Documentos de Académico
Documentos de Profesional
Documentos de Cultura
L
DT
; T
m
< T < T
m
DT
C
pl
; T _ T
m
DT
(3)
where C
ps
and C
pl
are the specic heats of the PCM in solid and
liquid state, respectively, T
m
is the temperature when melting
starts, and DT is the temperature range over which PCM melts.
The convective heat transfer within molten PCM is represented
using Eq. (4),
rC
p
vT
vt
rC
p
u$VT = V$(KVT) (4)
where u is the velocity eld of the molten PCM. The ow of molten
PCM due to density and gravity can be represented using Eq. (5),
r
vu
vt
r(u$V)u = V$
h
pI m(Vu) m(Vu)
T
i
grb
T T
ref
(5)
where I is the identity matrix, g is the standard gravitational ac-
celeration, r and b are density and coefcient of thermal expansion
of the PCM, respectively, and T
ref
is the reference temperature. The
non-linear, transient heat transfer Eq. (1)e(5) need to be solved
using a numerical procedure to determine the temperature prole
in the PCM inltrated metal foam.
2.2. The nite element model
The metal foam in this study was considered to have a face-
centered-cubic (FCC) structure. The morphology of metal foams is
dependent on the material and fabrication technique [25]. The use
of polymer sphere template is one of the techniques to fabricate
small pore size metal foams [26], which results in metal foams with
pores arranged in the FCC arrangement. A geometric model of the
metal foam was created by deducting 1/8 of a sphere at the eight
corners and 1/2 sphere from the six faces of a solid cube. A unit cell
of the metal foam in FCC conguration is shown in Fig. 1(a). The
unit cell structure of the phase change material was obtained by
taking an inverse of the metal foam structure, as shown in Fig. 1(b).
The two unit cells can be overlaid to form a single unit cell con-
taining both metal and PCM, which can then be replicated to forma
PCM-metal foam model of various dimensions.
Finite element simulation of transport properties of metal foams
has been previously conducted with a body-centered-cubic (BCC)
packing structure [27]. At a sphere radius of 0.5 of the unit cell size,
the BCC structure ceases to be completely open-celled, and the
corresponding porosity at this condition is 0.94. BCC models with
porosity lower than 94% over predicts experimental results,
because of the larger volume fraction of solid in the closed-cell
model. The FCC packing structure used in this study was gener-
ated with a constant pore size, such that models with porosities
above 75% remained an open-celled structure. The same model
structure has been used to predict the thermal conductivity of
micro- and nano-cellular polymer foams with satisfactory results
[28].
S.S. Sundarram, W. Li / Applied Thermal Engineering 64 (2014) 147e154 148
A FCC model with two overlaid unit cells replicated in one di-
rection (z-axis) is shown in Fig. 2. Metal slabs were added to the
two ends to represent the housing in which the PCM system was
enclosed. A thickness equal to one-tenth the unit cell size was
arbitrarily chosen for the metal housing. Another metal slab with
half the thickness was attached to one end of the PCM unit to
represent the heat source. Assuming the PCM system is at with a
thickness signicantly smaller than the lateral size, the congura-
tion shown in Fig. 2 was used to model a one-dimensional heat
transfer process through the thickness of the system. Because of
symmetry, the adiabatic condition was imposed on the four sides of
the model. A convective boundary condition was applied to the top
surface of the PCM system to represent a cooling situation by
natural or force air convection.
A no-slip boundary condition was enforced at the PCMand foam
ligament interfaces. A contact pair was dened to allow heat ex-
change between the two phases. A continuity constraint was
enforced so that the heat ux and temperature across the interface
was continuous,
n
PCM
$(KVT)
PCM
= n
metal
$(KVT)
metal
T
PCM
= T
metal
(6)
where n
PCM
and n
metal
are the normal vectors of the PCMand metal
surfaces, respectively.
The materials chosen in this study for metal foam and phase
change material were aluminum and parafn wax. Their properties
are listed in Table 1. The modied specic heat dened in Eq. (3)
was implemented using a Heaviside step function to account for
the latent heat of parafn. A temperature dependent viscosity was
used to account for the phase change from solid to liquid. Both of
these two properties are plotted in Fig. 3 as functions of tempera-
ture. A nite element model of the PCM-metal foam system was
developed using the commercial nite element analysis package
COMSOL. An extremely ne mesh with a minimum element size of
0.57 mm was generated for all the simulation models. This is the
Fig. 1. The solid model of a unit cell of (a) metal foamwith face centered cubic pore conguration and (b) the phase change material obtained by taking an inverse of the metal foam
structure.
Fig. 2. Solid model of phase change material inltrated metal foam system along with imposed boundary conditions: (a) a schematic representation of a PCM system used for
microelectronics cooling; (b) an element taken through the thickness direction for modeling.
Table 1
Properties of aluminum and parafn wax.
Aluminum Parafn wax
Density (Kg/m
3
) 2800 780
Specic heat (J/Kg-K) 910 2500 (solid and liquid)
Thermal conductivity (W/m-K) 237 0.20
Latent heat (KJ/Kg) e 180
Melting range (K) e 321e335
Viscosity (Pa S) e 0.024 (liquid)
S.S. Sundarram, W. Li / Applied Thermal Engineering 64 (2014) 147e154 149
nest mesh size possible in COMSOL. Models with high porosity
(>90%) could only be meshed using this mesh size because of the
presence of narrow struts. This mesh size was used for all the cases
to maintain the consistency of simulation results. A nonlinear time
dependent solver with backward differentiation and automatic
time stepping was employed for computation.
The developed model was validated with experimental data and
used to study the effects of pore size and porosity under various
heat generation and cooling conditions. The parameters used in the
study are shown in Table 2. The heat generation rates were selected
to represent typical low and high power generation situations in
portable electronics. The convective heat transfer rates were
selected to represent free and forced convection by air on metal
surfaces. The temperature at the center of the heat source, repre-
senting the average chip temperature, was used as a response
variable for analysis. The contribution of convective heat transfer
inside the molten PCM was also examined.
3. Results and discussion
3.1. Model validation
The developed model was validated with experimental data
found in literature. A copper foam with 2.5 mm pore size and 95%
porosity was used for comparison with data adapted from Zhao
et al. [5]. The thickness of the model was 25 mm. The power gen-
eration density was 3.2e5 W/m
3
. A copper foam with 635 mm pore
size and 90% porosity was used for comparison with data adapted
fromLi et al. [18]. The thickness of the model was 22.5 mm, and the
power density for heat generation was 5e7 W/m
3
. Both models
were insulated at the top surface. The comparisons are shown in
Fig. 4. As can be seen, the predicted temperature responses match
the experimental measurements closely.
Before the model prediction was compared to experimental
data, the effect of the number of unit cells in the x, y and z di-
rection was examined. It was found that varying the numbers of
unit cells in the x and y directions did not change the simulation
results. Hence, only one unit cell was used in the x and y directions
for all the cases in this study. The number of unit cells in the z
direction was determined based on the thickness of the PCM
system.
The thermal response of the PCM system with and without the
metal foam was also examined. A model with 2 mm pore size, 94%
porosity, and one unit cell on the z-axis was used for the case with
metal foam. A power generation density of 1.25e6 W/m
3
and an
adiabatic top surface condition were applied to both cases. The
temperature proles across the mid plane along z-axis for the two
cases during PCM melting are compared in Fig. 5. It is seen that the
temperature distribution in PCM is more uniform along the thick-
ness direction when metal foam is used. For pure PCM, melting is
close to the heat source, leading to a sharp increase in local
temperature.
3.2. The effect of convective heat transfer inside molten PCM
The above PCM-metal foam model was used to study the effect
of convective heat transfer in molten PCM. The transient temper-
ature response at the center of the power source with and without
the inclusion of the convection mechanism in the molten PCM is
shown in Fig. 6. It is seen that the temperature difference between
the two cases is insignicant, implying that convection could be
neglected in the model. The velocity of PCM movement during
melting across the mid plane has been found to be on the order of
10
8
m/s. This lowvelocity results in a Grashof number much lower
than the cutoff value (10
3
) for convection to be signicant. The high
viscosity of molten PCM (0.0269 Pa s) combined with its low co-
efcient of thermal expansion (3.085 10
4
K
1
) is the reason for
such a low velocity. As the pore size reduces further, the uid ve-
locity inside the porous structure will be even lower because of
increased tortuosity. Therefore, it is reasonable to assume that the
effect of convective heat transfer will be insignicantly small in
PCM inltrated microcellular metal foams; and only conduction
though the metal foam and PCM needs to be considered in the
models.
Fig. 3. Specic heat of the phase change material represented as a Heaviside step and
temperature dependent viscosity function to account for the phase change process.
Table 2
Parameters for studying pore size and porosity effect.
Parameter Value
Pore size (mm) 5, 10, 25, 50, 100
Porosity 75, 80, 84, 88, 94
Heat generation (W/m
3
) 2.5e8, 3.2e5, 5e7, 12.5e8
Convective heat transfer coefcient (W/m
2
-K) 1, 10
Fig. 4. Predicted temperature compared to experimental data for copper foams.
S.S. Sundarram, W. Li / Applied Thermal Engineering 64 (2014) 147e154 150
3.3. Effect of pore size
Models with pore size varying from 5 to 100 mm at a porosity of
94% were generated to study the pore size effect. The thickness of
the system was xed at 126 mm. A heating rate of 2.5e8 W/m
3
was
applied to the heat source and a convective cooling coefcient of
10 W/m
2
-K was applied to the top surface of the system. The
transient temperature response at the center of the power source is
shown in Fig. 7. It is seen that the temperature rises similarly for all
pore size cases until the PCM starts to melt at 321 K. A smaller pore
size results in a longer phase transition period. For example, the
transition zone lasts for 190 s for 5 mm pore size comparing to 120 s
for 50 mm pore size. After the PCM completes melting, the tem-
perature continues to rise. The temperatures for different pore sizes
eventually all converged to the same steady-state temperature.
However, before the equilibrium is reached, the temperatures are
dramatically different at any given time. For example, the temper-
ature at t =220 s is 335 K for the 5 mmpore size, while it is 378 K for
the 50 mm pore size. This temperature difference is signicant for
PCM-metal foam applications. For example, in microelectronic
cooling applications, the failure rate of IC chips is dependent on its
temperature with an Arrhenius relationship, as shown below [29],
l = Ce
Ea
kT
(7)
where l is the failure rate, E
a
is an activation energy, k is Boltz-
manns constant, T is the operating temperature, and C is a con-
stant. It can be found from the above equation that reducing the
operating temperature by 10