[27] B.B. Mandelbrot, The Fractal Geometry of Nature, Freeman, New York, 1983.
[28] C. Turk, A. Carbone, B.M. Chiaia, Fractal heterogeneous media, Phys. Rev. E 81
(2010) 026706.
[29] E.S. Kikkinides, V.N. Burganos, Structural and flow properties of binary media
generated by fractional Brownian motion models, Phys. Rev. E 59 (1999)
7185–7194.
[30] B.B. Mandelbrot, J.W. Vanness, Fractional Brownian motions fractional noises
and applications, SIAM Rev. 10 (1968) 422–437.
[31] J. Jilesen, J. Kuo, F.S. Lien, Three-dimensional midpoint displacement algorithm
for the generation of fractal porous media, Comput. Geosci. 46 (2012) 164–
173.
[32] A.M. Meirmanov, The Stefan Problem, DeGruyter, Berlin, 1992.