You are on page 1of 24

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/315643049

A Study of the Onset of Natural Convection During Melting of PCMs in a


Cylindrical Enclosure

Conference Paper · May 2017


DOI: 10.1615/ICHMT.2017.1240

CITATIONS READS
5 632

3 authors, including:

Mohammad Azad Dominic Groulx


Dalhousie University Dalhousie University
22 PUBLICATIONS   122 CITATIONS    166 PUBLICATIONS   2,618 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Smart sprayer View project

Spot-specific ferilizatio using Image processing View project

All content following this page was uploaded by Dominic Groulx on 25 March 2017.

The user has requested enhancement of the downloaded file.


Proceedings of CHT-17
ICHMT International Symposium on Advances in Computational Heat Transfer

May 28-June 1, 2017, Napoli, Italy

CHT-17-067

A STUDY OF THE ONSET OF NATURAL CONVECTION DURING MELTING OF PCMS


IN A CYLINDRICAL ENCLOSURE

Mohammad Azad*, Dominic Groulx*,§ and Adam Donaldson**


*Department of Mechanical Engineering, Dalhousie University, Halifax, NS, Canada
**Department of Process Engg. and Applied Science, Dalhousie University, Halifax, NS, Canada
§Correspondence author. Fax: +1 (902) 423 6711 Email: dominic.groulx@dal.ca

ABSTRACT The onset of natural convection was studied numerically during melting of
n-octadecane and dodecanoic acid within the annular space of a horizontal cylindrical enclosure.
COMSOL Multiphysics was used to model the two-dimensional heat transfer process in the system;
heating the PCM from the inner surface of the annulus with the outer surface insulated. The inner
surface of the enclosure was heated with temperatures corresponding to Stefan numbers of 0.2, 0.3
and 0.4, which caused the PCMs to melt. Melting of the PCMs was simulated in enclosures with inner
cylinders having diameters of 6.35 and 25.4 mm, while keeping the outer cylinder diameter constant
at 127 mm. For both the PCMs studied, natural convective heat transfer rapidly exceeded initial pure
conduction, with onset occurring faster at high Stefan numbers regardless of the diameter of the
heating surface. Also, the onset of natural convection was slightly slower in dodecanoic acid
compared to n-octadecane.

NOMENCLATURE

length [m /m]
A Constant in temperature dependent V(t) Volume of liquid PCM per unit axial
thermophysical properties [-]
B Constant in temperature dependent Vcr Volume of molten PCM per unit axial
thermophysical properties [-, K, 1/K] length at convection onset [cm3/m]
C Constant in temperature dependent vcr Velocity of molten PCM at convection
thermophysical properties [K, 1/K ] onset [mm/s]

( ) Temperature dependent coefficient of


Cp(T) Modified specific heat capacity [J/kg.K] Greek symbols
d Diameter of the inner cylinder of the

( ) Melt fraction [-]


enclosure [mm] thermal expansion [1/K]

thermophysical properties [-, 1/K ] ( ) Temperature


D Constant in temperature dependent
dependent viscosity

∆ℎ
D(T) Gaussian function [-] [kg/m∙s]


dcr Diameter of molten PCM at convection Latent heat of fusion [J/kg]
onset [mm] Mushy zone temperature range [K]
E Constant in temperature dependent Subscripts
thermophysical properties [1/K ] B Buoyancy
ET Total stored energy [kJ] m Melting point

1
INTRODUCTION

The most beneficial aspect of using phase change materials (PCMs) in thermal energy storage (TES)
systems is the high values of latent heat of fusion of the PCMs, which yield up to 14 times more
energy storage compared to the sensible storage systems [Murray et al. 2011]. However, the inherent
disadvantage of using PCMs is their low thermal conductivities. Most PCM-based TES systems have
low heat transfer rates in and out of the systems, i.e., the rate problem is the limiting factor for PCM-
based TES systems. Previous attempts to increase the heat transfer rates and thermal conductivities
of the PCMs have included addition of metallic matrix and fins to the heat exchangers, addition of
conductive nanoparticles, and enhancement of convective heat transfer through geometric
modifications.

Varol and Okcu [2013] compared numerically the melting process of paraffin wax in vertical
rectangular enclosures with five fins attached equidistance on the heating side and without fins.
Significant reduction in melting time of the paraffin wax in the finned configuration was observed in
their study. Fan et al. [2013] studied experimentally the effects of melting temperature and of the
presence of fins using two PCMs of similar thermophysical properties but different melting
temperatures in an electronic heat sink. A cubicle enclosure was used without fins and with fins -
placing two cross plates at the center of the enclosure. Performance of the heat sink improved in the
finned case regardless of the PCM used.

Dhaidan et al. [2013] studied both numerically and experimentally the melting of CuO-enhanced
n-octadecane in a square enclosure heated from a side with the opposite side insulated. Addition of
nanoparticles increased the thermal conductivity of the PCM and thus increased the heat transfer rate,
indicated by the mean temperature and melt fraction results. Sciacovelli et al. [2013] examined
numerically the melting of nano-enhanced PCM in a shell-and-tube vertical heat exchanger. Addition
of 4% nanoparticle by volume reduced the melting time by 15%. Ho and Gao [2013] studied
experimentally the melting of n-octadecane, enhanced with alumina particles in a vertical square
enclosure. One vertical side of the enclosure was heated isothermally and the opposite side was
thermally insulated. The fraction of nanoparticles used were 0, 5 and 10% by mass and their results
revealed that natural convection heat transfer in the melted area of the PCM degraded with increase
in the mass of nanoparticles. This decrease in natural convection heat transfer may be attributed to
the fact that dynamic viscosity of the nano-enhanced PCM increases by 20 and 28% for addition of
alumina particles at 5 and 10 wt.%, respectively. Arasu and Mujumdar [2012] also studied melting
of alumina enhanced paraffin wax in a square enclosure. Two cases were considered; first, heating
the PCM from one vertical side of the enclosure and second, heating the PCM from the bottom of the
enclosure. In both cases, increase in volumetric composition of alumina reduced the heat transfer rate.
Enhancement of PCMs with nanoparticles increases the conductivity of the PCMs and thus may result
in faster melting and solidification. In some cases, it degrades the impact of natural convection heat
transfer. Addition of too much nanoparticles also increases the viscosity and could cause stratification
of the PCM. Enhanced heat transfer is experienced in packed bed PCMs too, especially using small
capsules as reported by Regin et al. [2008] but it could result in significant pressure drop.

It is seen that most of the methods of enhancing heat transfer rate during melting of PCMs has their
own disadvantages. Moreover, any improvement in heat transfer rates by addition of nanoparticles to
the PCMs or addition of fins, or using metallic matrix to the heat exchangers comes with reduction
in storage volume in addition to the initial cost involved in modifying the heat exchanger or enhancing
the PCMs. Fast onset and presence of strong natural convection heat transfer increases the overall
heat transfer rate without any reduction in the storage volume and any extra cost involved in
modifying the heat exchangers or in enhancing the PCMs.

2
Darzi et al. [2012] conducted a numerical study on melting of n-eicosane in horizontal cylindrical
enclosure with two configurations – concentric annular enclosure and eccentric annular enclosure.
The PCM filled the annular space and heat was applied at the inner surface. Initially heat transfer was
due to conduction in both cases. The melt fraction and the melt profiles were identical for the first 15
minutes. Once natural convection in the melted region started playing a role, the heated fluid moved
upward causing more PCM to melt in the upper region of the concentric cylinder and leaving the
lower portion almost unaffected. However, in the eccentric cylinder arrangement, the volume of PCM
in the lower portion was significantly less than that at the upper potion and thus the melting was
significantly more in this case. Improved melting of PCM in a vertical annular enclosure was also
obtained by tilting the outer surface by a small angle of 5°, reported by Akgün et al. [2007]. Shmueli
et al. [2010] also found in their study of PCM melting in annular vertical enclosure that heat transfer
takes place by conduction in the beginning and convection heat transfer dominating shortly thereafter.
Natural convection driven melting has also been observed by Tan et al. [2009] and Duan et al. [2010]
in spherical containers heated from the surface, by Avci and Yazici [2013] and Tao and He [2015] in
horizontal cylindrical containers heated from the center, by Liu and Groulx [2014] in horizontal
cylindrical containers with longitudinal fins, by Murray and Groulx [2014] in vertical cylindrical
enclosures with radial fins and by Longeon et al. [2013] in vertical annular cylindrical enclosure.

As mentioned above, during the melting phase of PCMs, heat transfer takes place by conduction only
in the very beginning. Once enough PCM has melted, buoyancy driven natural convection starts to
play a role in the overall heat transfer and resulting melting; and soon becomes the dominant heat
transfer mode. During natural convection, hot fluid close to the heating surface rises upward in the
enclosure and cold fluid in the proximity of the solid phase moves downward. This causes more
PCM to melt in the upper portion of the enclosure. However, until most of the PCM in the upper
portion of the enclosure melts, convection heat transfer remains dominant over the conduction mode.
The importance of convection heat transfer during melting of PCMs warrants the answers to the
questions: What are the effects of heating surface diameter on onset of natural convection? What are
the effects of heating temperature on natural convection onset? Is the natural convection onset
dependent on the storage medium (i.e., PCM)? Is the natural convection onset dependent on the
geometry and orientation of the storage enclosure?

This paper presents the results of numerical investigation of the onset of natural convection during
melting of n-octadecane and dodecanoic acid within a horizontal annular cylinder heated from centre.
The diameter of the inner cylinder was set at 6.35 and 25.4 mm, with heated wall-temperatures set to
correspond to Stefan numbers of 0.2, 0.3 and 0.4. The instantaneous melt profiles, instantaneous
maximum velocities of molten PCMs, surface heat flux, melted volume and stored thermal energy were
compared. Melting was modelled accounting for convection heat transfer and appropriate results were
compared to those obtained considering conduction only heat transfer mode.

MATERIAL PROPERTIES

The thermophysical properties of n-octadecane and dodecanoic acid at solid state, which are assumed
constant, are given in Table 1. The thermophysical properties in liquid phase vary with temperature, and
therefore, temperature dependent properties were adopted from the Handbook of Thermodynamic and
Physical Properties of Chemical Compounds by Yaws [2003] and are defined in Eqs. (1) – (6). It is
emphasized here that all the temperature dependent property equations are valid at least up to 595 K,
which is significantly higher than the maximum temperature appearing in this study.

( )= + + [W/m ∙ K] (1)
$ '
( )= ×1000 [kg/m ]
!"#!$ &
%

( ) = 10 ×10 [Pa ∙ s]
(2)
.
",- -/0-10 2 & !
$ (3)

3
Table 1
Thermophysical properties of n-octadecane and dodecanoic acid

0.358 W/m ∙ K 0.160 W/m ∙ K


Properties n-octadecane Dodecanoic acid

814 kg/m 940 kg/m


: 189 kJ/kg 187.2 kJ/kg
27.5 °C 43.5 °C

A ( )= + + +B +C [kJ/kg ∙ K]
= + + [kJ/kg ∙ K]
(4)
A
0 E
(5)
( ) = × "1 − & [1/K]
0 %
(6)

The constants appearing in Eqs. (1) – (6) are given for n-octadecane and dodecanoic acid in Table 2 and
Table 3, respectively.

Table 2
Constants appearing in thermophysical property equations of n-octadecane

Properties Constants
C or F
0.211 2.397×10! 6.667×10!H − −
A B D E

0.234 0.251 747 0.274 −


−8.551 1670 0.0157 −1.234 −
A 4.846 −0.022 5.400×10!J −3.491×10!H −3.070×10!#
A 0.137 0.006 −6.896×10!K − −
4.986×10 !
−0.726 − − −

Table 3
Constants appearing in thermophysical property equations of dodecanoic acid

Properties Constants
C or F
0.163 1.922×10!J −2.107×10!K − −
A B D E

0.284 0.266 734 0.293 −


−8.602 1871.3 0.015 −1.204×10 !J −
A 0.254 0.011 −2.479×10!J 2.185×10!H −
A 1.011 −0.006 3.215×10!J − −
5.299×10 !
−0.707 − − −

4
The thermophysical properties of liquid PCMs at melting temperatures are given in Table 4. It should
be noted that in liquid phase, both the PCMs have similar properties except for the density and dynamic
viscosity. Specific heat in liquid phase is also a bit different but its significance mostly lies in sensible
heat transfer, which is low in the present study where both the PCMs are initially at the melting
temperature. Moreover, Table 1 shows that both the PCMs have similar latent heat of fusion. It is
noteworthy that the distinguishing thermophysical properties of the two PCMs are the melting
temperatures, dynamic viscosities and the densities. Any dependence of onset of natural convection or
subsequent influence of natural convection on melting could be correlated to these properties.

Table 4
The thermophysical properties of liquid PCMs at melting temperatures

0.145 W/m ∙ K 0.148 W/m ∙ K


Properties n-octadecane Dodecanoic acid

777.11 kg/m 872.39 kg/m


A 2239.54 J/kg ∙ K 2030.94 J/kg ∙ K
0.004 Pa ∙ s 0.008 Pa ∙ s
0.0073 1/K 0.0079 1/K

NUMERICAL MODEL

Melting of n-octadecane and dodecanoic acid in a horizontal cylindrical enclosure was simulated using
COMSOL Multiphysics 5.2a. The enclosure, shown in Fig. 1, had an outer diameter of 127 mm and
constant temperatures were applied on a concentric surface with inner diameter of 6.35, 12.7 and
25.4 mm. The PCMs, initially solid at its melting temperature, filled the annular space and the outer
surface of the enclosure was insulated. Two-dimensional melting process was simulated with thermal
characteristics unchanged in the longitudinal direction. Due to symmetry about the vertical central axis
only half of the geometry was simulated.

Figure 1. The computational domain.

5
The melting process was simulated using heat transfer in fluid and laminar flow physics of COMSOL.
A modified heat capacity-porosity method [Kheirabadi and Groulx 2015] was employed to simulate the
heat transfer process. In this method, the PCM is considered liquid regardless of its temperature and the
specific heat capacity of the PCM is modified to account for the latent heat of fusion. The contribution

Eq. (7). The value of this Gaussian function is non-zero in the temperature range of − ∆ /2 and +
of latent heat to the specific heat capacity of the PCM is modelled using a Gaussian function, defined in

∆ /2 and zero elsewhere; and the integral over the entire temperature range is 1.

exp O− R
(0!0P )2
∆$P 2
B( ) =
"
Q
&
U
ST "∆0P &
(7)

The solid and liquid phases of the PCM are distinguished by using Eq. (8),

0, < −
∆0P
Y
W0!(0P !∆0P /
( )= , − < < +
) ∆0P ∆0P

X ∆0P
W
(8)
1, > +
∆0P
V

where, ( ) = 0 and ( ) = 1 represent complete solid and liquid phases, respectively; any state in-
between is referred to as the mushy zone. The Gaussian function and the melt fraction function are used
to define the modified heat capacity as follows:

A( )= A +] A − A ^× ( ) + ∆ℎ ×B( ) (9)

Thermal conductivity and density are also modified similarly, excluding the latent heat term that appears
in Eq. (9).

The Boussinesq approximation, as defined in Eq. (10), was employed to account for the natural
convection movement in the melted portion of the PCMs. Since both the solid and liquid phases of the

the phases. As the melting process takes place over a temperature range of ∆ , the real viscosity of the
PCM exist in the same domain, the real viscosity of the liquid PCM is modified to distinguish between

+ ∆ /2 and above. For temperatures


− ∆ /2, the viscosity is kept tremendously high (in an order of 10_ ) to mimic the
liquid PCM (defined in Eq. (3)) is assumed at temperatures of
that are less than
solid phase.

aaaab
` E = cb ( − ) (10)

The detailed description of the simulation method, including the treatment of the viscosity using the
mushy-zone contanst through porosity method, can be found in Kheirabadi and Groulx [2015].

Meshing Free triangular elements were used in the present study to create a domain mesh, as shown in
Fig. 2. Boundary layer elements were also added along the two walls of the PCM domain, as seen in the
zoomed-in figure. The maximum element size in the domain was 0.5 mm and the minimum size was
0.05 mm. The maximum and minimum element sizes in the boundary layers were 0.05 mm and
0.01 mm, respectively. The total number of elements in the domain was 156188 for core diameter of
6.35 mm. This meshing was chosen because the results obtained using this mesh follow the experimental
results very closely, which is discussed in the following subsection.

6
Figure 2. Meshing used in the simulation.

Numerical Method A total of 24 simulations (12 for conduction only, 12 accounting for convection)
were performed for the two PCMs over two of the inner diameters (6.35 and 25.4 mm) . Since the study
focuses on the onset of natural convection, the simulations were carried out for 30 minutes only, which
is above the time requirement for natural convection onset in any of the cases studied. The simulations
were performed on an Intel Xeon quad core processor (3.07 GHz, 12 GB RAM). For the selected mesh
size, the simulations took up to 35 times longer to run than the simulated time (30 minutes); with
simulation times increasing at higher Stefan numbers and at larger diameters of the inner cylinder.

Validation of the model The numerical model was validated against experimental results, for which, a
cylindrical cavity of 127 mm diameter and 50.8 mm length was made in a 152.4 × 76.2 mm plexiglass
block. The inside tube of 6.35 mm outer-diameter was made of stainless-steel. The system was equipped
with heat transfer plates on the axial wall surfaces, which were maintained at the melting point
temperature of the PCM to prevent sub-cooling of the solid PCM in adhesion to the axial surface.
Heating through the central tube was provided via an external recirculating bath. A digital camera was
used to capture images that were processed offline using the image processing software ImageJ®.

Melting of n-octadecane, initially at melting temperature, was observed for 30 minutes by heating it from
the centre at 40 (Ste = 0.148) and 60 ºC (Ste = 0.385). The images obtained at an interval of 15 minutes
were processed to calculate the melt fractions. Melt fractions were also obtained through the COMSOL
simulations, which are described in the present study, for central tube temperatures of 40 and 60 ºC. Melt
fractions from experiments and simulations are presented in Fig. 3. The melt fractions obtained from
experiments and from simulation are comparable given the potential variability in experimental systems.
The melt fractions observed after 30 minutes were lower under experimental conditions, likely stemming
from the axial surfaces of the experimental system being maintained at near-melting point conditions
compared to an assumed extension of 2-D conditions in the axial direction during simulation.

In addition to the overall melt fraction, it is useful to compare the melted state of n-octadecane after 30
minutes, which are shown in Fig. 4. Images showing an area of 76.2 × 38.1 mm on the symmetric right
side near the centre of the experimental setup are shown here. The core cylinder appears to be bigger in

7
Figure 3. Numerical and experimental melt fraction of n-octadecane.

the figures for experimental results. It should be noted that the cylindrical part seen at the centre of the
experimental images is not the core cylinder but rather the tubing outside the enclosure used to carry the
circulating water. In the experimental images, the white portions represent the solid PCM and the see-
through portions represent the liquid PCM. For the numerical melt profiles, the dark blue portions
represent the solid PCM with all other colours representing liquid PCM at various temperatures; the
yellow line separating the two regions representing the solid-liquid interface. The corresponding
experimental and numerical images closely match each other. For experimental results at 40 ºC, the solid
and liquid portions are not as distinguishable as they are for 60 ºC, which may also cause inaccuracy in
calculating the experimental melt fraction and in turn the comparatively large difference between the
experimental and numerical melt fractions at 40 ºC.

RESULTS AND DISCUSSION

As mentioned earlier, melting of n-octadecane and dodecanoic acid in a horizontal cylindrical annular
enclosure was studied numerically. The diameter of the outer cylinder was fixed at 127 mm and two
diameters for the inner cylinder were used: 6.35 and 25.4 mm. The surface of the inner cylinder was
heated with temperatures corresponding to Stefan numbers, as defined in Eq. (11), for PCMs initially at
melting temperature of 0.2, 0.3 and 0.4. The temperatures corresponding to the Stefan numbers of 0.2,
0.3 and 0.4 were 44.4, 52.8 and 61.3 ºC, respectively for n-octadecane and 61.9, 71.2 and 80.4 ºC,
respectively for dodecanoic acid. In the following few subsections the results are presented and discussed
for each PCM separately.

( − )
Ste = A f
g∆ℎ (11)

8
Figure 4. Melt profiles of n-octadecane after 30 minutes of melting.

n-octadecane Melt profiles, maximum velocity in the melted area and the surface heat flux for n-
octadecane are presented in the following subsections. Melt profiles for Stefan number Ste = 0.3 are not
presented because their characteristics are identical to the other corresponding results presented here.

Melt profiles. The melt profiles of n-octadecane for core diameter of 6.35 mm are shown in Fig. 5 and
that for core diameter of 25.4 mm are shown in Fig. 6. It should be noted that only the inner portions
close to the heating surface are shown here. For core diameter of 6.35 mm, a region of 25.4 × 12.7 mm

9
Figure 5. Melt profiles of n-octadecane for core diameter of 6.35 mm (T in K).

is shown; for core diameter of 25.4 mm, 50.8 × 25.4 mm. The profiles are shown for Stefan numbers of
0.2 and 0.4. The time scale for the melt profiles are chosen such that it shows the phases of pure
conduction, formation of instability, initiation of convection heat transfer and presence of multi-cell
convection current. Therefore, the time scales for different Stefan numbers or different diameters are
different. For Ste = 0.2, regardless of the core diameter, the melt profile is purely circular for the first
240 seconds, which indicates that heat transfer is purely by conduction. Up to 180 seconds the fluid
layers remain stable. At 240 seconds, the molten PCM layer at the top of the hot surface becomes
unstable and starts to form circulating cells, which are dominant at 480 seconds. A single circulating cell
(in the right half of the enclosure) is observed in the case of small core diameter, while two counter-
rotating cells are generated (again, in the right half of the enclosure) in the case of the core diameter of
25.4 mm. Hence, it can be concluded that for Ste = 0.2, the heat transfer takes place by conduction up to
240 seconds regardless of the core diameter. After 240 seconds, convection heat transfer starts to play a
role and becomes dominant after 480 seconds when the melt profiles become noncircular. The melt
profiles for Ste = 0.4 develop very similarly to that for the Ste of 0.2 except that for Ste = 0.4, the PCM
melts much faster.

Maximum velocity. The maximum velocities in the melted area of n-octadecane are presented in Fig. 7.
The velocities in the figures represent the maximum velocity anywhere in the melted area at any time.
Velocity of the fluid is very low in the beginning because convection heat transfer is not significant yet.

10
Figure 6. Melt profiles of n-octadecane for core diameter of 25.4 mm (T in K).

It is seen from Figs. 5 and 6 that at low Stefan number, it takes approximately 240 seconds (4 min) for
natural convection to start playing a role; that is when the velocity starts to increase rapidly regardless of
the core diameter. For Ste = 0.4, velocity increases almost immediately as convection heat transfer starts
to dominate as soon as after 1 minute. In the beginning of the melting phase the fluid velocity depends
more on the Stefan number than on the diameter of the heating surface. For the core diameter of 6.35
mm, the maximum velocity increases steadily after a fast increase in the beginning. However, for core
diameter of 25.4 mm, spikes are observed in the maximum velocity profiles which show a more chaotic
natural convection regime which comes from the appearance of two convection cells in this case. The
maximum velocity profiles are chaotic when at least two cells of convection current of similar
prominence are present in the molten PCM. Much fluctuation in velocity profiles is not observed when
one cell (in right half of the enclosure) becomes much more prominent than the other cells. To illustrate
the phenomenon, melt profiles of n-octadecane for core diameter of 25.4 mm and Ste = 0.2 and 0.4 are
presented in Fig. 8. It should be noted from Fig. 7b) that the maximum velocity profile for Ste = 0.2 is
smooth up to 29 minutes and after that a spike appears. It is seen from Fig. 8 that the convection current
in molten PCM for Ste = 0.2 is indeed weak in the beginning. After 29 minutes, strong convection cells
of similar prominence forms in the molten PCM, which causes the spikes that are seen in the maximum
velocity profiles. Also, Fig. 7b) shows that the maximum velocity profile for Ste = 0.4 is chaotic between
4 and 12 minutes and smooth afterward. For Ste = 0.4, multi-cell convection current of similar
prominence exists at 7 minutes, while at 15 minutes, one cell is much more prominent than the other, as
shown in Fig. 8.

11
Figure 7. Maximum velocity in the melted area of n-octadecane for core diameter: a) 6.35 mm,
b) 25.4 mm.

12
Figure 8. Influence of multi-cell convection current in molten n-octadecane for core diameter of
25.4 mm

Surface heat flux. The surface heat flux of n-octadecane for heating surafce diameters of 6.35 mm and
25.4 mm are presented in Fig. 9. The presented surface heat fluxes are for the two cases studied: first,
considering heat transfer takes place by conduction only, and second, accounting for convection heat
transfer in the simulations. The surface heat flux per unit axial length was obtained using COMSOL
built-in heat flux at a boundary function. For a particular inner cylinder diameter and a particular Stefan
number, the heat flux decreases sharply in the beginning of the melting process, as thermal resistance
between the solid surface and the melt interface builds up by small fraction of molten PCM and the
temperature difference driving the heat transfer decreases due to presence of hot molten PCM adjacent
to the hot surface. Initially the heat flux is the same for the conduction only heat transfer case and the
one where convection heat transfer is accounted for. This indicates that in the beginning heat transfer
takes place only by conduction. Once convection heat transfer begins in the molten PCM, the circulation
of the hot molten PCM reduces the thermal resistance, redistributes the energy and brings the colder
PCM close to the heating surafce. This causes the heat flux in the convection heat transfer case to increase
slowly. The heat flux in conduction only heat transfer case continues to decrease.

The heat flux in the convection heat transfer case supercedes that of the conduction only case fastest at
the highest Stefan numbers, emphasizing the fact the Stefan number plays a significant role in the onset
of natural convection. However, close observation of Fig. 9 reveals that the time at which the difference
in the magnitude of heat flux between conduction and convection case does not depend significantly on
the diameter of the inner cylinder. This indicates, as did the melt profiles and the maximum velocity
results that onset of natural convection is weakly dependent on the diameter of the inner cylinder. In the
beginning, the heat flux decreases more in the case of 25.4 mm core diameter than for the 6.35 mm. This
larger decrease in heat flux at the beginning of the melting process in enclosure with large core diameter
is due to higher thermal resistance present in that system.

Dodecanoic Acid The melt profiles and the maximum velocity profiles in the melted area of dodecanoic
acid are presented for core diameters of 6.35 mm and 25.4 mm in the following subsections. The heat
flux results of dodecanoic acid are not presented as they are qualitatively the same and quantitatively
varies only slightly from that of n-octadecane.

Melt profiles. The melt profiles of dodecanoic acid are presented in Figs. 10 and 11, for core diameter
of 6.35 mm and 25.4 mm, respectively. As was done for n-octadecane, the inner 25.4 × 12.7 mm
segments for core diameter of 6.35 mm and 50.8 × 24.5 mm segments for core diameter of 25.4 mm are

13
Figure 9. Surface heat flux of n-octadecane for core diameter: a) 6.35 mm, b) 25.4 mm.

14
Figure 10. Melt profiles of dodecanoic acid for core diameter of 6.35 mm (T in K).

presented. The melt profiles are presented for the same times as they were presented for n-octadecane to
facilitate comparison. Dodecanoic acid melts in a very similar fashion to n-octadecane although a bit
slowly. It should be noted that dodecanoic acid has many similar thermophysical properties to that of n-
octadecane (presented in Table 1 and Table 4). However, the density of dodecanoic acid is significantly
higher and the thermal conductivity in solid state is significantly lower than that of n-octadecane. In the
present study, the thermal conductivity of solid phase will not have any contribution to the heat transfer
rate since the PCM is initially at its melting temperature. It also needs to be noted that dynamic viscosity
of dodecanoic acid is double (at the melting point) that of n-octadecane. The slower melting (by volume)
of dodecanoic acid is attributed partly to its high viscosity and to its high density. It was found by Ho
and Gao [2013] that increased viscosity inhibits convection heat transfer.

Maximum velocity. The maximum velocity of molten dodecanoic acid at different times are presented
in Fig. 12 for core diameter of 6.35 mm and 25.4 mm. Again, similar trend in the maximum velocity
curves as that of n-octadecane is observed. However, the magnitudes of maximum velocities at
corresponding times is higher for n-octadecane than for dodecanoic acid, indicating stronger convection
current in the melted n-octadecane. For larger core diameter, the fluctuations in maximum velocity curve
at Stefan numbers of 0.3 and 0.4 are associated with the formation of multi-cell rotating convection
currents in the melted area and their interaction with each other, as was the case for n-octadecane.
However, the multi-cell convection currents in the molten dodecanoic acid are not as prominent as they
are in molten n-octadecane. This is again attributed to the high viscosity of dodecanoic acid.

15
Figure 11. Melt profiles of dodecanoic acid for core diameter of 25.4 mm (T in K).

Comparative Study In the following subsections, the melted volume of the PCMs, surface heat flux
and the total stored energy in the PCMs are compared.

Melted volume. The melted volumes, simulated accounting for convection heat transfer, of n-octadecane
and dodecanoic acid for all the Stefan numbers studied are presented for core diameter of 6.35 mm and
25.4 mm in Fig. 13. As was indicated by the melt profiles and the maximum velocity curves, dodecanoic
acid melts at a slower rate, although slightly. At low Stefan numbers, the melting of dodecanoic acid
follows that of n-octadecane very closely and the difference becomes pronounced as the Stefan number
gets larger.

Surface heat flux. The surface heat flux results, obtained from simulations accounting for convection
heat transfer, of n-octadecane and dodecanoic acid for all Stefan numbers studied are presented in
Fig. 14 for core diameter of 6.35 mm and 25.4 mm. Again, the heat flux results were obtained from
COMSOL using its built-in function. The heat flux for dodecanoic acid follows the same trend as that of
n-octadecane but being slightly higher. The high heat flux of dodecanoic acid compared to that n-
octadecane may appear contradictory to the melted volume results at first sight. However, it should be
noted that dodecanoic acid has significantly higher density than that of n-octadecane, which means
higher mass of melted dodecanoic acid compared to n-octadecane. This requires more energy to melt
dodecanoic acid than to melt n-octadecane and thus higher surface heat flux in dodecanoic acid. At high

16
Figure 12. Maximum velocity in the melted area of dodecanoic acid for core diameter: a) 6.35 mm,
b) 25.4 mm.

17
Figure 13. Melted volume of the PCMs for core diameter: a) 6.35 mm, b) 25.4 mm.

18
Figure 14. Surface heat flux of the PCMs for core diameter: a) 6.35 mm, b) 25.4 mm.

19
Stefan numbers and small core diameters, the difference between the heat flux of dodecanoic acid and
n-octadecane is more pronounced. The oscillations seen in surface heat flux in Fig. 14 are the result of
transient convection cells that are progressing, changing size and strength over time as more PCM melts.

Total stored energy. The total stored energy in the PCMs is the summation of the energy required to melt
the PCMs and the energy required to raise the temperatures of the molten PCMs. The total stored energy
in the PCMs obtained from COMSOL are shown in Fig. 15 for core diameter of 6.35 mm and 25.4 mm.
Regardless of the diameter of the heating surface, dodecanoic acid stores slightly more energy than n-
octadecane after 30 minutes of melting. As mentioned earlier for surface heat flux, the higher energy
storage of dodecanoic acid is due to melting of more mass. It should be noted that the same Stefan
numbers for the two PCMs were obtained by varying the heating surface temperature. The differences
between wall and melting temperatures for dodecanoic acid were 9.2% higher than for n-octadecane,
which could contribute to faster heat transfer rate in dodecanoic acid. Again, higher density and dynamic
viscosity of dodecanoic acid would also influence the difference in heat transfer between dodecanoic
acid and n-octadecane.

The critical values, at the onset of convection heat transfer, of the maximum velocity, volume and
diameter of the melted PCM are presented in Table 5 for n-octadecane and Table 6 for dodecanoic acid;
the critical diameter at which the onset of convection starts is expressed in units of diameter of the tube
d. All the quantities mentioned above are highly dependent on the Stefan number. The metrics show that
dodecanoic acid sees convection starts at diameters and melted volume slightly larger than n-octadecane,
while its melted PCM velocity is also slower. The results also show that the impact of the critical
diameter of molten PCM becomes less at the high Stefan numbers.

Table 5
Critical diameter, velocity and volume of molten n-octadecane at convection onset

Ste Small diameter (d = 6.35 mm) Large diameter (d = 25.4 mm)


dcr /d vcr (mm/s) Vcr (cm3/m) dcr /d vcr (mm/s) Vcr (cm3/m)
0.2 1.75 0.598 33.7 1.20 0.751 115.6
0.3 1.59 0.787 25.7 1.15 0.957 90.0
0.4 1.54 1.04 22.6 1.14 1.32 79.1

Table 6
Critical diameter, velocity and volume of molten dodecanoic acid at convection onset

Ste Small diameter (d = 6.35 mm) Large diameter (d = 25.4 mm)


dcr /d vcr (mm/s) Vcr (cm3/m) dcr /d vcr (mm/s) Vcr (cm3/m)
0.2 1.77 0.517 36.3 1.21 0.672 123
0.3 1.62 0.666 27.2 1.16 0.858 93.5
0.4 1.58 0.907 24.4 1.15 1.03 84.8

20
Figure 15. Stored energy in the PCMs for core diameter: a) 6.35 mm, b) 25.4 mm.

21
CONCLUSIONS

The onset of natural convection during melting of n-octadecane and dodecanoic acid in annular
horizontal cylindrical enclosures heated from the inner surface was studied numerically using COMSOL
Multiphysics. The diameter of the outer surface was 127 mm in all the simulations and the diameter of
the inner surface was varied: 6.35 and 25.4 mm. The inner surface of the enclosure was heated at
temperatures corresponding to Stefan numbers of 0.2, 0.3 and 0.4. A two-dimensional melting process
was simulated only on the right half of the enclosure about the vertical axis of symmetry. The simulations
were performed for 30 minutes of melting time, which was sufficient for onset of natural convection.

The study revealed that the melting phenomenon in n-octadecane and dodecanoic acid is essentially the
same. The volumetric melting of n-octadecane was slightly faster than that of dodecanoic acid, and more
pronounced over that of dodecanoic acid at high Stefan numbers. However, heat flux results indicated
slightly higher heat transfer in dodecanoic acid. Although the volumetric melt fraction of n-octadecane
was higher than that of dodecanoic acid, energy storage was higher in dodecanoic acid due to its high
density. Onset of natural convection was faster at high Stefan numbers and mostly independent of the
heating surface diameter. However, larger diameter of the inner surface augments conduction heat
transfer in the beginning of the melting process.

ACKNOWLEDGEMENT

The authors are thankful for financial contributions from the Nova Scotia Graduate Scholarship
(NSGS) funds, the Natural Sciences and Engineering Research Council of Canada (NSERC), the
Canada Foundation for Innovation (CFI) and the Faculty of Graduate Studies (FGS) at Dalhousie
University

REFERENCES

Akgün, M., Aydın, O. and Kaygusuz, K. [2007], Experimental study on melting/solidification


characteristics of a paraffin as PCM, Energy Conversion and Management, Vol. 48, pp. 669-678.

Arasu, A. V. and Mujumdar, A. S. [2012], Numerical study on melting of paraffin wax with Al2O3 in a
square enclosure, International Communications in Heat and Mass Transfer, Vol. 39, pp. 8-16.

Avci, M. and Yazici, M. Y. [2013], Experimental study of thermal energy storage characteristics of a
paraffin in a horizontal tube-in-shell storage unit, Energy Conversion and Management, Vol. 73, pp.
271-277.

Darzi, A. R., Farhadi, M. and Sedighi, K. [2012], Numerical study of melting inside concentric and
eccentric horizontal annulus, Applied Mathematical Modelling, Vol. 36, pp. 4080-4086.

Dhaidan, N. S., Khodadadi, J. M., Al-Hattab, T. A. and Al-Mashat, S. M. [2013], Experimental and
numerical investigation of melting of phase change material/nanoparticle suspensions in a square
container subjected to a constant heat flux, International Journal of Heat and Mass Transfer, Vol. 66,
pp. 672-683.

Duan, Y., Hosseinizadeh, S. F. and Khodadadi, J. M. [2010], Effects of Insulated and Isothermal Baffles
on Pseudosteady-State Natural Convection Inside Spherical Containers, Journal of Heat Transfer, Vol.
132, pp. 062502.

22
Fan, L.-W., Xiao, Y.-Q., Zeng, Y., Fang, X., Wang, X., Xu, X., Yu, Z.-T., Hong, R.-H., Hu, Y.-C. and
Cen, K.-F. [2013], Effects of melting temperature and the presence of internal fins on the performance
of a phase change material (PCM)-based heat sink, International Journal of Thermal Sciences, Vol. 70,
pp. 114-126.

Ho, C. J. and Gao, J. Y. [2013], An experimental study on melting heat transfer of paraffin dispersed
with Al2O3 nanoparticles in a vertical enclosure, International Journal of Heat and Mass Transfer, Vol.
62, pp. 2-8.

Kheirabadi, A. C. and Groulx, D. 2015. The Effect of the Mushy-Zone Constant on Simulated Phase
Change Heat Transfer. CHT-15 ICHMT International Symposium on Advances in Computational Heat
Transfer. Rutgers University, Piscataway (USA). 22 p.

Liu, C. and Groulx, D. [2014], Experimental study of the phase change heat transfer inside a horizontal
cylindrical latent heat energy storage system, International Journal of Thermal Sciences, Vol. 82, pp.
100-110.

Longeon, M., Soupart, A., Fourmigué, J.-F., Bruch, A. and Marty, P. [2013], Experimental and
numerical study of annular PCM storage in the presence of natural convection, Applied Energy, Vol.
112, pp. 175-184.

Murray, R. E., Desgrosseilliers, L., Stewart, J., Osbourne, N., Marin, G., Safatli, A., Groulx, D. and
White, M. A. 2011. Design of a Latent Heat Energy Storage System Coupled with a Domestic Hot Water
Solar Thermal System. World Renewable Energy Congress 2011. Linköping. 8 p.

Murray, R. E. and Groulx, D. [2014], Experimental study of the phase change and energy characteristics
inside a cylindrical latent heat energy storage system: Part 1 consecutive charging and discharging,
Renewable Energy, Vol. 62, pp. 571-581.

Regin, A. F., Solanki, S. C. and Saini, J. S. [2008], Heat transfer characteristics of thermal energy storage
system using PCM capsules: A review, Renewable and Sustainable Energy Reviews, Vol. 12, pp. 2438-
2458.

Sciacovelli, A., Colella, F. and Verda, V. [2013], Melting of PCM in a thermal energy storage unit:
Numerical investigation and effect of nanoparticle enhancement, International Journal of Energy
Research, Vol. 37, pp. 1610-1623.

Shmueli, H., Ziskind, G. and Letan, R. [2010], Melting in a vertical cylindrical tube: Numerical
investigation and comparison with experiments, International Journal of Heat and Mass Transfer, Vol.
53, pp. 4082-4091.

Tan, F. L., Hosseinizadeh, S. F., Khodadadi, J. M. and Fan, L. [2009], Experimental and computational
study of constrained melting of phase change materials (PCM) inside a spherical capsule, International
Journal of Heat and Mass Transfer, Vol. 52, pp. 3464-3472.

Tao, Y. B. and He, Y. L. [2015], Effects of natural convection on latent heat storage performance of salt
in a horizontal concentric tube, Applied Energy, Vol. 143, pp. 38-46.

Varol, Y. and Okcu, M. 2013. Numerical Investigation of Fins Effect for Melting Process of Phase
Change Materials. ASME International Mechanical Engineering Congress and Exposition, November
15-21, 2013 USA, 6 p.

Yaws, C. L. 2003. Yaws' Handbook of thermodynamic and Physical Properties of Chemical Compounds,
Knovel.

23
View publication stats

You might also like