8th Annual International Energy Conversion Engineering Conference

25 - 28 July 2010, Nashville, TN

AIAA 2010-6680

Numerical and Analytical Simulation of a Thermal Storage

Unit for Building Hot Water
Monica F. Bonadies1, Son Ho2, and J.S. Kapat3
University of Central Florida, Orlando, FL, 32816

The modeling of a latent heat storage unit can be done analytically and numerically, with the
two approaches intersecting to help to validate simpler models. In this paper, numerical
simulation with a two dimensional FLUENT model is used to determine the accuracy of the
fin resistance used in a heat exchanger. This resistance model can then be used to determine
which changes within the unit can improve it with a relatively quick calculation. FLUENT
tracks the melting and solidification of a material through the enthalpy-porosity method,
which is utilized in this simulation. The one dimensional analytical model is based upon the
resistance model approach from heat transfer, and can be used to quickly analyze what
effect a change in the storage unit has on its operation. In the future, this model will be used
to parametrically analyze the storage unit and optimize it in terms of the number of fins, the
heat exchange material, and the heat transfer fluid. The analytical model developed shows
the same trends as the FLUENT model within an acceptable amount of error on both the
melting and solidification processes. The comparison of the analytical and FLUENT models
to established analytical models and experimental data shows the fact that several factors
need to be taken into consideration when modeling a phase change process, including the
shape of the melt front, natural convection within the melted wax, and the fin dimensions of
the heat exchanger.

Nomenclature
A
Cp
D
erf
h
H
HTF
I
k
K
L
m
l
PCM
q
q
r
R
S

=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=

area
constant pressure specific heat
fin half thickness
error function
heat transfer coefficient
latent heat of fusion
heat transfer fluid
Bessel function of the first kind
thermal conductivity
Bessel function of the second kind
length of pipe
fin heat transfer parameter
length
phase change material
heat in Watts
heat flux (Watts per meter squared)
radius
resistance
phase change material thickness

Graduate Research Assistant, Department of Mechanical, Materials, and Aerospace Engineering, 4000 Central
Florida Boulevard, Building 40, Room 307, Orlando, FL 32816, Student Member of AIAA.
2
Post-Doctoral Research Associate, Department of Mechanical, Materials, and Aerospace Engineering, 4000
Central Florida Boulevard, Building 40, Room 307, Orlando, FL 32816, AIAA Member.
3
Professor, Department of Mechanical, Materials, and Aerospace Engineering, 4000 Central Florida Boulevard,
Building 40, Room 307, Orlando, FL 32816, AIAA Associate Fellow.
1
American Institute of Aeronautics and Astronautics
Copyright 2010 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.

t
T
x

= time
= temperature
= solution of transcendental equation

Greek

= coefficient of thermal expansion

= thickness of freezing or melting front

Subscript
0
1
2
cond
conv
fin, f
freeze
fusion
mean
melt
old
pipe
unit
wall, w
wax
x
y

=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=

zero order Bessel function

pertaining to inner radius of fin or first order Bessel function
pertaining to outer radius of fin
pertaining to conduction
pertaining to convection
pertaining to the fin
pertaining to the freezing process
indicates latent heat of fusion
pertaining to mean temperature
pertaining to the melting process
pertaining to previous time iteration
pertaining to pipe portion of the heat exchanger
pertaining to storage unit
pertaining to the wall
pertaining to wax
pertaining to x direction
pertaining to y direction

I. Introduction and Review of the Literature

olar energy has been used for many years to make use of the energy that falls to earth each day from the sun.
The collection methods vary from photovoltaic cells to evacuated tubes to parabolic dish collectors at the largest
scale. The benefit of using solar thermal as opposed to photovoltaic is that a solar thermal plant produces steam,
which is then used in a steam turbine an already existing technology with several decades of improvement. [1]
Furthermore, a smaller scale amount of solar thermal collectors can be used for home hot water heating through a
heat exchange system. This provides an alternative to the conventional hot water heater, which makes up a
significant portion of a typical
familys electricity usage.[2]
The use of solar energy has the
benefit of being environmentally
friendly in terms of nearly zero
carbon emissions, but has the
drawback of only being available
while the sun shines. This means
that there is a benefit in adding a
storage
unit
for
minimal
interruption in the thermal energy
supply [3]. The most compelling
reason to use thermal energy
storage is the fact that the storage
unit effectively evens out the
thermal energy available over a
day, since it stores the energy in
excess of that which is used
during the day. For instance, the
Figure 1. Thermal Storage with Internal Fins [5]
maximum solar radiation may
occur at mid-day, but the hot
2
American Institute of Aeronautics and Astronautics

water need at that time is not necessarily the greatest. So, the excess radiation can be stored in a solar thermal unit
to provide hot water in the evening.
Storage media is divided into two main groups: sensible heat storage and latent heat storage. Sensible heat
storage has been implemented in terms of heated tanks of water or beds of rock underneath a home, but latent heat
storage provides a unique benefit via its phase change. During a phase change process, the temperature remains
constant and the energy received by the medium melts it. With this fact, it can be determined that the latent heat
storage provides more energy storage than a sensible heat storage unit of the same volume. There are three
necessary components in a latent heat energy storage system: a suitable PCM, a suitable heat exchange surface, and
a suitable vessel that is compatible with the PCM. The materials available for this application may be organic (such
as paraffin wax), inorganic (such as molten salts), or a combination of both (such as a graphite matrix impregnated
with paraffin wax). In this study, an organic material, paraffin wax, is used. The structure of paraffin wax is a chain
of n-alkanes, and the melt point and latent heat of fusion increase with chain length. [4]
Modeling of a phase change material within a storage configuration can be done analytically or numerically,
with advantages and disadvantages corresponding to each. These will be explained in specific detail along with an
example of application of these methods, but the main advantage of using an analytical or numerical model is that a
design can be quickly analyzed, compared to the time it takes to construct a prototype. Numerous analytical and
numerical models exist, with their main differences being the design of the method of heat exchange from the
working fluid of the solar thermal system to the phase change material. In the case presently studied, the analytical
model of Lamberg and Siren most closely matches the conditions, with a close second being the circular fin model
as described in a typical heat transfer text. Both these models will be explored in more detail, along with a
numerical simulation using FLUENT, based upon other thermodynamically valid FLUENT models
One of the simplest forms of enhancing thermal conductivity is the addition of rectangular fins within the phase
change material. Of course, the configuration and number of fins must be optimized so that natural convection,
which speeds the melting process, can take place. Also,
the fins cannot be too far apart, as this would create too
big an area for them to transfer heat across. Reddy[5]
studied the configuration of rectangular fins within an
integrated collector system with paraffin wax with
embedded fins set beneath a specially coated piece of
glass, as shown in Figure 1. Solar radiation would be
trapped beneath this glass and the thermal energy
transferred to the paraffin wax. The wax in turn would
transfer heat to the water bath beneath it, which would be
pumped in during the night time. The entire unit was
tilted to optimize the suns position in the sky as well.
Once the boundary conditions and heat loss values were
determined for this configuration, FLUENT was used to
numerically analyze the temperature distributions for 0, 4,
9, and 19 fins within the paraffin wax.
FLUENT does not directly give the location of the
melting front, however. Instead, a mushy region can be
tracked, with 0 being completely solid material and 1
being a completely melted material. This modeling
Figure 2. Energy Flows in a Differential
method was used along with a sinusoidal heat transfer to
Element of Finned Thermal Storage [7]
represent the daily variation of the sun. Results can be
shown in terms of temperature contours and liquid
fraction variations, which can be given after a certain amount of time. The most important result was that the ninefin configuration maintained the best heat level, allowed for the maximum amount of natural convection, and had
the highest average water temperature and lowest night time temperature drop. This study shows that although
placing fins in the phase change material does increase its thermal conductivity, there is a maximum amount of fins
one can use to optimize the performance of the storage unit.

3
American Institute of Aeronautics and Astronautics

FLUENT has also been analyzed in

terms of its method of solving phase
change problems.
Pinelli et al[6]
identified the fact that FLUENTs phase
change solving method was meant for a
few specific applications, so it must be
modified accordingly to solve problems
outside that realm of applications. They
applied FLUENT in solving a phase
change occurring within a cylinder with
the top of the cylinder heated at a
temperature greater than the melting
point of the phase change material. This
works goal is to determine the
usefulness of FLUENT when analyzing
phase change materials, in this case nFigure 3. Symmetry Cell of PCM Storage with Internal Fin [8]
octodecane. The numerical results are
compared with existing experimental
data first before using the program to analyze a phase change condition. Thus, some tweaking is necessary to utilize
the code for this application to a cylinder of n-octodecane wax heated from the top with a source of temperature
greater than its melting point. The temperature distributions found numerically agreed quite closely with the
experimental results previously found by Pinelli. However, the phase change boundary seemed to move more
quickly in the numerical model. This difference can be accounted for by the fact that some properties had to be
treated as temperature dependent, due to FLUENTs method of handling the phase change the enthalpy-porosity
model was used, with 0 being a solid material and 1 being a fully liquid material. The authors emphasize that the
methodology that FLUENT uses should be carefully worked through before using the code, so that the user is aware
of the processes. This then allows for changes and assumptions to be made that will help to arrive at a reasonable
answer once the code is run.
On the other hand, a relevant analytical model to the current study is the one created by Lamberg et al.[7] The
model in this paper is for a prediction of a solid-liquid interface location and temperature distribution of a fin in the
melting process of a PCM storage. The basic energy balance of such a model is shown in Figure 2. The value of the
model for the melting process is that it determines when the thickness of the phase change material is such that
natural convection takes over from conduction. The
analytical results and numerical results are in agreement
with the exception of the speed of the solid-liquid interface
during the phase change process. The main complications
in solving phase change material problems are that the
liquid and solid states have different properties; there is a
nonlinear moving liquid solid interface. Furthermore, the
melting front location is not known a priori. There are
three possibilities when there is heat transfer to the PCM
from a fin: conduction plus natural convection form the fin,
pure conduction from the fin, and pure natural convection
from the fin. The numerical solution is based upon the
Neumann solution and provides a prediction of the items
mentioned in the model. However, determining the
resistance of the fin in this case would require the
evaluation of Equation 1 at x = 0 (the fin base). This
would require some evaluation using LHospitals rule,
based upon the form of the equation. This specific
equation can be found in[7].
Siren and Lamberg worked to also create a model for
solidification so that a complete melting and freezing
process could be simulated for a given finned thermal
storage unit.[8] Two models were analyzed for the
Figure 4. Solar Collector and Storage
simulation of solidification of a PCM storage system with
4
American Institute of Aeronautics and Astronautics

internal fins, one dimensional and two dimensional. Most of the heat transfer in the case of solidification is done by
conduction, so that is the focus of the driving equations. The fin energy balance is also used to create the one
dimensional model. The cell is split into the region with the fin and the region with just the pipe for the analysis (as
shown in Figure 3, since the Neumann solution can be used for just the pipe. A dimensionless rate of change is
developed first for the fin, and then it is multiplied by the length to determine the solid-liquid interface location. A
finite-difference solution is developed also for the solidification process for the two dimensional model. From the
results, it can be seen that the cell geometry most affects the main solidification direction and the result of the
analytical models. When the length of the fin is more significant, the speed of the interface coming off the fin
determines the solidification, and vice versa (the speed is
influenced by the pipe heat transfer). In the case of the fin
being longer, the one dimensional model becomes more
accurate. A method for determining the solidified fraction of
3
PCM is also presented, which is useful in determining the time
of the unit to solidify. Determining a fin resistance from the
equations derived through this model, however, would be
difficult due to the form of the solution, as seen in [8].
Therefore, for the purposes of this study, we analyze a
2
5
simpler fin model, that of the circular fin. The approximation
made to do this calculation breaks up the plate fin into multiple
4
circular fins, each with their diameter measuring the width of
the plate fin. The analytical model corresponding to the
circular fin is contained in Incroperas heat transfer text.[9]
FLUENT will be used alongside the analytical model to
6
1
simulate the same portion of the storage unit. This FLUENT
model will be compared to the simpler resistance model given
by the circular fin. This process will be completed for melting
and solidification of the wax. This initial result is that of
solidification to show the extraction of heat from the wax. A
TC Location
4
simple finite-differencing analytical scheme has been
Heat load loop
developed for the analytical calculations, while a two
dimensional FLUENT simulation has been run for the
Heat exchange loop
numerical calculations. A photo of the physical rig whose
storage section is modeled is shown in Figure 4.
Figure 5. Top View of Storage Unit

II. Methodology
A. Storage Modeled
The storage unit modeled is a copper heat exchanger with
copper fins brazed on. This unit is placed in an aluminum box,
and melted wax is poured in over it so that the heat exchanger is
completely encased in the wax. There are two loops, one for the
heat exchanged to the wax from the heat transfer fluid, and one
for the heat extracted from the melted wax when the unit is fully
C
charged. The entire assembly of wax and heat exchanger is
A
enclosed in an aluminum box two feet on each side. This box is
surrounded by one foot of Styrofoam insulation on all sides to
reduce heat loss from the tank once it is fully charged. This
insulation is surrounded by a wooden structure, which houses the
storage unit as well as a pump for charging the unit. To charge
B
the unit, it is wheeled outside on a sunny day and the charging
Figure 6. Computational Domain
pump turned on. Thermocouples embedded within the wax
indicate the time at which the wax begins to melt. Once all the
wax is melted, the discharging process begins. The charging pump is turned off after the unit is rolled back inside,
and the discharge pump is turned on. The top view of the heat load and heat transfer tube is shown in Figure 5, as
well as the location of the thermocouples embedded within the wax to measure the wax temperature during melting
and freezing tests. The fins are also shown as horizontal lines with dimensions of 1/16 thick, 23.75 long and 7.75
5
American Institute of Aeronautics and Astronautics

wide. A glycol / water mix will be used as the heat transfer fluid for charging the tank, while water will be used as
the heat transfer fluid for discharging the tank. Ideally, the water flowing out of the unit would be used in a home,
but in this case a small radiator is used in this case to discharge the heat. Testing was completed between the
months of November and June. The results from the analytical model will be compared to a sample test date to
examine their proximity to the results in terms of the resistance.
B. Numerical Model
The FLUENT model is developed using the two-dimensional double precision mode. It is important to note that
the
FLUENT
model
takes
into
consideration the mushy zone of solidifying
Fin
h, Tinf
wax, while the analytical model does not
make this distinction.
The first
consideration given to the numerical model
was appropriate assumptions to create the
model to be run in FLUENT. From the
storage unit, the first fin and tube section is
modeled in two dimensions. There are
several assumptions made here to analyze
Wax
only two dimensions. First, the effect of
gravity is neglected, so there is no natural
convection modeled here. Furthermore, the
heating up of the fluid as it moves past the
melted wax is not modeled, and it is
assumed that the flow to all tubes in the
heat transfer array is split evenly among
each tube. The area modeled is shown
enclosed in red in Figure 6. The boundary
Figure 7. FLUENT representation of Computational Domain
condition was set as adiabatic for faces A
and D, symmetric for face B, and
convection with a heat transfer coefficient corresponding to the Dittus-Boelter Correlation for water flowing at 298K
on face C. The interface between the wax and copper became a shadow wall coupled with the copper fin. This
enabled conduction from the fluid, to the pipe wall, to the fin and ultimately to the wax.
A schematic of the portion meshed can be
seen in Figure 7. This shows the small
portion on side C that is exposed to the
convection condition to simulate the water
flow through the pipe. A resistance is also
added to this small portion to account for the
fact that there would be a resistance value
from the pipe wall in the actual storage unit.
Before the numerical model was run for
the complete case desired, a simplified model
was first created to test that the solidification
process made sense.
This model was
composed of only the wax within a three inch
by three inch square and was exposed to a
negative heat flux on both walls. The results
of this model were compared to the analytical
Stefan problem for solidification and it was
found that it was accurate. This simple model
Figure 8. Control Volume Types
was then used as the base for the more
complex model.
The variable of interest in this case was the resistance across the fin, so an adjustment to the model was made
after an initial run. The equivalent resistance from the length of the pipe wall modeled was added to the fin base,
and the length of pipe wall without the fin was set to adiabatic. Thus, there was only convection at the fin base.
This was done to determine the resistance across the fin by dividing the heat transfer across the base and to the wax
6
American Institute of Aeronautics and Astronautics

Rwax

Rpipe

Rconv
Type A

Rfin

Rpipe

Rconv

by the temperature difference between the melting temperature of wax and the
fluid temperature in the free steam within the pipe.
When creating a base for the model in GAMBIT, the mesh was first
constructed on the half-fin section. Once this edge was established, a boundary
layer mesh could be implemented. GAMBITs built in function for the boundary
mesh was used to establish a finer mesh around the four sides of the wax. This
ensured that a finely structured mesh was in place along the crucial edge portions
of the wax next to the copper heat exchanger. Finally, a uniform mesh was set on
the remainder of the wax portion.
When running the model, the solidification / melting model was selected,
which automatically enabled the energy equation calculations. The appropriate
properties were entered for the materials involved in the calculation, which in this
case were the paraffin wax and copper. Since the solidification / melting model
was used, the latent heat of fusion, solidus temperature, and liquidus temperature
were entered for the paraffin wax in addition to its other liquid properties. The
boundary conditions were set as indicated in the computational domain, and the
model was initialized according to the case studied. For the melting case, the wax
was initialized at 300 K, while in the freezing case it was initialized at the melting
point given by the manufacturer, 323 K. An unsteady calculation was run
because the melting front movement had to be observed for various times. As
such, the model was run for one hour to five hours, with data being saved at each
hour mark. To check the model, it was run for a one hour then a two hour
calculation, and checked to ensure that the freezing front had advanced. Once
these models were run, the flux across the portion of the fin exposed to
convection was calculated using FLUENTs Report feature. This gave the flux in
watts across the desired surface. This flux value is then used to figure out the
resistance over the wall, which will be compared to the analytical resistance value
found.

C. Analytical Model
The analytical model was developed using the one dimensional resistance
model as described in Incropera et. als heat transfer text[9] and guided by the
principles presented in Patankars text.[10] It also utilizes the control volume
approach, which is widely used in the heat transfer and fluid mechanics fields.
The process presented here will be validated using the work of Lamberg and Siren, as well as the experimental data.
Through this validation one can identify areas of improvement for the model and areas for further study within the
storage unit.
There are two different types of control volumes analyzed, so the analytical model will be explained in terms of
these two volumes. Control volumes A and B are specified in Figure 8 in terms of their location within the storage
unit. The two control volumes share three common elements: the convection condition within the pipe, the pipe
wall, and the freezing thickness of the wax. These three elements have corresponding resistance values, which are
calculated based on the cylindrical heat transfer resistance model found in Incroperas text.[9] The difference
between these two control volumes, however, is the fact that control volume B is up against the fin, whereas control
volume A is not. The convection resistance was determined by the following equation:

Type B
Figure 9. Resistance
Chain Types

Rconv

1
hA

(1)

The conduction through the wall is determined using the cylindrical conduction resistance, as in equation 2. This
is used to simulate the conduction through the circular medium radially from the outside of the pipe to the inside of
the pipe during the freezing process and from the inside of the pipe to the outside of the pipe during the melting
process.

7
American Institute of Aeronautics and Astronautics

Rcond ,wall

r2, pipe

ln
r

1, pipe

2k pipe L

(2)

The convection resistance and the conduction resistance through the pipe wall were assumed constant throughout
the models run, however, the frozen wax layers resistance changed with time. The resistance of the wax changes
during the melting process because of the melted wax layer, and changes because there are two possibilities for the
heat transfer coefficient around the fin. Either way, the solidification or melting distance of the wax was calculated
using the Stefan equation:

xe x erf ( x)
In this case, x represents:

C p (Twax Tmean )
H

2 t

(3)

(4)

The expression in (3) was solved using a simple transcendental solver for x, then (4) was solved for to determine
the wax thickness as it froze or melted. Next, the wax thickness was placed in the cylindrical resistance formula to
determine the resistance resulting from the increasing wax thickness at each time step.
The first control volume, A, has a total resistance determined by the following equation. This resistance is
shown in schematic in Figure 9.

Rtotal

r2, pipe
r

ln freeze
ln
r

1
1, pipe
2, pipe

hmean A 2Lk pipe

2Lk wax

(5)

However, for control volume type B, the fact that there is heat transfer through the sides of the fin must be taken
into consideration. This is done by developing an expression for the equivalent resistance for the fin. In the case of
melting, the heat transfer coefficient will be determined by natural convection methods. For the freezing process,
we replace the heat transfer coefficient with the thermal conductivity of the solid wax divided by the distance the
wax has frozen at that point. This represents an equivalent heat transfer coefficient in the solidification case. The
thermal resistance for a circular fin is given by the following:

R fin A fin

k fin

2r

1, fin
2
2
2 , fin
1, fin

m( r

K1 (mr1, fin ) I1 (mr2, fin ) I1 (mr1, fin ) K1 (mr2, fin )

) K 0 (mr1, fin ) I1 (mr2, fin ) I 0 (mr1, fin ) K1 (mr2, fin )

(6)

with the term m defined as in (7) for freezing, (8) for melting without natural convection, and (9) for melting with
natural convection.

2
m

freeze

k fin t fin

2
m

k fin
(7)

k fin

melt

k fin t fin
hwax Pfin
k fin Ac

So, the total resistance with the fin included is as shown below, and the resistance chain can be seen in Figure 9.

8
American Institute of Aeronautics and Astronautics

(8)

(9)

Rtotal

r2, pipe

ln
r

k fin
2r1, fin
K1 (mr1, fin ) I1 (mr2, fin ) I1 (mr1, fin ) K1 (mr2, fin )
1
1, pipe

A fin
2
2
hmean A 2Lk pipe
freeze m(r2, fin r1, fin ) K 0 (mr1, fin ) I1 (mr2, fin ) I 0 (mr1, fin ) K1 (mr2, fin )
(8)
Having defined the resistances to be used
in the model, the analytical code could be run
for the length of the pipe and for the time
desired. First, the storage unit was initialized
for all lengths during the first time, and for
the first portion of the tube length for all
times run. The fluid always entered the heat
load section at 298 K (27 C), since it was
assumed that the water coming out from the
heat load will be used up. When the
analytical model was run in comparison to the
experimental results, the temperature into the
model was the same condition in the
experimental testing so that the comparison
would be as close as possible. In the melting
case, the fluid enters the heat exchange loop
at 338 K, which is the same temperature used
in the FLUENT calculations. Furthermore, it
was assumed that for the very first section of
the pipe at the first second of heat extraction
from the melted wax, the wax did not freeze.
The same assumption is made for the first
section of pipe at the first second of heat
transfer to the solid wax.
Once the
initialization was complete, the main code
was run. First, the mean fluid temperature in
the section of interest for calculation was set
to be the exit fluid temperature from the
previous section.
From the resistance
determined from control volume A or B, the
heat transfer from the wax to the fluid was
determined using:

qwall

1
Rtotal

(Twax Tmean )

(9)
Next, the temperature out of the small
section was determined using the fact that the
heat transferred from the wax to the water
could be set equal to the heat gained by the
fluid from the wax as it passed through that
small section of pipe. The following equation
was then solved for the temperature exiting
the small section of pipe.

mdotC p (Tmean ,in Tmean ,out )

1
Rtotal

(Twax Tmean )
(10)

Figure 10: FLUENT Freeze Fraction Contours, Hours 1-4

9
American Institute of Aeronautics and Astronautics

Using this equation, the temperature coming

out from the small section of pipe was
determined. This temperature was then passed
on to the next section of pipe, and the process
repeated over the entire length of the pipe. The
entire heat exchanger was modeled so that the
temperature at the exit of the heat load or heat
transfer loop could be determined, but for the
sake of this work, the first portion of the pipe is
taken for analysis. It is also important to note
that although the analytical model does take
into account the fact that there is solidification
occurring on both sides of the fin, it does not
take into account the fact that the solidification
front will take on a curve shape as it nears the
fin. This is taken into account in the FLUENT
model, obviously, so this will be a point of
discrepancy between the two models. The
effect of this discrepancy will be addressed
when discussing the results.
Finally, the resistance over the entire section
is determined by dividing the log mean
temperature difference by the total heat transfer
from the wax to the fluid, with both values
taken over the entire section.
This total
resistance will be determined at one hour
through five hours in one hour increments and
will be compared to the result found by the
FLUENT simulation.

III. Results and Discussion

Figure 11: FLUENT Melt Fraction Contours, Hours 1-4

A. Numerical Model
The FLUENT simulation was run with the
boundary conditions and initial conditions
described in the Methodology section. To
provide intial verification, a two-dimensional
model was developed. This included one half
of the wax section between fins, a half-width of
the fin, and the equivalent wall of the pipe with
the convective heat transfer from the liquid
going through the pipe. FLUENT models were
run for one through four of heat extraction and
heat addition. Also, the contours of liquid
fraction were noted, as these indicated the
mushy zone of the wax during the freezing
process. The contours of liquid fraction during
the freezing process are shown in sequence in
Figure 10. The process starts at the one hour
mark at the top plot in the set, then is shown for
the remaining hours. The melting front and
freezing front move as expected over time for
both cases.
The other main result from the FLUENT
simulation is the heat transfer over change in
temperature value at the fin, which corresponds

10
American Institute of Aeronautics and Astronautics

to its resistance. This value is computed by having FLUENT compute the heat transfer through the fin, then
dividing it by the temperature difference (in this case, it is between the melting point of the wax, 323 K, and the
temperature of the fluid through the pipe, 298 K). This value is shown in Figure 13 along with the values found
from the freezing analytical model, and will be discussed in a following section.
Figure 11 shows the corresponding melt process for one through four hours. It also makes sense in that the
melting front moves away from the pipe area. Also, it occurs slower than the solidification process, which is
expected as a lower mass flow rate is used for the melting process than the freezing process. The comparison of the
analytical and numerical models resistance values will be discussed in a following section.
B. Analytical Model
The analytical model is first validated based upon the experimental results from the rig and the results achieved
by Lamberg and Siren. Since their freezing study used a one dimensional model of a similar process, this is a good
comparison. The melting model is also compared to this literature to ensure a complete consideration has been
made of both processes.
First, the freezing model is considered. The one-dimensional freezing analytical model from Lamberg and Siren
represents a symmetrical cell with an internal fin, and the analytical model developed for the current study was used
with the same conditions. The wax properties and fin material properties and the dimensions of the enclosure used
by Lamberg and Siren were placed in the analytical model. These results are compared in Figure 12. The first
evident result is that the developed analytical model grossly overpredicts the distance of the freezing front from the
wall and the fin when compared to the results of Lamberg and Siren.
0.25

Solid-Liquid Interface Distance (m)

0.2

0.15
Siren and Lamberg (Case 1)
Calculated (Case 1)
Siren and Lamberg (Case 2)
Calculated Values (Case 2)
0.1

0.05

0
0

0.01

0.02

0.03

0.04

0.05

0.06

Fin Length (m)

Figure 12: Validation of Freezing Model

Next, the melting model is considered. Lamberg and Siren used their derived analytical model to determine the
location of the solid-liquid interface as compared to the wall and the fin within the semi-infinite region considered.
The current analytical model compares more favorably to Lamberg and Sirens model. The values of the melting
front distance from the endwall are more in agreement, however the same trend is not seen. This can be explained
by the fact that in the one-dimensional model, the fin is collapsed as a resistance in the model, not as a separate
entity.
11
American Institute of Aeronautics and Astronautics

The results of the analytical model compared to the actual experimental data are compared in Figure 14. The
melting model compares favorably to the experimental melting data. However, the freezing model does not
compare as favorably to the actual data. This could be due to the fact that the natural convection is not considered in
the freezing case. Furthermore, the curvature of the freezing and the melting fronts was not considered in either
analytical model. The fact that the model overpredicts the freezing temperatures in the experimental data is in
agreement with the validation.

60

50

Fin Temperature (C)

40

Analytical Model
Siren and Lamberg

30

20

10

0
0

0.02

0.04

0.06

0.08

0.1

0.12

Endwall Distance (m)

Figure 13: Validation of Melting Analytical Model

C. Comparison of Results
When first examining the trends, a check of the physical processes as time progresses should first be made. The
amount of resistance in degrees Kelvin per Watts increases over time during the freezing process in the FLUENT
model and analytical model. This indicates that more Watts of heat are require to push one Kelvin of temperature
from the melted wax, through the frozen wax, through the fin, and ultimately to the fluid in the pipe in this case.
This makes sense the increasing thickness of solid wax provides a larger and larger resistance over time in the
freezing process. This is accounted for in the analytical and FLUENT model also, the trend of both curves is
almost the same, which indicates that there some correspondence between the models. The melting process
resistance as time goes by is a more complicated consideration as there is not only conduction heat transfer to melt
the wax, but also convection heat transfer once the melted wax reaches a critical radius. As time goes by, the
convection resistance decreases because the surface area over which the convection occurs increases, as does the
heat transfer coefficient as natural convection begins to take hold. The conduction resistance of the melted wax is
active in the melting model in terms of the fact that it is another medium through which heat must pass before
reaching the solid wax, but it is negligible when compared to the convection resistance.
The analytical model does not exactly compare to the ones developed by Lamberg and Siren, which indicates
that there must be some error in the calculation, which corresponds with the simplifications made in the present
analytical model. The present model relies upon the heat transfer resistance model to transfer heat within the fin
during the melting and solidification process. From the melting model results we see that this works, supported by
12
American Institute of Aeronautics and Astronautics

the fact that the fin resistance is used and therefore takes into account the convection on both sides of the fin. Also,
the natural convection beyond a critical thickness is accounted for, which helps to accelerate the melting of the wax
by gradually stripping away mushy wax at the solid-liquid interface. The analytical model also does not take into
account the dimensions of the site as closely as the model in literature; the literature model nondimensionalizes fin
parameters based upon the sizing of the enclosure. The temperatures are also nondimensionalized based upon the
wall temperature and the melting or freezing temperature of the material. This is done to solve the energy equation.
The solution is more sophisticated than that of the current analytical model. Although the idea of having a
simplified model is attractive, there must be an appropriate balance between time the model requires to run and
accuracy.
70

60

Temperature (C)

50

40
Experimental Melting
Temperatures
Analytical Melting
Temperatures

30

Experimental Freezing
Temperatures
Analytical Freezing
Temperatures

20

10

0
0

0.5

1.5

2.5

Time (h)

Figure 14: Comparison of Analytical Model to Experimental Results

In analyzing the trends in the resistance values for the fins for both cases, several important results are found.
The first result is that there is indeed a discrepancy between the analytical and numerical models for resistance
values. This can be attributed to the fact that the analytical model does not take into account the curvature of the
solidifying front as it approaches the fin. Furthermore, as pointed out by Pinelli, the solidification or melting front
calculated with FLUENT moved faster than the front in actual experiments. This could be another factor in the
difference between the resistance values for the analytical and numerical cases. Finally, it is also recognized that the
computational method of FLUENT takes into account the mushy zone that occurs where the liquid wax is
somewhere between the melting and solidification areas. However, this is not accounted for in the analytical
calculation, so the analytical model does not take into account the fact that the wax gradually loses energy.

IV. Conclusions
Based upon the numerical study done thus far, the analytical model resistance values fall within the same order
of magnitude as the resistances predicted by FLUENT. However, the resistances achieved by the analytical model
are approximately five times that achieved by the FLUENT model for the melting and freezing cases. When the
analytical model is verified using experimental data and published models, the melting model is most accurate in
both comparisons, while the freezing model overpredicts the amount of frozen wax and the temperature exiting the
13
American Institute of Aeronautics and Astronautics

system. These results are useful because they show the complexities of the melting and solidification process, as
well as the fact that the melt and freezing curvature along the fins must be considered, an energy equation should be
incorporated into the simulation, whether it is within the analytical model or within the FLUENT calculations.

Acknowledgments
Monica Bonadies acknowledges financial support from the University of Central Florida in the form of a
Provosts Fellowship.
14.000

12.000

Resistance (K/W)

10.000

8.000

FLUENT Freezing Resistance

Analytical Freezing Resistance
FLUENT Melting Resistance
Analytical Melting Resistance

6.000

4.000

2.000

0.000
0

2000

4000

6000

8000

10000

12000

14000

16000

Time (s)

Figure 15: Comparison of Analytical Model to FLUENT Results

All authors acknowledge the donation of solar thermal tubes from GEAR Solar of Greenville, South Carolina as
well as the donation of circulating pumps from Grundfos. We would also like to thank Dr. Arup Guha for his
management of the cluster on which the FLUENT cases were run.

References
[1] Johnson, G., Plugging Into the Sun, National Geographic, September 2009, pp. 28-51.
[2] http://www.energysavers.gov/your_home/water_heating/index.cfm/mytopic=13050, accessed November 16,
2009.
[3] Zalba, B., Marin, J.M., Cabeza, L.F., Mehling, H., 2003, Review on thermal energy storage with phase change:
materials, heat transfer analysis and applications, Applied Thermal Engineering, 23, pp.251-283.
[4] Sharma, A., Tyagi, V.V., Chen, C.R., Buddhi, D., Review on thermal energy storage with phase change
materials and applications, Renewable and Sustainable Energy Reviews, 2009, 13, 318-345.

14
American Institute of Aeronautics and Astronautics

[5] Reddy, K.S., 2007, Thermal Modeling of PCM-Based Solar Integrated Collector Storage Water Heating
System, J. Solar Energy Engineering, 129, pp. 458-464.
[6] Pinelli, M., Piva, S., 2003, Solid/Liquid Phase Change in Persence of Natural Convection: A Thermal Energy
Storage Case Study, J. Energy Resources Technology, 125, pp. 190-198.
[7] Lamberg, P., Siren, K., 2003, Analytical model for melting in a semi-infinite PCM storage with an internal fin,
Heat and Mass Transfer, 39, pp. 167-176.
[8] Lamberg, P., Siren, K., 2003, Approximate analytical model for solidification in a finite PCM storage with
internal fins, Applied Mathematical Modeling, 27, pp. 491-513.
[9] Incropera, Fundamentals of Heat and Mass Transfer, 6th ed., Wiley and Sons, New York, 2007, Chaps. 3,5.
[10]

Patankar, Numerical Heat Transfer and Fluid Flow, Hemisphere Publishing, New York, 1980.

15
American Institute of Aeronautics and Astronautics