1 s2.0 S0307904X03000805 Main

Applied Mathematical Modelling 27 (2003) 491–513
www.elsevier.com/locate/apm
Approximate analytical model for solidification in a finite

PCM storage with internal fins
Piia Lamberg *, Kai Siren *
Helsinki University of Technology, HVAC Laboratory, P.O. Box 4400, 02015 Hut, Finland
Received 13 August 2002; received in revised form 10 February 2003; accepted 28 March 2003
Abstract
In latent heat storage internal heat transfer enhancement techniques such as fins have to be used because
of the low heat conductivity of the phase change material. Internal heat transfer enhancement is essential,
especially in a solidification process where the main heat transfer mode is conduction. The aim of this paper
is to present a simplified analytical model which predicts the solid–liquid interface location and temperature
distribution of the fin in the solidification process with a constant end-wall temperature in the finned two-
dimensional PCM storage. The analytical results are compared to numerical results and they show that the
analytical model is more suited to the prediction of the solid–liquid interface location than the temperature
distribution of the fin in the PCM storage. A new factor called the fraction of solidified PCM is also in-
troduced. The fraction of solidified PCM easily gives a good picture of how much of the storage is solidified
after a given time.
2003 Elsevier Science Inc. All rights reserved.
Keywords: PCM; Heat transfer; Analytical model; FEMLAB
1. Introduction
Energy storage is essential whenever the supply or consumption of energy varies independently
with time. Basically, there are three methods of storing thermal energy: sensible, latent and
thermochemical heat or cold storage. Thermochemical storage offers an order of magnitude larger
heat storage capacity compared to sensible heat storage. However, this technology is still in the
development phase. A latent heat storage system (LHTS) is preferable to sensible heat storage in
*
Corresponding authors. Tel.: +358-9-451-3597; fax: +358-9-451-3418 (P. Lamberg), Tel.: +358-9-451-3602; fax:
+358-9-451-3418 (K. Siren).
E-mail addresses: piia.lamberg@hut.fi (P. Lamberg), kai.siren@hut.fi (K. Siren).
0307-904X/03/$ - see front matter 2003 Elsevier Science Inc. All rights reserved.
doi:10.1016/S0307-904X(03)00080-5
492 P. Lamberg, K. Siren / Appl. Math. Modelling 27 (2003) 491–513
Nomenclature
cp heat capacity, J kg1 K1

D half thickness of the fin, m
E energy storage capacity, J
E0 energy storage per unit length, W m1
E00 energy storage per unit area, W m2
fi local solid fraction, –
H total enthalpy, J
h enthalpy, J
k heat conductivity, W m1 K1
l length, m
L latent heat of fusion, J kg1
q heat flow, W
q0 heat flow per unit length, W m1
q00 heat flux, W m2
S location of the phase change interface, m
St Stefan number, St ¼ ðcp DT Þ=L
T temperature, C
DTm solidification temperature range, C
t time, s
x distance in the x-direction, m
y distance in the y-direction, m
Dx space increment in the x-direction
Dy space increment in the y-direction
Greek symbols
a thermal diffusivity, m2 s1
q density, kg m3
e rate of solidified PCM
k root of the transcendental equation
h ¼ TTwTm
Tm
dimensionless temperature distribution
s ¼ ðqckpfÞt l2 dimensionless time
f f
Sy
c¼ lc
dimensionless rate of solid–liquid interface recession
g ¼ lxf dimensionless x-coordinate
k ¼ llcf cell aspect ratio
W ¼ lDc dimensionless half thickness of the fin
j ¼ kkfs ratio of the heat conductivities
ðqcp Þf ðTw Tm Þ
n¼ Lqs
modified Stefan number
P. Lamberg, K. Siren / Appl. Math. Modelling 27 (2003) 491–513 493
Subscripts and superscripts

c convection
d depth
f fin
l liquid
m solidification
old old value
p phase change material
s solid
t theoretical
w wall
x x-direction
y y-direction
2D two-dimensional
applications with a small temperature swing because of its nearly isothermal storing mechanism
and high storage density.
A common problem in latent heat thermal storages is the poor conductivity of the phase change
materials. The crystallising and thickening agents which prevent supercooling and phase sepa-
ration in the PCM lower the thermal conductivity of the PCM and the inhibiting convection
motion in the liquid PCM [1,2]. During the phase change the solid–liquid interface moves away
from the heat transfer surface. Thus, the surface heat flux decreases due to the increasing thermal
resistance of the growing layer of molten or solidified PCM. The problem arises especially in
solidification processes where the sole heat transfer mode is conduction.
PCM melts more quickly than it solidifies because natural convection speeds up the melting of
PCM. In solidification the heat transfers by conduction in solid PCM out from the storage. If
there is a temperature gradient in liquid PCM, natural convection exists in the liquid–solid in-
terface. But even very strong natural convection has a negligible effect on the solid–liquid interface
position compared to the effect of heat conduction in solid PCM [3,11]. PCM solidifies on the heat
transfer surface and acts as a self-insulator because of low heat conductivity. In practice some
kind of heat transfer enhancement technique has to be used in LHTS because of the low heat
conductivity of the PCM. Heat transfer in LTHS can be enhanced by the following techniques:
(a) Active methods such as agitators/vibrators, scrapers and slurries [4].

(b) Using microencapsulated PCM [5].
(c) Using PCM containing dispersed high conductivity particles or lessing rings [6].
(d) Using PCM graphite composite material [7].
(e) Using extended surfaces such as fins and honeycombs [1,2,11,12].
This paper focuses on the finned PCM storage and heat transfer enhancement in the solidifi-
cation process. Heat transfer during solidification in a storage with internal fins has been studied
numerically and experimentally by many authors. Velraj et al. [8] made an experimental analysis
and numerical modelling of inward solidification on a finned vertical tube for a latent heat storage
unit. The solidification time in a finned tube was reduced approximately by a factor 1=n (where n
is the number of fins in the tube) compared to the case without fins. Velraj et al. [2] also studied
experimentally and numerically different heat transfer enhancement techniques for a solar thermal
storage system focusing on fins, lessing rings and air bubbles.
Al-Jandal [9] studied experimentally what effects the fin, metal honeycomb and copper matrix
structure have on the total melting and solidification time. The results showed that the average
thermal conductivity enhancement factors for solidification are in the order of 1.7 and those for
melting in the order of 3.3. Natural convection has a significant effect on the acceleration of
melting. The average thermal conductivity enhancement factor was determined as a ratio of so-
lidification or melting time with fins and without fins.
Bugaje [10] conducted experiments on the use of methods for enhancing the thermal response of
paraffin wax heat storage tubes by the incorporation of aluminium fins and star structures. The
conclusion was that internal fins performed much better than the star matrices, reducing the
loading time in the order of 2.2 and the unloading time in the order of 4.2.
Padmanabhan and Murthy [1] presented a theoretical analysis for phase change in a cylindrical
annulus with axial rectangular fins between inner and outer tubes. The finite-difference method
was used. Based on the analysis, a working formula for the volume of the melt/frozen fraction for
PCM was introduced for engineering purposes. The analysis also indicated that the addition of
fins in the cylindrical annulus is advantageous for energy storage applications. The melting or
solidification time decreases with an increase in the number of fins. The fins should be long and
thin and they should be made of good thermal conductors.
Stritih and Novak [11] handled numerically and experimentally heat transfer enhancement in
the solidification process in a finned PCM storage with a heat exchanger. The conclusion arrived
at was that the biggest influence on heat transfer in the solidification process was the distance
between the fins. The thickness of the fin is not as influential.
Humphries and Griggs [12] studied numerically a rectangular phase change housing, using
straight fins as a heat transfer enhancer in a two-dimensional grid. The data were generated over a
range of realistic sizes, material properties and different kinds of thermal boundary conditions.
This resulted in a design handbook for phase change energy storage.
It is obvious that fins enhance the internal heat transfer of a phase change storage. Fins have a
larger influence on the solidification process than on the melting process. If fins are near each
other, the effect of natural convection decreases when conduction takes place as the prime heat
transfer mode in a melting process. Melting slows down while solidification speeds up when the
fins enhance conduction heat transfer in the storage. The loading and unloading times have to be
optimised to achieve the best technical performance for the storage. Here the geometry of the
latent heat storage plays a very important role.
Heat transfer in finned PCM storages has been widely studied but analytical models for the
solidification process have not been published. The main advantages of an analytical model are
simplicity and short computation times. These features can be valuable in certain situations, such
as in the pre-design stage of the storage. Absolute precision is not important in such a situation
but the speed of the calculation is, as it enables us to compare several alternatives within a rea-
sonable time.
Lamberg et al. have derived an analytical model for the melting process in a finned semi-infinite
storage [13]. This paper continues work already carried out on the solidification process. The
objective of this work was to develop an approximate analytical model for the solidification
process in a finite PCM storage with internal fins.
2. Finite PCM storage with internal fins
In this work the solidification process in a finite two-dimensional PCM storage with internal
fins is studied (Fig. 1). Basically, the PCM storage consists of a metallic housing in the shape of a
parallelepiped which is filled with PCM and straight metal fins. These improve heat transfer
between the housing and the PCM.
Initially the PCM is in a liquid phase in the storage. During solidification the PCM storage
operates as a heat source when the right and left end-walls are kept at a constant temperature, one
which is lower than the initial temperature of the liquid PCM (Tw < Ti ). The phase change ma-
terial releases energy primarily via the phase change process. Heat transfers by conduction along
the fins and through the solidified phase change material from the solid–liquid interface to the
end-walls.
This problem is a so called Stefan-type problem or a moving boundary problem, whose basic
feature is that the regions in which the partial differential equations are held are unknown and
must be found as part of the solution to the problem. This amounts to a non-linearity of geo-
metric nature even when the rest of the equations appear to be linear. Non-linearity is the source
of difficulties in moving boundary problems. The non-linearity destroys the validity of generally
used solving methods such as the superposition method and the separation of variables method
[15]. The analytical solution to this kind of two-dimensional heat transfer problem with a phase
change has not been found. However, the problem is sufficiently simple to be solved numerically
by using well-known numerical methods such as the enthalpy method or the effective heat ca-
pacity method, along with the finite difference or finite element procedure.
Fig. 1. A finite PCM storage with internal fins.

Fig. 2. Methodology to find out the effect of different simplifications.
In this paper a simplified one-dimensional analytical model based on a quasi-linear, transient,

thin-fin equation is presented. The model predicts the solid–liquid interface location and the
temperature distribution of the fins during the solidification process in the storage. A new factor
called the fraction of solidified PCM may be calculated from the results of the analytical solution.
In addition to the one-dimensional analytical model, the heat transfer in the PCM storage is
calculated with a simplified one-dimensional numerical model and a two-dimensional numerical
model. The one-dimensional numerical calculation is carried out using a programme known as
FEMLAB, which is a simulation package solving systems and coupled equations through the
finite element method in one-, two- and three-dimensions [14]. The two-dimensional numerical
calculation is based on the enthalpy method and implemented in the Digital Fortran 5.0 envi-
ronment. By comparing the results of different methods it is possible to draw conclusions about
the accuracy of the one-dimensional analytical model (Fig. 2).
3. One-dimensional approach
The heat transfer in a PCM storage with internal fins cannot be solved analytically in a two-
dimensional case. In this paper a simplified one-dimensional model is introduced to determine
analytically the temperature of the fin and the location of the solid–liquid interface during the
solidification.
The PCM storage has a symmetrical structure. Therefore, the analysis is reduced to handle only
a one symmetry cell where the planes of symmetry are in the middle of the fin and midway be-
tween the two adjacent fins (Fig. 3). The symmetry cell is divided into two regions. In region 1, the
only heat sink is the constant temperature end-wall. Here the fin does not influence the solidifi-
cation process. In region 2, both the wall and the fin transfer heat from the phase change material
to the environment.
3.1. Assumptions
Due to the non-linear, unsteady nature of the problem several assumptions have been made to
simplify the two-dimensional heat transfer problem.
Fig. 3. The symmetry cell of PCM storage with an internal fin.
1. Initially the liquid PCM and the fin are at the solidification temperature of the phase change
material Tm ¼ Tl ¼ Tf . Therefore, the heat conduction and natural convection in the liquid
PCM are considered to be negligible. The sole heat transfer is by conduction in the solid PCM.
2. The solidification temperature (Tm ) is assumed to be constant. In reality, a phase change ma-
terial has a solidification range (DTm ).
3. The temperature distribution of the fin is considered to be one-dimensional in the x-direction
because the fin is thin and the conductivity of the fin material is high.
4. In region 1, heat transfer is one-dimensional only in the x-direction. The fin does not affect heat
transfer in this region.
5. In region 2, the solid–liquid interface moves only one-dimensionally in the y-direction because
the heat is mainly transferred through the fin to the environment.
6. In region 2, the sensible heat of solid PCM is assumed to be negligible. The latent heat of fusion
is assumed to be the principal mode of energy storage. In region 1, the sensible heat is taken
into account.
7. The physical properties such as heat conductivity, heat capacity and density of the phase
change material and the fin are assumed to be constant.
3.2. Mathematical formulation
Following the previously introduced regions the mathematical formulation will also be divided
into two parts. In region 1 the solidification can be handled as a one-dimensional one-phase
Stefan problem [15], which is the simplest explicitly solvable moving boundary problem with
constant end-wall temperature and with constant thermophysical properties in the materials. In a
one-phase Stefan problem the heat equation for a solid phase change material and for a solid–
liquid interface with an initial and boundary conditions is:
o2 Ts 1 oTs
2
¼ ; t > 0; 0 6 x 6 lf ð1Þ
ox as ot
oSx ðtÞ oTs ðSx ; tÞ

ðqLÞs ¼ ks ; t>0 ð2Þ
ot ox
Sx ð0Þ ¼ 0 ð3Þ
Ts ðSx ; tÞ ¼ Tm ð4Þ
Ts ð0; tÞ ¼ Ts ðlf; ; tÞ ¼ Tw ð5Þ

In region 2 the movement of solid–liquid interface is assumed to occur only in the y-direction.
An arbitrary differential element, dx, is separated from the PCM storage to outline energy bal-
ances in Fig. 4. The energy balance of the element dx yields two equations, one for the fin and one
for the PCM.
The energy balance for the fin is
Ef00 ¼ q00x þ q00xþdx q00c ð6Þ
The rate equations are substituted into the energy balance Eq. (6) which can be rewritten with
initial and boundary conditions as
oTf o2 Tf ks
ðqcp Þf D ¼ kf D 2 ðTf Tm Þ; t>0 ð7Þ
ot ox Sy
Tf ðx; 0Þ ¼ Tm ð8Þ
Tf ð0; tÞ ¼ Tw ð9Þ
Tf ðlf ; tÞ ¼ Tw ð10Þ
The energy balance for the solid–liquid interface recession in the y-direction is
Ep0 ¼ q0w þ q0c ð11Þ
The released heat due to solidification in a dx wide element is
Fig. 4. Energy flows in a differential element in the finned PCM storage.

oSy
Ep0 ¼ ðqLÞs dx ð12Þ
ot
The heat transfer from the solid–liquid interface to the fin takes place by conduction and is
ks
q0c ¼ ðTf Tm Þ dx ð13Þ
Sy
The rate of heat flow per unit length from the solid–liquid interface to the constant temperature
end-walls qw is determined by conduction through the solid in the x-direction:
ks oSy
q0w ¼ ðTw Tm Þ dx ð14Þ
x ox
After substituting Eqs. (12)–(14) into the energy balance Eq. (11) it can be rewritten with the
initial condition as
oSy ks oSy ks
ðqLÞs ¼ ðTw Tm Þ þ ðTf Tm Þ; t>0 ð15Þ
ot x ox Sy
Sy ðx; 0Þ ¼ 0 ð16Þ
To enable a simpler manipulation, Eqs. (7)–(10), (15) and (16) are converted into dimensionless
form. The following dimensionless variables are introduced [16]:
T Tm
h¼ ; dimensionless temperature distribution
Tw Tm
kf t
s¼ ; dimensionless time
ðqcp Þf l2f
Sy
c¼ ; dimensionless rate of solid–liquid interface recession
lc
x
g¼ ; dimensionless x-coordinate
lf
lf
k¼ ; cell aspect ratio
lc
D
W¼ ; dimensionless half thickness of the fin
lc
ks
j¼ ; ratio of the heat conductivities
kf
ðqcp Þf ðTw Tm Þ
n¼ ; modified Stefan number
Lqs
Eqs. (7)–(10), rewritten using the dimensionless variables determine the dimensionless tem-
perature distribution as a function of dimensionless time and length along the finite fin with the
initial and boundary conditions:
oh o2 h k2 j h
¼ ; s>0 ð17Þ
os og2 W c
hðg; 0Þ ¼ 0 ð18Þ
hð0; sÞ ¼ 1 ð19Þ
hð1; sÞ ¼ 1 ð20Þ
Eqs. (15) and (16), rewritten in dimensionless form, determine the dimensionless rate of solid–
liquid interface recession as a function of dimensionless place along the fin:
oc nj oc h
¼ þ njk2 ; s>0 ð21Þ
os g og c
cðg; 0Þ ¼ 0 ð22Þ
Eqs. (17)–(22) form the basis for the simplified one-dimensional model.
3.3. Analytical solution of the simplified one-dimensional model
In region 1 the Stefan problem Eqs. (1)–(5) has a well-known analytical solution. According to
Neumann [15], the distance of the solid–liquid interface from the left end wall can be solved from
Eqs. (23) and (24):
pffiffiffiffiffiffi
Sx¼0 ¼ 2k as t ð23Þ
where k is a root of the transcendental equation
2 Stl cp ðTw Tm Þ
kek erfðkÞ ¼ pffiffiffi ¼ pffiffiffi ð24Þ
p L p
Due to the symmetry, the distance of the solid–liquid interface from the right end wall is
Sx¼Lf ¼ lf Sx¼0 ð25Þ
In region 2, for the dimensionless equations (17)–(20) no mathematically exact solution exists.
Eq. (17) is a parabolic partial differential equation and it is impossible to solve with known
methods because the dimensionless rate of the solid–liquid interface recession c is a variable in the
equation. Eq. (21) shows that c is a function of the dimensionless place g and time s. The following
assumptions are made to make the dimensionless equations solvable:
1. The dimensionless rate of the solid–liquid interface recession c is assumed to be constant in Eq.
(17). By introducing a new parameter m ¼ ðk2 jÞ=ðWcÞ, Eq. (17) it can be rewritten as
oh o2 h
¼ mh; s > 0 ð26Þ
os og2
2. The dimensionless temperature distribution h is assumed to be constant in Eq. (21) and the di-
mensionless rate of the solid–liquid interface recession c is assumed to be only the function of
dimensionless time. Hence, Eq. (21) can be rewritten as
oc h
¼ njk2 ; s>0 ð27Þ
os c
First, the dimensionless temperature distribution of the fin is solved from Eq. (26) and (18)–
(20). Eq. (26) is a parabolic partial differential equation and its solution is well-known with given
initial and boundary conditions [17]:
pffiffiffi
4 X 1 2 2
coshððg 0:5Þ mÞ ð1Þn eðð2nþ1Þ p sÞ
h¼ pffiffiffi ms cosðð2n þ 1Þpðg 0:5ÞÞ
coshð0:5 mÞ pe n¼0 ð2n þ 1Þ½1 þ fm=ðð2n þ 1Þ2 p2 Þg
ð28Þ
The dimensionless rate of the solid–liquid interface recession c is solved from the Eqs. (27) and
(22) to give
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
c ¼ 2njk2 hs ð29Þ
Eqs. (28) and (29) are solved and the solution for the dimensionless temperature distribution of
the fin and the dimensionless rate of the solid–liquid interface recession are found. Finally, the
temperature distribution of the fin and the distance of the solid–liquid interface in the y-direction
are
T ¼ Tm þ hðTw Tm Þ ð30Þ
and
Sy ¼ lc c ð31Þ
In conclusion, the analytical solution for the simplified one-dimensional heat transfer problem
consists of the Neumann solution for the solid–liquid interface location Sx in the x-direction in
region 1 (Eqs. (23)–(25)) and of the derived analytical solution for the temperature distribution of
the fin Tf and the solid–liquid interface location Sy in y-direction in region 2 (Eqs. (28)–(31)).
3.4. Numerical solution
Eqs. (7)–(10), (15) and (16) are solved numerically with the FEMLAB programme [14].
FEMLAB is based on the finite element method and is a powerful tool for solving coupled sys-
tems of non-linear partial differential equations in one-, two- or three-dimensions. The geometry
of the storage is defined. The equations are written in partial differential form in line with pro-
gramme definitions, and initial and boundary conditions are determined. The numerical solution
achieved is compared to the analytical solution to find out how the assumptions made influence
the simplified solution.
4. Two-dimensional approach
To determine the accuracy of the one-dimensional approach and the analytical model, a two-
dimensional numerical model is also developed and implemented. Here, the numerical analysis is
also reduced to only handle a one symmetry cell where the planes of symmetry are located in the
middle of the fin and midway between two adjacent fins (Fig. 2).
4.1. Assumptions
Assumptions 1, 2 and 7 made in Section 3.1 are also valid for the two-dimensional model. The
heat transfer is two-dimensional in the storage and the sensible heat of PCM is taken into account.
The whole storage is initially in its solidification temperature. Thus, there is no temperature
gradient in the liquid and natural convection is negligible and it can thus be ignored.
4.2. Mathematical formulation
The following two-dimensional heat equation can be applied for the fin and for the phase
change material:

o2 T o2 T 1 oT
þ ¼ ; t>0 ð32Þ
ox2 oy 2 a ot
T ðx; y; 0Þ ¼ Tm ð33Þ
T ð0; y; tÞ ¼ T ðlf ; y; tÞ ¼ Tw ð34Þ
where T is the temperature of the fin or PCM, a is the thermal diffusivity of the fin or PCM
respectively.
The two-dimensional heat equation for the solid–liquid interface location is [18]:
" 2 #
oTs oSy oSy
ks 1þ ¼ qs L ; y ¼ Sy ð35Þ
oy ox ot
The two-dimensional model with initial and boundary conditions can be solved numerically,
but it is not possible to find an analytical solution for this kind of heat transfer problem with a
phase change.
4.3. Numerical solution
The enthalpy method with a finite volume procedure is used to simulate the temperature dis-
tribution and the solid–liquid interface location in the storage [4,15]. The enthalpy method is used
in a particular way so that the only unknown variable is the temperature of the phase change
material and the solidification occurs at a uniform temperature. A Fortran code was developed to
carry out the calculations.
Zivkovic and Fujii [19] have used this method in a one-dimensional PCM storage. In this paper
the method is conducted, presented and used in a two-dimensional form. The enthalpy equation
for the conduction-controlled heat transfer in PCM and the fin can be written as

oH k o2 T o2 T
¼ þ ð36Þ
ot q ox2 oy 2
where H is the total enthalpy. An alternative form for Eq. (36) can be written by splitting the total
enthalpy into sensible and latent heat components:
H ¼ h þ Lfi ð37Þ
where h is sensible enthalpy and fi the local solid fraction. The sensible heat component can
further be written as
Z T
h¼ cp dT ð38Þ
Tm
where cp is the heat capacity of the material. For an isothermal phase change the local solid
fraction fi in the solidification process is defined as:
8
< 0 1 if T ¼ Tm
fi ðT Þ ¼ 1 if T < Tm ð39Þ
:
0 if T > Tm
Substituting Eq. (37) into Eq. (36) gives the following with initial and boundary conditions:

oh k o2 T o2 T ofi
¼ 2
þ 2 L ; t >0 ð40Þ
ot q ox oy ot
Tf ðx; y; 0Þ ¼ Ts ðx; y; 0Þ ¼ Tm ð41Þ
Tf ð0; y; tÞ ¼ Ts ð0; y; tÞ ¼ Tf ðlf ; y; tÞ ¼ Ts ðlf ; y; tÞ ¼ Tw ð42Þ
Tf ðx; D; tÞ ¼ Ts ðx; D; tÞ ð43Þ

In cases where solidification occurs in the node ði; jÞ, the solid fraction lies in the interval [0,1]
and oh=ot ¼ 0. Eq. (40) is fully implicit discretized in each internal node ði; jÞ and it can be written
as
ki Dt ki Dt
fi;j ¼ fi;jold þ 2
ðTi1;j 2Ti;j þ Tiþ1;j Þ þ ðTi;j1 2Ti;j þ Ti;jþ1 Þ ð44Þ
qi L Dx qi L Dy 2
where Dx is the space increment in the x-direction and Dy is the space increment in the y-direction.
Eq. (44) updates the solid fraction during the phase change process.
In cases where the PCM is fully solid or liquid, Eq. (40) reduces to the ordinary heat diffusion
equation and its discretized form is
ki Dt ki Dt
Ti;j ¼ Ti;jold þ 2
ðTi1;j 2Ti;j þ Tiþ1;j Þ þ ðTi;j1 2Ti;j þ Ti;jþ1 Þ ð45Þ
qi cpi Dx qi cpi Dy 2
At the end of each time step, the check for the start and/or end of the phase change has to be
performed throughout the entire domain. If solidification begins or ends during the calculated
time step, the calculation has to be performed again. When the solidification begins, the sensible
heat released to lower the temperature of the control volume to the solidification temperature can
not be used for solidification in the PCM, and when solidification ends the heat released to
complete solidification cannot be used either to lower the temperature of the PCM.
A Visual Fortran 5.0 code was developed to solve Eqs. (41)–(45) with different time steps. In the
simulation model the storage consists of a two-dimensional fixed grid. The space increments
depend on the dimensions of the storage.
After each time step each node gets a new value for the temperature and for the solid fraction.
The location of the solid–liquid interface is interpolated between the nodes.
The model was tested by comparing the released energy from the storage during a specified
calculation period to the theoretical storage capacity. The theoretical capacity Et0 of the storage
with a temperature difference T1 T2 is
Et0 ¼ mp cp ðT1 T2 Þ þ mf cpf ðT1 T2 Þ þ mp hf ð46Þ
where mp is the mass of the PCM and mf the mass of the fin. This was compared with the released
0
energy E2D from the storage during the two-dimensional numerical calculations
X
m X
n
kj DTk;j
0
E2D ¼ Dy Dtk ð47Þ
k¼1 j¼1
Dx
X
m
Dtk ¼ t; j ¼ 1; 2; . . . ; n and k ¼ 1; 2; . . . ; m
k¼1
where t is the total calculation time and DTk;j the temperature difference in node j in time k.
The error produced was smaller than 0.1% when the theoretical storage capacity and the re-
leased energy during the two-dimensional calculation were compared with each other. Zivkovic
and Fujii [19] examined the accuracy of the method in a one-dimensional case. They obtained only
0.12% error in the location of the solid–liquid interface in one-dimensional calculations when
compared to the enthalpy based approach where the enthalpy is a function of the temperature.
5. Results
5.1. Test cases
To find out the accuracy and performance of the one-dimensional approach and the analytical
solution, six different test cases were chosen. In the test cases laboratory grade pure n-octadecane
paraffin with an aluminium fin (three cases) and a commercial grade salt hydrate (Climsel 23) with
a steel fin (three cases) were chosen. n-Octadecane paraffin has a relatively uniform melting point,
but the salt hydrate melts and solidifies in temperature range 21–25 C [16,20]. However, it has
been assumed that the both PCMs have a discrete melting temperature, though in reality they
have a melting region. The melting temperature of the salt hydrate was fixed at 23 C, which is the
Table 1
Physical properties of the phase change and the fin materials
Property Paraffin n-octadecane Salt hydrate Climsel Aluminium fin Steel fin
23
Density (q), kg m3 777 1480 2713 7854
Heat conductivity (k), W m1 K1 0.149 0.6 180 60.5
Heat capacity (cp ), J kg1 K1 2660 2660 960 434
Latent heat of fusion (L), J kg1 241,360 148,000 – –
Melting/solidification temperature
(Tm ), C 28 23 – –
average temperature in the melting and solidification range of salt hydrates. The physical prop-
erties of the phase change and the fin materials are shown in Table 1 [16,21].
In the test cases the initial temperature of the storage is the solidification temperature and the
PCM is in liquid state. The temperature difference (Tm Tw ) in all cases is fixed at 15 C. The half
thickness D of the fin has a constant value. Otherwise, the geometry of the storage is varied in
different cases. The width to height ratio k ¼ Lf =Lc varies between 0.2 and 5.
The geometry of the storage affects the speed of the solidification. When k is small, the heat
transfers mainly through the wall from the solid–liquid interface to the environment. When k is
close to one, the influence of the wall and the fin are of the same magnitude and when k is large,
the fin has a major role in heat transfer. Therefore, when k is small, the speed of the solid–liquid
interface in the x-direction determines the solidification time and when k is large, the speed of the
interface in the y-direction determines the solidification time. Three different k values were used.
The calculation time is also varied. The geometry, initial temperatures, materials and calculation
times used in different test cases are shown in Table 2.
For each test case the location of the solid–liquid interface and the temperature of the fin are
calculated using three different methods:
1. simplified one-dimensional analytical model,

2. simplified one-dimensional numerical model, and
3. two-dimensional numerical model.
Table 2
The geometry of the storage, initial temperatures, materials and calculation time used in different test cases
Case 1 Case 2 Case 3 Case 4 Case 5 Case 6
PCM Paraffin Paraffin Paraffin Salt hydrate Salt hydrate Salt hydrate
Fin Aluminium Aluminium Aluminium Steel Steel Steel
Tm, C 28 28 28 23 23 24
Tw, C 13 13 13 8 8 8
Lf , m 0.01 0.05 0.05 0.01 0.05 0.05
Lc, m 0.05 0.05 0.01 0.05 0.05 0.01
k ¼ Lf =Lc 0.2 1 5 0.2 1 5
D, m 0.0005 0.0005 0.0005 0.0005 0.0005 0.0005
t, s 723 1085 1085 169 4226 1127
s, – 500 30 30 30 30 6
The results achieved with different methods are compared with each other in line with the
methodology shown in Fig. 2.
5.2. Accuracy of the one-dimensional approach
Several assumptions were made in the simplified one-dimensional model (see Section 3.1). By
comparing the numerical results of the simplified one-dimensional model to the numerical results
of the two-dimensional model it is possible to draw conclusions about the error made when using
a one-dimensional model instead of a two-dimensional model.
The one-dimensional numerical results in region 2 are calculated using FEMLAB (Eqs. (7)–(10)
and (15) and (16), and the two-dimensional numerical results using the enthalpy method (Eqs.
(44)–(46)) with Visual Fortran 5.0 code. In region 1, the location of the solid–liquid interface has
an exact analytical solution from Neumann. Therefore, the one-dimensional numerical solution in
region 1 does not need to be calculated. The step size is Dg ¼ 0:1 in the x-direction both in one-
and two-dimensional numerical results. The step size does not affect the results. In Fig. 5a–f, the
temperature distributions of the fin in different test cases are shown.
It seems that the accuracy of the one-dimensional numerical solution for the temperature
distribution of the fin is good in cases where the width to height ratio (k) is small. In this case the
heat transfers mainly through the wall to the environment. The fin temperature approaches
the wall temperatures quite quickly and the small amount of heat, which is transferred through
the fin, is conducted effectively. The heat transfer is close to one-dimensional in the x-direction
and, therefore, the one-dimensional model gives good results.
The biggest error is made when k approaches value one. In this scenario, the tempera-
ture difference between one-dimensional and two-dimensional numerical results is 1–2 C
in both paraffin and salt hydrate storage. When k is equal to one the wall has a strong effect
on the solidification process, but in the simplified one-dimensional model it is assumed that
the solid–liquid interface moves only in the y-direction when the temperature distribution of
the fin is solved. It will distort the temperature distribution of the fin. In this case the storage
should be handled two-dimensionally instead of one-dimensionally to achieve more accurate
results.
When k is large, the accuracy of a one-dimensional numerical solution for the temperature
distribution of the fin is relatively good because the heat transfers mainly through the fin to the
environment, as is assumed in the one-dimensional model.
In Fig. 6a–f, one- and two-dimensional numerical predictions of the solid–liquid interface lo-
cation in the different test cases are shown.
When k is small, the speed of the solid–liquid interface in the one-dimensional cases is slower
than in the two-dimensional cases. It is possible to see from the figures that the cross-section of the
one-dimensional results in region 1 and region 2 is sharp. It should be more round in the corners
because of the two-dimensional heat transfer in that section. When k increases, the location of the
solid–liquid interface in the one-dimensional model will go ahead of the interface location in the
two-dimensional model. This is due to the heat capacity of the PCM, which was ignored in region 2
in the one-dimensional approach. When k is large, the fin plays the most important role in
transferring heat out from the storage. In reality the heat capacity of the PCM slows down the
speed of the solid–liquid interface in the solidification process. However, the error made in the
Fig. 5. (a–f) The one-dimensional analytical and numerical results and two-dimensional numerical results of the
temperature distribution of the fin in different test cases.
solid–liquid interface location when using the one-dimensional model is not large in the calculated
cases.
The one-dimensional model gives quite good results for the solid–liquid interface location but
the results for the temperature distribution of the fin are not so good, except where k is small. All
Fig. 6. (a–f) The one-dimensional analytical and numerical results and two-dimensional numerical.
in all, the one-dimensional model is better at predicting the solid–liquid interface location than the
temperature distribution of the fin.
5.3. Accuracy of the analytical solution
The simplified one-dimensional model consists of a Neumann solution in region 1 and a derived
analytical solution in region 2. Since the Neumann solution is an exact solution, there is no need
to study its accuracy. However, several assumptions are made when solving Eqs. (17)–(22) in
region 2 to achieve the derived analytical solution for the location of solid–liquid interface in the
y-direction and the temperature distribution of the fin.
By comparing the derived analytical solution to the one-dimensional numerical solution, with
both having the same physical assumptions and mathematical formulation, it is possible to draw
conclusions about the accuracy of the analytical solution.
The analytical results are calculated using Eqs. (23)–(25) and (28)–(31). The one-dimensional
numerical calculation based on Eqs. (7)–(10), (15) and (16) is carried out using FEMLAB [14]. In
Fig. 5a–f, the one-dimensional analytical and numerical results of the temperature distribution of
the fin in different test cases are shown.
The maximum temperature difference (DTmax ) between one-dimensional analytical and nu-
merical results is small, especially in paraffin storage. The heat conductivity of the paraffin is poor
whilst the heat conductivity of the aluminium fin is good. The temperature distribution of the fin
approaches the temperature of the wall quite quickly. At the same time the paraffin solidifies
slowly. It can be assumed that the solid–liquid interface location is only a function of time in
paraffin storage. This assumption is used when solving the equations. In salt hydrate storage the
steel fin does not rapidly achieve the temperature of the wall when compared to the aluminium fin.
This is because of its poorer heat conductivity. In salt hydrate storage the solid–liquid interface
location is more a function of position and time than a function of time, as has been assumed
when solving the equations. This causes an error in the results of the temperature distribution of
the fin in salt hydrate storage.
All in all, the best results are achieved when k is small. The error increases when k increases.
However, the errors made are small, a maximum 0.012 C for paraffin storage and 0.58 C in salt
hydrate storage in these cases. The mathematical solution for the temperature distribution of the
fin works well, especially when paraffin storage with an aluminium fin is concerned.
In Fig. 6a–f, the one-dimensional analytical and numerical results of the solid–liquid interface
location in different test cases are shown.
The difference in solid–liquid interface location Sy between one-dimensional analytical and
numerical results is relatively small in both paraffin and salt hydrate storage. The biggest error is
made in the cases when is large because of the assumption that the location of the solid–liquid
interface is only a function of time. In a long fin the location should be a function of time and
position. However, the mathematical solution for the solid–liquid interface location performs
adequately.
Thus, the assumptions made when solving the equations mathematically do not greatly affect
the accuracy of the one-dimensional analytical solution.
5.4. Accuracy of the one-dimensional analytical model
Finally, when the one-dimensional analytical model is compared to the two-dimensional nu-
merical model it is possible to draw conclusions about the accuracy of the analytical model. Fig.
5a–f shows one-dimensional analytical and two-dimensional numerical results of the temperature
distribution of the fin in the different test cases.
The results are similar to Section 5.2. The maximum temperature difference (DTmax ) between the
one-dimensional analytical results and the two-dimensional numerical results are small when k is
small, and are largest when k approaches unity. The analytical model predicts the fin temperature
well when the geometry of the cell is close to a one-dimensional case (k is small or large) and fails
to some extent in a pure two-dimensional situation (k ¼ 1).
The one-dimensional analytical and two-dimensional numerical results for the solid–liquid
interface location are shown in Fig. 6a–f.
When k is small, the solid–liquid interface in the y-direction moves more slowly in the one-
dimensional analytical model than in the two-dimensional model. When k increases, the location
of the solid–liquid interface in a one-dimensional analytical model will go ahead of the interface
location in a two-dimensional model. The same conclusions were evident in Section 5.2.
In general, the simplified analytical model more accurately predicts the solid–liquid interface
location than the temperature distribution of the fin in the PCM storage with internal fins. The
assumptions made when the two-dimensional heat transfer problem was changed to the one-
dimensional problem more greatly affect the accuracy of the one-dimensional analytical model
than the assumptions made when solving the one-dimensional equations mathematically.
5.5. Fraction of the solidified PCM
The fraction of solidified PCM describes how much of the storage is solidified after a certain
time. The factor takes values between 0 (the storage is totally liquid) and 1 (the storage is totally
solid). It describes with one value the relative amount of released energy from the storage.
Therefore, it is a useful quantity when different methods are compared with each other.
The factor is defined as the volume of the solidified PCM related to the total volume of PCM in
the storage:
2Sx ðLc D Sy Þ þ Sy Lf
e¼ ð48Þ
ðLc DÞLf
where Sy is the average value of the solid–liquid interface location in the y-direction along the fin
length and Sx and Sy can be calculated from Eqs. (28)–(31).
The rate of solidified PCM calculated from the one-dimensional analytical (Eq. (47)) and two-
dimensional numerical results in different test cases are shown in Table 3.
The difference is smaller than 2.8 % in all cases. If we consider Fig. 6a–f, it can be seen that in
many cases the analytical results of the solid–liquid interface location are ahead of the numerical
Table 3
The rate of the solidified PCM calculated with one-dimensional analytical and two-dimensional numerical models in
different test cases
Case k ¼ Lf =Lc Time, s Tw Tm , C D, m Fraction Fraction Error, %
analytical numerical
Case 1, para 0.2 724 15 0.0005 0.787 0.799 )1.2
Case 2, para 1 1085 15 0.0005 0.258 0.233 2.4
Case 3, para 5 1085 15 0.0005 0.543 0.539 0.4
Case 4, salt 0.2 169 15 0.0005 0.733 0.749 )1.7
Case 5, salt 1 4226 15 0.0005 0.804 0.826 )2.3
Case 6, salt 5 1127 15 0.0005 0.819 0.791 2.8
results of the interface. At the same time the cross-section of the analytical results in region 1 and
region 2 is sharp, though it should be more rounded at the corners. Thus, the numerical results for
the solid–liquid interface location are ahead of the analytical results for the interface locations in
the corners. The interaction of these errors diminishes the error made when calculating the rate of
solidified PCM.
The error is largest in the cases when k is equal to one, except in case 6. In case 6 the solid–
liquid interface (Sy ) in the one-dimensional calculation is much further ahead of the interface
achieved in the two-dimensional numerical calculation because the heat capacity of the PCM is
ignored in the one-dimensional calculation. When k is large, the value Sy is dominant in Eq. (48),
therefore the error in calculating Sy in a one-dimensional case affects the rate of solidified PCM.
It seems that the factor Eq. (48) gives a good picture of how much of the storage is solidified
without the need to draw a solid–liquid interface figure. It is possible to see from the results that
with small k values, the speed of the solidification is slower in the analytical model than in the two-
dimensional numerical model. In reality, when k is smaller than one, the heat transfers mainly in
the x-direction in the storage. In this paper it has been assumed that the heat transfers only in the
x-direction and the heat transfer in the y-direction is assumed to be negligible in region 1.
Therefore, the location of the solid–liquid interface moves more slowly in analytical results than in
numerical results.
When k increases, the solid–liquid interface calculated by the analytical model starts to move
faster than the interface calculated with the two-dimensional analytical model. When k is over
one, the heat transfer is dominant in the y-direction.
In the approximate analytical model in region 2 the sensible heat of the PCM is ignored.
Therefore, the analytical method overestimates the actual solid–liquid interface location.
The factor is also useful when dimensioning PCM storage and also when different geometry are
compared with each other. With the analytical model it is possible to access how the geometry of
the storage and the material properties effect the solidification time of the storage. However, the
fraction of solidified PCM is a bulk measure and it misses microscopic effects in the PCM op-
eration. The factor is a useful tool together with analytical results that indicate the location of the
solid–liquid interface. For example, Fig. 7 shows the rate of the solidified PCM as a function of
time for paraffin storage with the aluminium fin (case 2) and for salt hydrate storage with the steel
Fig. 7. The fraction of solidified PCM for n-octadecane storage with the aluminium fin and for salt hydrate storage
with the steel fin, Tw Tm ¼ 15 C, D ¼ 0:0005 m, k ¼ Lc =Lf ¼ 0:05=0:05 ¼ 1.
fin (case 4). The volume of the storage as well as the temperature difference between the wall and
the phase change material are the same in both cases.
It can be seen that salt hydrate storage solidifies much more quickly than paraffin storage
because of salt hydrates better heat conductivity. The mass of the phase change material in the
storages is equal. The storages release different amounts of heat because of different densities and
the latent heat of fusion. When the storage volume is in both cases 125 cm2 (the storage size
0.05 0.05 0.05 m), the released energy from the paraffin is E ¼ 25:5 kJ and from salt hydrate
storage E ¼ 27:1 kJ during the phase change. In this case the salt hydrate storage works more
effectively than the paraffin storage.
6. Conclusions
This paper presents a simplified analytical model based on a quasi-linear, transient, thin-fin
equation which predicts the solid–liquid interface location and temperature distribution of the fin
in a solidification process with constant end-wall temperatures in a finite PCM storage. The
analytical solution consists of the well-known Neumann solution for predicting the solid–liquid
interface location Sx in the x-direction, as well as a derived analytical solution to predict the
temperature distribution of the fin Tf in the x-direction and the solid–liquid interface location Sy in
the y-direction. Some simplifying assumptions are made to make it possible to find an analytical
solution to the problem.
A new factor, called the fraction of solidified PCM, is also presented. The factor indicates how
much PCM has solidified after a certain time. In addition to the one-dimensional analytical
model, the heat transfer in the PCM storage is calculated using the simplified one-dimensional
numerical model and the two-dimensional numerical model. Six different test cases were chosen to
compare the results by using same initial and boundary values and material properties. The
following conclusions are made concerning the accuracy of the one-dimensional analytical model:
• The geometry of the computational domain (cell) seems to be one of the most important factors
in explaining the performance of the developed one-dimensional model. When the width to
height ratio (k) is small or large, the accuracy of the one-dimensional approach is relatively
good because the heat transfer is near to a one-dimensional model in the x-direction (k is small)
or the y-direction (k is large). When k is equal to one, the heat effectively transfers both in the
x- and y-direction. By using a one-dimensional approach instead of a two-dimensional
approach, the accuracy is slightly reduced. However, the results are still satisfactory, especially
for the solid–liquid interface location.
• The assumptions made in order to enable the analytical solution have a minor influence on the
results. Thus, the accuracy of the analytical solution is good.
• The assumptions made in simplifying the two-dimensional heat transfer problem into a one-
dimensional form affect the accuracy to a greater extent than the assumptions made when solv-
ing the one-dimensional equations analytically.
• All in all, the accuracy of the approximate analytical model is good. The model is useful in the
pre-design stage of the storage. With the model it is possible to compare several storage alter-
natives in a reasonable time.
Acknowledgements
The support of The Academy of Finland and the Graduate School of Building Physics at HUT
are greatly acknowledged by the authors.
References
[1] P.V. Padmanabhan, M.V. Murthy, Outward phase change in a cylindrical annulus with axial fins on the inner tube,
International Journal of Heat and Mass Transfer 29 (12) (1986) 1855–1868.
[2] R. Velraj et al., Heat transfer enhancement in a latent heat storage system, Solar Energy 65 (3) (1999) 171–180.
[3] P.G. Kroeger, S. Ostrach, The solution of a two-dimensional freezing problem including convection effects in the
liquid region, International Journal of Heat and Mass Transfer 17 (1973) 1191–1207.
[4] G. Lane, in: Solar Heat Storage: Latent Heat Material, vol. 1, CRC Press, 1983.
[5] www.frisby.com.
[6] R. Siegel, Solidification of low conductivity material containing dispersed high conductivity particles, International
Journal of Heat and Mass Transfer 22 (1977) 1087–1089.
[7] H. Mehling, S. Hiebler, F. Ziegler, Latent heat storage using a PCM-graphite composite material, in: Terrastock
8th International Conference on Thermal Energy Storage, 28 August–1 September, 2000, Proceedings, vol. 1, 2000,
pp. 375–380.
[8] R. Velraj, R. Seeniraj, B. Hafner, et al., Experimental analysis and numerical modelling of inward solidification on
a finned vertical tube for a latent heat storage unit, Solar Energy 60 (5) (1997) 281–290.
[9] S. Al-Jandal, Experimental study of temperature dependent heat transfer during melting and solidification
processes, in: 2nd World Renewable Energy Congress, Reading University, UK, 13–18 September 1992, pp. 1097–
1105.
[10] I. Bujage, Enhancing thermal response of latent heat storage systems, International Journal of Energy Research 21
(1997) 759–766.
[11] U. Stritih, P. Novak, Heat transfer enhancement at phase change processes, in: Terrastock 8th International
Conference on Thermal Energy Storage, 28 August–1 September, 2000, Proceedings, vol. 1, 2000, pp. 333–338.
[12] W. Humphries, E. Griggs, A design handbook for phase change thermal control and energy storage devices, NASA
TP-1074, 1977.
[13] P. Lamberg, K. Siren, Analytical model for melting in semi-infinite PCM storage with an internal fin, Heat and
Mass Transfer 39 (2003) 167–176.
[14] www.femlab.com.
[15] V. Alexiades, A.D. Solomon, Mathematical Modelling of Melting and Freezing Processes, Hemisphere Publishing
Corporation, USA, 1993.
[16] R. Henze, J. Humphrey, Enhanced heat conduction in phase-change thermal energy storage devices, International
Journal of Heat and Mass Transfer 24 (1981) 450–474.
[17] H. Carslaw, J. Jaeger, Conduction of Heat in solids, second ed., Oxford University Press, New York, 1959.
[18] K.A. Rathjen, M.J. Latif, Heat conduction with melting or freezing in a corner, Journal of Heat Transfer (1971)
101–109.
[19] B. Zivkivic, I. Fujii, An analysis of isothermal phase change of phase change material within rectangular cylindrical
containers, Solar Energy 70 (1) (2001) 51–61.
[20] www.climator.com.
[21] F.P. Incropera, D. De Witt, Fundamentals of Heat and Mass Transfer, John Wiley & Sons Inc., Canada, 1990.

1 s2.0 S0307904X03000805 Main

Uploaded by

Copyright:

Available Formats

1 s2.0 S0307904X03000805 Main

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

1 s2.0 S0307904X03000805 Main

Uploaded by

Copyright:

Available Formats

Applied Mathematical Modelling 27 (2003) 491–513

Approximate analytical model for solidiﬁcation in a ﬁnite

Keywords: PCM; Heat transfer; Analytical model; FEMLAB

cp heat capacity, J kg1 K1

Subscripts and superscripts

(a) Active methods such as agitators/vibrators, scrapers and slurries [4].

2. Finite PCM storage with internal ﬁns

Fig. 1. A ﬁnite PCM storage with internal ﬁns.

Fig. 2. Methodology to find out the effect of different simplifications.

In this paper a simpliﬁed one-dimensional analytical model based on a quasi-linear, transient,

Fig. 3. The symmetry cell of PCM storage with an internal ﬁn.

3.2. Mathematical formulation

oSx ðtÞ oTs ðSx ; tÞ

Ts ð0; tÞ ¼ Ts ðlf; ; tÞ ¼ Tw ð5Þ

Fig. 4. Energy flows in a differential element in the finned PCM storage.

3.3. Analytical solution of the simpliﬁed one-dimensional model

3.4. Numerical solution

4.2. Mathematical formulation

T ð0; y; tÞ ¼ T ðlf ; y; tÞ ¼ Tw ð34Þ

4.3. Numerical solution

Tf ðx; y; 0Þ ¼ Ts ðx; y; 0Þ ¼ Tm ð41Þ

Tf ð0; y; tÞ ¼ Ts ð0; y; tÞ ¼ Tf ðlf ; y; tÞ ¼ Ts ðlf ; y; tÞ ¼ Tw ð42Þ

Tf ðx; D; tÞ ¼ Ts ðx; D; tÞ ð43Þ

5.1. Test cases

1. simpliﬁed one-dimensional analytical model,

5.2. Accuracy of the one-dimensional approach

5.3. Accuracy of the analytical solution

5.4. Accuracy of the one-dimensional analytical model

5.5. Fraction of the solidiﬁed PCM

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.