International Journal of Heat and Mass Transfer 86 (2015) 465–477

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer

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

Close-contact melting in a horizontal cylindrical enclosure with

longitudinal plate fins: Demonstration, modeling and application to
thermal storage
T. Rozenfeld, Y. Kozak, R. Hayat, G. Ziskind ⇑
Heat Transfer Laboratory, Department of Mechanical Engineering, Ben-Gurion University of the Negev, P.O.B. 653, Beer-Sheva 84105, Israel

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

Article history: The present study deals with melting in a horizontal double-pipe concentric storage unit with a
Received 3 August 2014 longitudinally finned inner tube. This geometry, applicable for thermal energy storage, has been
Received in revised form 19 February 2015 extensively studied in the past. However, close-contact melting (CCM), which may significantly increase
Accepted 24 February 2015
the melting rate, has not been explored for this sort of systems.
In the present experiments, a laboratory-scale three-fin unit is transparent and thus the processes
inside it are observed and recorded. Close-contact melting is achieved by supplying heat to the outer shell
Keywords:
of the unit: the solid phase is detached from the shell and moves in the liquid phase. In the upper part of
PCM
Heat storage
the unit, the solid phase translates vertically and melts on the inclined fin surfaces. In the lower parts of
Longitudinal fins the unit, the solid phase approaches the vertical fin by rotation, while its outward surface slides on the
Close-contact melting shell. It is shown that close-contact melting shortens the melting time by a factor of 2.5 in that specific
Modeling laboratory-scale device.
Dimensional analysis A novel theoretical model includes gravity-induced rotational motion of the solid, primary melting on a
vertical fin with non-uniform temperature distribution, secondary melting at the shell, and frictional
resistance at the latter. The model is validated using the experimental results, and then used for a
parametric investigation and dimensional analysis. It is revealed that the melt fraction depends on the
Fourier and Stefan numbers combined as FoSte3/4, whereas the Nusselt number depends also on an
additional group, Ste1/4. Further generalization is achieved when different angles between the fins are
considered.
The present paper presents a further proof of the special role that close-contact melting (CCM) can play
in thermal energy storage units and finned systems in general. It is demonstrated that the fins, when
properly designed and oriented, induce CCM in their vicinity, leading to a very significant increase in
the melting rate.
Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction of various sizes and configurations have been tried, from common
fins [5–7] to heat pipes [8–11] and thermosyphons [12].
The main attractiveness of latent-heat thermal energy storage One possible implementation of a latent heat thermal storage
systems (LHTES) stems from the large heat-storage capacity of device is a finned tube with an array of fins. The tube can be
phase-change materials (PCMs). Different systems have been enclosed in a cylindrical shell filled with PCM. A heat transfer fluid
studied in the literature, addressing their overall characteristics (HTF) flows through the tube and heat is conducted from the tube
[1], materials [2], solar energy applications [3], and modeling to the fins that are in contact with the PCM. The thermal storage
methods [4]. Of particular concern in the design of these systems charging/discharging process is driven by a hot/cold HTF inside
is the low thermal conductivity of the PCMs, and various ways to the tube that causes the PCM to melt/solidify. This type of storage
overcome this deficiency have been suggested. As in the other unit can be set in different orientations, usually with the tube axis
fields of heat transfer and thermal engineering, extended surfaces vertical or horizontal. Two types of fins are the most common,
radial (circumferential) fins and longitudinal fins, although many
other configurations, including somewhat exotic ones, have been
⇑ Corresponding author. Tel.: +972 8 6477089. also suggested [13].
E-mail address: gziskind@bgu.ac.il (G. Ziskind).

http://dx.doi.org/10.1016/j.ijheatmasstransfer.2015.02.064
0017-9310/Ó 2015 Elsevier Ltd. All rights reserved.
466 T. Rozenfeld et al. / International Journal of Heat and Mass Transfer 86 (2015) 465–477

Nomenclature

cp specific heat, kJ/(kg K) dr molten layer thickness at the fin, m

g gravitational acceleration, m/s2 h angular coordinate
k thermal conductivity, W/(m K) l dynamic viscosity, kg/(m s)
L latent heat, kJ/kg m kinematic viscosity, m2/s
M moment per unit depth, N q density, kg/m3
p pressure, Pa s shear stress, Pa
q’’ heat flux, W/m2 u auxiliary angle used at the envelope
r radial coordinate, m / angle between fins
R radius of the solid phase, m
s solid–liquid interface location, m Subscripts
S cross-section area, m2 Al aluminum
t time, s avg average
T temperature, °C eff effective
u velocity component, m/s env envelope
x Cartesian coordinate, m i inner
Xc center of mass coordinate, m l liquid
y Cartesian coordinate, m m melting
Yc center of mass coordinate, m o outer
r radial
Greek letters s solid
a angular displacement of the solid w wall
al thermal diffusivity, m2/s w0 fin base
d molten layer angular dimension

Latent heat thermal storage units with radial (circumferential) the same group [34–35]. Hamdani and Mahlia [36] examined
arrays of fins have been comprehensively explored in the past, various cases for melting in a vertical tube: the case of a tube
for both solidification and melting. A recent paper [14] surveys without fins was compared with a tube with radial and longitudinal
some representative examples [15–22]. Most recent papers fin arrays. Mat et al. [37] studied a horizontal storage unit with
address such diverse issues as comparison with pinned tubes three concentric tubes, where the HTF flows in the most inner
[23], effectiveness analysis [24], and entropy generation [25]. and most outer tubes, and the PCM is in the middle tube. The effect
Latent heat storage units with longitudinal fins have been stud- of longitudinal fins attached to the inner and outer tubes was
ied both experimentally and theoretically for more than 30 years. tested. This configuration was further explored by the same group
Sparrow et al. [26] conducted experiments for solidification on a [38–39]. Most recently, Liu and Groulx [40] studied experimentally
vertical tube with and without a longitudinal fin array. It was found the melting and solidification processes in a horizontal finned tube
that the presence of fins significantly enhances the solidification with four longitudinal fins. Two different fin orientations, which
process. Bathelt and Viskanta [27] studied experimentally melting may be defined as ‘‘X’’ and ‘‘ + ’’, were tested. The same authors
and solidification on a horizontal finned tube with three longitudi- studied a similar vertical system under different conditions
nal plate fins, and compared the results with the case of an unfinned [41–42]. The most recent work on this subject introduces a certain
tube. Also, the effect of different fin orientations was examined. The configuration of fins [43].
same configuration was investigated experimentally by Betzel and Fig. 1 presents an original two-dimensional CFD simulation by
Beer [28], where two different fin materials were used: copper and the authors that illustrates patterns commonly encountered in
PVC. The results were generalized and a correlation for the Nusselt units with a longitudinally-finned inner tube with a horizontal
number was derived. Padmanabhan and Krishna Murthy [29] axis. This type of CFD simulation has been introduced by [44]
numerically simulated melting/solidification on a tube with an and allows to study the role of convection in the melt, the effects
arbitrary number of longitudinal fins. A two-dimensional numerical of fin design and the influence of the storage device working
model was developed under the assumption that the only heat conditions. Here and below we are going to define the presented
transfer mode is conduction. The simulation results were general- 3-fin system as ‘‘Y’’ configuration, noting that it has much in
ized by a correlation for the melt fraction. Al-jandal and Sayigh common with ‘‘X’’ and ‘‘ + ’’ configurations [40] which have 4 fins.
[30] compared experimentally melting on a vertical finless tube One can see in Fig. 1 a number of important features of the
with radially and longitudinally finned tubes. It was found that process. For instance, in the upper part of the unit melting is
the fins enhance the heat transfer rate and that natural convection dominated by conduction at the early stages of the process,
plays a significant role in the melting process. Ismail et al. [31] whereas at the later stages convection plays the decisive role.
developed a two-dimensional numerical model, based on heat In the lower parts of the unit, the role of convection also increases
conduction only, for the solidification process on a vertical with time. As a result, melting in the unit occurs mostly by con-
longitudinally finned tube. The numerical findings were close to vection, and as heating from below is more effective, melting in
experimental results. Also, the effects of fin geometry and tube the upper part of the unit is more rapid. Also, due to convection
temperature were examined. Solomon and Velraj [32] tested the solid phase in the lower parts of the unit is not located
experimentally a tube with eight longitudinal fins compared symmetrically as related to the fins. It will be demonstrated later
with a finless tube. Different parameters which were examined that the patterns in Fig. 1, which prove the positive effect of fins,
included the fin length. Agyenim et al. [33] explored a storage unit are indeed encountered experimentally under commonly used
based on a horizontal tube, with and without fins. Two types of fin conditions. Still, the molten material between the hot fin surfaces
configurations were compared: radial and longitudinal. Later on, and the remaining solid presents a significant thermal resistance
the longitudinal finned storage unit was further investigated by which impedes the process.
 T. Rozenfeld et al. / International Journal of Heat and Mass Transfer 86 (2015) 465–477 467

[46], and assumed that sensible heating can be neglected, the

process is quasi-steady, the thin molten layer thickness is small
relatively to the radius, and the solid–liquid interface retains its
circular shape. They found that 85% of heat transfer rate was due
to CCM from the bottom of the shell, and only 15% was due to
natural convection. This type of analysis with similar assumptions
has been further developed for melting in a spherical enclosure
[47–49], hot spherical [50], cylindrical [51] and arbitrary-shape
[52] body migrating through solid PCM, melting in an isothermal
rectangular capsule [53], and melting of a vertical solid cylinder
heated from below [54]. Bejan [55] found a generalization for five
simple geometries: cylindrical, spherical and rectangular capsules
and a migrating cylinder or sphere.
Kozak et al. [14] presented the first ever analysis of close-
contact melting (CCM) effects in a storage unit with fins, built of
a circumferentially finned inner tube concentric with a circular
shell. It was demonstrated that CCM enhances significantly the
heat transfer rate, shortening the melting time by almost 2.5 times
in that specific laboratory-scale device. The results showed that it
is favorable to supply heat to the outer shell of a latent-heat
storage unit in order to initiate close-contact melting. Since melt-
ing occurs on the fins, their role becomes much more significant
than just extended surfaces.
In the present work, this idea is further explored for a storage
unit with longitudinal fins. In order to have CCM in such system,
its axis should not be vertical. Thus, a horizontal tube with a
number of fins is explored, attempting at maximizing the positive
CCM effects. Since the fins themselves are not necessarily horizon-
tal, see Fig. 1 above, it may be expected that the solid would not
move along the vertical direction. There are no works addressing
similar problems in the literature, except for Kumano et al. [56]
who studied close-contact melting of a rectangular parallelepiped
solid on a hot surface under asymmetric loads. The motion of the
solid in that case is more complex than linear sinking motion, as
the asymmetric load on the top of the solid bulk causes the solid
to rotate, as well. Thus, not only the equation for the force balance
is solved, but also the moment balance due to the solid bulk
Fig. 1. Typical melting patterns in a horizontal concentric latent heat storage unit rotation.
with longitudinal fins in ‘‘Y’’ configuration with uniformly spaced fins (blue – the In the next section, an experimental set-up is introduced and
solid phase, red – the liquid phase). (For interpretation of the references to color in experimental results are presented. They clearly demonstrate
this figure legend, the reader is referred to the web version of this article.)
melting enhancement and acceleration due to the close-contact
melting achieved on the fin surfaces inside the unit. Then, a hybrid
However, a certain feature remains mistreated: one can see in analytical–numerical model is developed for the lower part of the
Fig. 1 that the solid PCM, which is denser than the liquid, remains unit, where close-contact melting occurs mostly on a vertical sur-
stationary in the upper side of the storage unit and does not sink face. A comparison between the predictions and experimental
due to the gravitational forces. Neither does the solid move in results is presented and discussed. Then, the model is further used
the lower parts. This phenomenon, also seen in experiments, takes for a dimensional analysis which reveals the governing dimen-
place because the solid phase does not melt near the shell and sional parameters and leads to generalization of the results. We
remains stuck to it. It is obvious that if the PCM solid bulk were note that the upper part of the unit may be modeled analytically,
detached from the shell, it would sink and approach the hot similarly to the model reported in [57] for a half-cylinder on a
surfaces. This effect of the sinking solid PCM in a horizontal storage horizontal heated surface.
unit with longitudinal fins has not been addressed experimentally
or theoretically in the available literature. 2. Experimental study
Actually, the phenomenon of a sinking solid bulk that melts
directly on a hot surface is well known and called close-contact The experimental set-up is shown in Fig. 2(a), whereas the stor-
melting (CCM). A thin molten layer is formed between the solid age unit appears in Fig. 2(b). The experimental set-up includes the
and the hot surface, as the melt is squeezed to the sides by the following main parts: (1) two thermostatic circulators, (2) control
descending solid bulk. Thus, the PCM is melted continuously, valves, (3) transparent water tank, (4) PCM storage unit, and (5)
providing for the flow in the thin molten layer. CCM allows high immersion thermostat. Additional parts include piping, stands,
melting and heat transfer rates due to the relatively thin layer of etc. A detailed description is presented in a previous work [14],
a liquid PCM, across which heat is conducted to the solid phase. hence only the essentially new particulars are given in detail here.
Close-contact melting in various configurations has been The main objective of the experimental work is to demonstrate
explored both experimentally and theoretically. Usually the the role of fins in melting. They are not just extended surfaces
theoretical models are developed under various assumptions, in which contribute to a larger heat transfer area: the fins serve as
order to simplify the problem. Bareiss and Beer [45] studied the surfaces on which the close-contact melting occurs. For this
melting in a horizontal cylindrical enclosure, considered also in reason, the unit is installed horizontally. Two groups of
468 T. Rozenfeld et al. / International Journal of Heat and Mass Transfer 86 (2015) 465–477

shell has an external diameter of 50 mm, internal diameter of

1 2 3 4 5 44 mm and height of 218 mm. Note that unlike the previous work
[14], in this study it was essential to observe the process in the
axial direction, and thus transparent flanges were designed and
built. Thus, the geometry and dimensions allow to obtain and
observe all the features that characterize a future real-size unit.
The PCM used in the present study is a commercial-grade
eicosane (96% C20H42, Roper Thermals, Clinton, CT) [14]. The total
mass of PCM used in the experiments is 0.18 kg, leaving enough
room for volumetric expansion, which occurs mostly due to the
solid–liquid transition.
In a typical experiment, the storage device is held by two
clamps that keep it strictly horizontal, as can be seen in Fig. 2(a).
When the experiment is performed in air, the entire storage device
is initially at room temperature, and the PCM is solid. The HTF inlet
valve is opened, allowing hot water from a circulator to flow in the
tube, initiating melting. When the experiment is performed in a
warm water bath, the unit is kept first in the water without
allowing the HTF to enter the inner tube. Once the initial melting
near the shell is achieved and the solid is free to move, the valve
that controls the HTF flow is opened and the process starts.

(a)
19 min

23 min

(b)
Fig. 2. Experimental setup: (a) overview (1 – thermostatic circulators, 2 – control
valves, 3 – transparent water tank, 4 – PCM storage unit, 5 – immersion thermostat)
and (b) PCM storage unit with three uniformly spaced (/ = 2p/3) longitudinal fins.

experiments are performed, namely, with the unit shell exposed to

ambient air, like in the literature, and with the unit shell exposed
to a heated controlled environment, like in [14]. A transparent shell
of the unit, including its flanges, allows to achieve visualization. In
order to enable close-contact melting for demonstration purposes,
in the present study the entire storage device may be submerged in 31 min
a specially-designed transparent bath filled with heated water. As
shown in the previous work [14], the only purpose of the shell
heating is to detach the solid from it and to allow the former to
sink, whereas its contribution to the overall melting is very small.
The PCM storage unit, depicted in Fig. 2(b), includes a finned
tube, a transparent shell, transparent flanges and seals. The PCM
is stored between the tube/fins and the shell. The one-piece finned
tube is made of aluminum 7075 and manufactured by machining.
The tube is 280 mm long with an internal diameter of 8 mm and
wall thickness of 2 mm, meaning that its external diameter is
12 mm. The fin array includes three longitudinal plate fins. Each
fin has a dimension of 180 mm along the tube, 15.5 mm along
the radial direction and 3 mm thick. The finned tube is sealed, Fig. 3. Typical observed melting patterns in the lower part of the unit when the
using flanges and O-rings, inside a Perspex cylindrical shell. The PCM sticks to the shell.
 T. Rozenfeld et al. / International Journal of Heat and Mass Transfer 86 (2015) 465–477 469

Fig. 3 presents an experiment in which the unit shell was

exposed to ambient air at room temperature. The melting
temperature of eicosane is 36.7 °C. The HTF flows in the tube and
its temperature is 56.7 °C, remaining practically uniform along
the unit axis [14]. Thus, melting occurs only at the tube and fins,
whereas the shell plays no role in the process. In each of the lower
parts, heat is transferred by convection, whereas the solid remains
attached to the shell. The shape of the remaining solid is not
symmetric here, because a warmer liquid flows up and enhances
melting in a manner similar to tall and cylindrical enclosures
[58]. In general, melting is very similar to that predicted by the
simulation shown in Fig. 1, including in the upper part of the unit,
i.e. is typical for the systems of this kind.
Fig. 4 shows that typical melting with a slightly heated envel- 60 s
ope differs most considerably. Now, the solid in the upper part of
the unit (V-shape) sinks and melts directly above the longitudinal
fins. A thin molten layer is formed between the solid and the fin. It
appears that this layer does not grow noticeably throughout the
melting process. Although the hot surface is inclined, its chevron
shape allows for a strictly vertical motion only.
For the lower parts of the unit, Fig. 4 demonstrates that the
remaining solid moves in such a way that it always approaches
the vertical longitudinal fin, whereas the outer radius of the solid
bulk remains almost constant throughout the entire process.
Thus, the contribution of melting at the heated shell to phase
change proper is only minor in these parts, as well. On the other
hand, its contribution to the overall result is overwhelming: the
process is greatly enhanced because now the close-contact melting
is allowed. Quantitatively, the difference between CCM-dominated
and ‘‘regular’’ melting is illustrated in Fig. 5, which shows time-de- 180 s
pendent melt fractions for the lower two-thirds of the unit
corresponding to Figs. 3 and 4, respectively. For these specific
cases, where the temperature difference between the HTF and
PCM is 20 °C, the CCM-dominated melting is completed above
2.5 times quicker than the regular process. This result is consistent
with the role of CCM in other geometries [14]. We note, however,
that according to the numerical results of Fig. 1 and experimental
observations of Fig. 4, both ‘‘regular’’ and close-contact melting are
more rapid in the V-shaped upper part of the unit, so the overall
melting time in the entire unit will not be affected by its
weighted-average inclusion in the melt fraction (although the
instant values along the process will be slightly higher).
It should be noted that close-contact melting in the lower parts
of the unit occurs in a rather unfamiliar manner: it happens on a
vertical surface. In order to reach it, the solid should now rotate, 300 s
rather than translate, which is also highly unusual. Based on these
observations, it was decided to study the CCM in this geometry
more closely, by means of novel theoretical and numerical models
presented below.

3. Modeling

Close-contact melting in the upper part of the Y-configuration,

although it occurs on an inclined surface, is still similar to CCM
studied in the literature because it deals with a purely translational
motion of the solid in the vertical direction. In the lower part of the
Y-configuration, melting occurs primarily on the vertical fin.
Secondary melting from the envelope, whereas separating the solid
from the envelope, should not contribute substantially to the total
[14]. Therefore, one may conclude that the solid is rotating in that
420 s
part of the unit, with its outer surface sliding on the envelope with
a certain resistance. To the best of our knowledge, ‘‘rotational’’ Fig. 4. Typical observed melting patterns when the PCM is allowed to move inside
close-contact melting has never been addressed in the literature. the unit.
470 T. Rozenfeld et al. / International Journal of Heat and Mass Transfer 86 (2015) 465–477

therefore,
1 R Ro R /a R Ro R /a
Ri 0
r sin hrdrdh Ri 0
r cos hrdrdh
Xc ¼ Yc ¼ ð3Þ
R2o  R2i R2o  R2i
0.8 ð/  aÞ ð/  aÞ
2 2
and finally the center of gravity is given by:
Melt fraction

0.6 !
2ð1  cosð/  aÞÞ R3o  R3i 2ð1  cosð/  aÞÞ R2o
Xc ¼ ¼ Ri þ
2 2
3ðRo  Ri Þ ð/  aÞ 3ð/  aÞ Ro þ Ri
0.4
!
2
2 Ro
Yc ¼  sinð/  aÞ Ri þ
3ð/  aÞ Ro þ Ri
0.2 ð4Þ
CCM
Regular The melting is dominated by heat conduction through a thin
0 vertical molten liquid layer located adjacent to the vertical fin as
0 500 1000 1500 2000 2500 3000
shown in Fig. 6, whereas heat transfer at the cylindrical surface
Time, s
plays a secondary role and will be introduced later. The thickness
Fig. 5. Comparison of melting rates for regular and close-contact melting in the of the liquid layer in the close-contact area is very small relative
lower part of the unit. to the dimension of the solid. Hence, the thin layer approximation
is applied and the following assumptions are used: (1) the melting
process is quasi-steady, hence sensible heating is neglected, (2)
heat transfer by convective flow is negligible as compared with
that by heat conduction, which is assumed to be one-dimensional
Ro across the layer, (3) thermo-physical properties of the material are
constant, (4) the flow of the liquid layer is symmetric related to the
layer plane of symmetry, (5) the flow in the thin molten liquid
layer is laminar and quasi-steady, (6) the solid bulk motion is
Ri quasi-steady, (7) shear stresses on the solid bulk at the fin are
Liquid
α neglected. Also, we assume that there is no melting from the upper
fin and the tube itself, and that the molten PCM flows upwards
φ Solid along the fin, accordingly the pressure gradient is only in the radial
Slightly heated direction. The momentum and energy equations for the thin liquid
envelope layer adjacent to the vertical fin are given in polar coordinates, r
Center of
mass Tenv=const and h, as follows, respectively:
Xc !
Thin molten 1 @ 2 ur dp
layer g ll ¼ ð5Þ
r2 @h2 dr
Longitudinal
fin Tw(r) Menv @2T
¼0 ð6Þ
@h2
where ll is the dynamic viscosity of liquid, p is pressure, ur is
Fig. 6. Schematic diagram of the model for the lower part of the unit. velocity in the r-direction. Integration of Eqs. (5) and (6) with the
following boundary conditions:
u ¼ 0; T ¼ T w ðrÞ@h ¼ 0
3.1. Full model ð7Þ
u ¼ 0; T ¼ T m @h ¼ d
We assume that the solid motion may be approximated as rota- yields a solution for the velocity and temperature distribution in the
tion about the axis of the unit, thus allowing for a formulation in liquid layer, respectively, as:
polar coordinates, r and h. We define a as the angular displacement
of the solid, whereas / is the angle between the fins, as shown in r 2 dp 2
ur ¼ ðh  dhÞ ð8Þ
Fig. 6. 2ll dr
The cross-section area of a circular sector, bounded by Ri and Ro,
is given by: h
TðrÞ ¼ T w ðrÞ þ ðT m  T w ðrÞÞ ð9Þ
Z Z d
/a Ro
R2o  R2i
S¼ rdrdh ¼ ð/  aÞ ð1Þ The heat balance equation is given in the present geometry as:
0 Ri 2
Z  Z d
r
1 @TðrÞ
The center of gravity of a circular sector is:  kl  dr ¼ ql L ur rdh ð10Þ
Ro r @h h¼0 0
R R
xdS ydS
Xc ¼ R ; Yc ¼ R ð2Þ where kl is the thermal conductivity of liquid, L latent heat of fusion,
dS dS ql density of liquid, and d liquid-layer angular dimension in radians.
where Note that the temperature along the layer, T(r), is, generally,
non-uniform, due to the temperature distribution in the fin, Tw(r).
x ¼ r sin h; y ¼ r cos h Substitution of Eqs. (8) and (9) into Eq. (10) yields
 T. Rozenfeld et al. / International Journal of Heat and Mass Transfer 86 (2015) 465–477 471

Z r
dr q Lr3 dp 3 yielding the following expression for the moment per unit depth of
kl ðT m  T w ðrÞÞ ¼ l d ð11Þ frictional resistance at the shell:
Ro rd 12ll dr
Z /a
The heat balance equation – essentially the Stefan condition Menv ¼ R2o senv du ð21Þ
reformulated for a cylindrical geometry – is given by neglecting 0
sensible heat in the liquid layer: Although / can be greater than p/2, like in the 3-fin case, the

1 @TðrÞ @a solid can only retreat from the shell above p/2. For this reason,
 kl ¼ qs Lr ð12Þ
r @h h¼0 @t the resistance at that part is neglected. This relation is integrated
numerically and then substituted in Eq. (15), along with the rela-
Substituting the temperature gradient at h = 0, the angular tion for pressure in the vertical molten layer, Eq. (14), for which
dimension of the molten liquid layer is: the condition p = 0@r = Ri is used. The resulting ordinary dif-
kl ðT w ðrÞ  T m Þ ferential equation for a(t) is then integrated numerically with
d¼ ð13Þ a(0) = 0 as the initial condition, yielding the melt fraction at any
@a
qs L r2 instant throughout the process. Note, that the equation for a(t) is
@t
solved simultaneously with a numerical solution for the time-
Substituting d from Eq. (13) into Eq. (11), we obtain a dependent temperature distribution in the fin, Tw(r):
differential equation for the pressure:
@T w @2T w k 2
 4  3   ðqcp ÞAl ¼ kAl  ðT w  T m Þ ð22Þ
@a L ll 2 3 dt dr
2 dr lt
6 qs R r  r5
dp @t kl ql o where only a half of the fin thickness, lt, is used because in reality
¼ ð14Þ
dr ðT w ðrÞ  T m Þ3 melting takes place symmetrically on both sides of the vertical
fin. For this equation, the following boundary conditions are used:
The moment, rather than force, acting on the solid bulk, is now
Tw(r) = Tw0@r = Ri, @Tw(r)/@r = 0@r = Ro, where Tw0 is assumed to be
given by the following relation:
! equal to the HTF temperature. Initially, the fin temperature is taken
Z Ro
R2  R2i R2o uniform and equal to the melting temperature. Note that under the
ðqs  ql Þg o Ri þ ð1  cosð/  aÞÞ þ pðrÞrdr þ M env ¼ 0 assumption that contribution of melting from the shell to the total
3 Ro þ Ri Ri
melting is rather small, the solid displacement towards the shell is
ð15Þ
neglected in the calculation.
where Menv is the resistance per unit depth caused by friction at the
envelope. In order to predict Menv, the secondary close-contact 3.2. Simplified analytical model
melting is taken into account, where a thin molten layer is formed
between the slightly heated shell and the solid PCM. Since the rota- The model can be considerably simplified if the resistance at the
tion of the solid is rather slow, the velocity inside this liquid layer is envelope and the temperature distribution in the fin are neglected,
assumed to have a Poiseuille profile. Then, according to [45] who i.e. Menv = 0 and Tw = const. This leads to an analytical solution
solved close-contact melting of a horizontal cylinder, the pressure which, as shown in [14] for a different case, is indispensable in
distribution in the secondary thin molten layer is: finding the governing dimensionless groups. In addition, the
 3   analytical solution will be used to verify the numerical model.
L   ds 4
penv ðuÞ ¼ 3mR2o q4s cos4 ðuÞ  cos4 ð/  aÞ For the simplified case, substitution of Eq. (13) into Eq. (11)
kl ðT env  T m Þ dt
yields
ð16Þ
 4  3  
where u is an auxiliary angular coordinate relevant to the envelope dp @a cpl
¼ 6 qs ml r 5  r3 R2o ð23Þ
only, and Tenv is the inner temperature of the envelope, estimated dr @t kl  Ste
from the temperature difference between the water tank and the
where the Stefan number is defined as Ste = cpl(Tw  Tm)/L.
PCM, taking into account thermal conductivities and thicknesses
Integration with the boundary condition of p = 0@r = Ri yields
of the shell and the secondary liquid PCM layer.
For the secondary melting, we can formulate the following force  4  3  
@a cpl 3 3
balance on the solid bulk: p ¼ qs ml r 6 þ r 4 R2o  R4i R2o þ R6i ð24Þ
Z @t kl  Ste 2 2
R2o  R2i /a
ðqs  ql Þg ð/  aÞ þ Ro penv ðuÞ cosðuÞdu ¼ 0 ð17Þ Substitution of Eq. (24) into Eq. (15) yields, after some inter-
2 0
mediate steps, the following result:
Substitution of Eq. (16) into Eq. (17) yields the following 2 31=4
ordinary differential equation for s:
2 31=4 da 6 2 2
6 8ðq  ql ÞgðRi þ Ri Ro þ Ro Þð1  cosð/  aÞÞ 7
7
R2o  R2i ¼6  s 3 7 ð25Þ
6 5ð q  q Þg ð a  /Þ 7 dt 4 c pl 5
ds 6 s l
2 7 3 ml q4s ðRi  Ro Þ2 ðRi þ Ro Þ3 ð3R2i þ R2o Þ
¼6  3 7 kl  Ste
dt 4 3 L 3
5
6Ro mqs
4 sin ða  /Þð7 þ 3 cosð2a  2/ÞÞ
kl ðT env  T m Þ The liquid layer angular dimension is:
2 31=4
ð18Þ
Now, the layer thickness is given by: kl ðT w  T m Þ 6 2 2
6 8ðqs  ql ÞgðRi þ Ri Ro þ Ro Þð1  cosð/  aÞÞ 7
7
d¼ 2 6  3 7
kl ðT env  T m Þ Lr 4 c pl 5
denv ¼ ð19Þ 3 ml q4s ðRi  Ro Þ2 ðRi þ Ro Þ3 ð3R2i þ R2o Þ
ds kl  Ste
Lqs cos u
dt ð26Þ
The shear stress exerted on the solid is given by: The heat flux is given by:
denv dpenv 1 @T Tw  Tm
senv ¼  ð20Þ q00 ¼ kl ¼ kl ð27Þ
2Ro du r @h dr
472 T. Rozenfeld et al. / International Journal of Heat and Mass Transfer 86 (2015) 465–477

" #1=4
where dr = dr approximates the local liquid layer thickness, assum- Ste1=4
2 3
3q ðR  1Þ ðR þ 1Þ ð3R2 þ 1Þ
ing that d = sind for small values of d. This approximation appears to d¼ ð38Þ
be valid for the most of the melting process, both spatially and
r 2 =R2o 8ArPrðR2 þ R þ 1Þð1  cosð/  aÞÞ
temporarily. Note that the liquid layer thickness is not uniform in
the present case. " #1=4
Substituting Eq. (13) into Eq. (27) yields, r=Ro 8ArPrðR2 þ R þ 1Þð1  cosð/  aÞÞ
Nu ¼ 2 3
ð39Þ
@a Ste1=4 3q ðR  1Þ ðR þ 1Þ ð3R2 þ 1Þ
q00 ¼ qs L r ð28Þ
@t Eq. (39) shows that local heat transfer to the PCM, expressed in
The Nusselt number is defined as: terms of the Nusselt number, depends on a number of parameters,
@a
qs L r Ro including the combination FoSte3/4, via a, and in addition also on
q00 Ro @t
Nu ¼ ¼ ð29Þ Ste1/4 and the normalized location along the fin, r/Ro.
ðT w  T m Þ kl ðT w  T m Þ kl
Fig. 7a presents a comparison between the analytical model and
and finally the numerical one, where the latter is used for a simplified case
2 31=4 Tw = const, Menv = 0. The fin temperature, set uniform and equal
to the HTF temperature, is 20 °C above the melting temperature
qs Lr R0 6 2 2
6 8ðqs  ql ÞgðRi þ Ri Ro þ Ro Þð1  cosð/  aÞÞ 7
7 of the PCM. One can see that a very good agreement is reached.
Nu ¼ 6  3 7
ðT w  T m Þ kl 4 c pl 5 Fig. 7(b) compares predictions of the full numerical model for dif-
3 ml q4s ðRi  Ro Þ2 ðRi þ Ro Þ3 ð3R2i þ R2o Þ ferent time steps and number of grid cells inside the fin. One can
kl  Ste
see that independence of the solution on these factors is achieved.
ð30Þ
Eq. (25) may be rewritten, using a trigonometric identity, as:
1
2 31=4
0.9
da 6 8ðqs  ql ÞgðR2i þ Ri Ro þ R2o Þ 7
6 7
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
/a ¼ 6  3 7 dt 0.8
2 1=4
sin 2 4 c pl 4 2 3 2 2
5
3 ml qs ðRi  Ro Þ ðRi þ Ro Þ ð3Ri þ Ro Þ
kl  Ste 0.7
ð31Þ
Melt fraction

0.6
Integrating with the initial condition a ¼ 0@t ¼ 0, the following
0.5
solution is obtained in terms of elliptic functions [59]:


 0.4
1 1
EllipticF ðp þ a  /Þ; 2  EllipticF ðp  /Þ; 2
4 4 0.3
2 31=4
0.2
16
6 ðqs  ql ÞgðR2i þ Ri Ro þ R2o Þ 7
7
¼ 6  3 7 t ð32Þ 0.1 Analytical
24 cpl 5
3 ml q4s ðRi  Ro Þ2 ðRi þ Ro Þ3 ð3R2i þ R2o Þ 0
Numerical
kl  Ste
0 100 200 300 400 500 600 700 800
Eqs. (25) and (32) can be written in a dimensionless form, using Time, s
the following dimensionless groups: (a)
3
Ri q  Ro g al t 1
R ¼ ; q ¼ l ; Ar ¼ ð1  q Þ 2 ; FoSte3=4 ¼ Ste3=4 ð33Þ
Ro qs m l R2o 0.9

" #1=4 0.8

da 8ArPrq3 ðR2 þ R þ 1Þð1  cosð/  aÞÞ
¼ 2 3
ð34Þ
dðFoSte3=4 Þ 3ðR  1Þ ðR þ 1Þ ð3R2 þ 1Þ 0.7
Melt fraction

0.6
and finally



1 1 0.5
EllipticF ðp þ a  /Þ; 2  EllipticF ðp  /Þ; 2
4 4 0.4
" #1=4
2 
1 ArPr q3 ðR þ R þ 1Þ 0.3
¼ FoSte3=4 ð35Þ
2 3ðR  1Þ2 ðR þ 1Þ3 ð3R2 þ 1Þ
0.2
Case 1
Eqs. (26) and (30) can also be written in a dimensionless form, Case 2
0.1
q Ste1=4 Case 3
d¼ ð36Þ 0
@a r2 0 100 200 300 400 500 600 700 800 900
Time, s
@ðFoSte3=4 Þ R2o
(b)
@a r
Fig. 7. Verification of the numerical model: (a) comparison of the analytical and
@ðFoSte3=4 Þ Ro numerical model predictions for the simplified case with uniform fin temperature
Nu ¼ ð37Þ
q Ste1=4 and no friction at the envelope and (b) predictions of the full numerical model for
different time steps and number of grid cells inside the fin: Case 1 – Dt = 103 s, 50
Substitution of Eq. (34) into Eqs. (36) and (37) yields: cells. Case 2 – Dt = 103 s, 100 cells. Case 3 – Dt = 0.5  103 s, 100 cells.
 T. Rozenfeld et al. / International Journal of Heat and Mass Transfer 86 (2015) 465–477 473

Thus, one may conclude from Fig. 7 that the numerical approach gravity, whereas the rotational moment created by the gravity itself
developed in the present study is quite reliable. Further validation decreases not only due to the decrease in the solid mass, i.e. abso-
of the numerical model is done in the next section. lutely, but also due to the decrease in the lever of the gravity with
regards to the center of rotation, i.e. relatively. Moreover, at the
very late stage of the process, when above 90% of the PCM has
4. Results and discussion
already melted, the model predicts that the solid would cease to
4.1. Comparison with the experiments rotate at all. Thus, melting there would switch from CCM to ‘‘regu-
lar’’. Finally, a partial addition of the sensible heat brings the predic-
Fig. 8 shows a comparison of the model predictions with experi- tion closer to the experimental results, although this effect is less
mental results, in terms of the calculated and observed melt frac- significant than those of fin temperature and envelope friction.
tion in the lower parts of the unit. The HTF temperature is 20 °C One can see that the agreement between the full numerical
above the melting temperature of the PCM. In order to demonstrate model predictions and experimental observations is very good. It
the effects of various physical factors on the model, four different is worth to note that the model, in all its versions, under-predicts
curves are included: (1) the prediction made using the analytical the melting rate at small melt fractions. This is because in the model
model, see Eq. (32), i.e. the case with a uniform fin temperature, melting occurs only on the vertical fin, whereas in reality it takes
Tw = const, and zero friction resistance at the shell, Menv = 0; (2) place also on the downward side of the inclined fin. Since at the ini-
the prediction made by the numerical model when temperature tial stages the melting is quite similar on the surfaces of various ori-
distribution along the fin is included, Tw(r) – const, but the friction entations, a contribution of the melting on the inclined downward-
resistance is still zero, Menv = 0; (3) the prediction made by the facing surface to the early-stage total melting rate is rather signifi-
numerical model when temperature distribution along the fin and cant. It also appears that in the second part of the melting process,
the friction resistance are included, Tw(r) – const and Menv – 0; the model slightly over-predicts the experimentally observed
and (4) the prediction made by the numerical model when tem-
perature distribution along the fin and the friction resistance are 1
included, Tw(r) – const and Menv – 0, and in addition the latent
heat, L, is replaced with an effective heat, Leff = L + 0.5cpl(Tw0  Tm), 0.9
in order to take into account sensible heating of the PCM in the 0.8
same manner as done in [45].
We note that in general the difference between the analytical 0.7
model and the experiments is similar to that encountered in [14]
Melt fraction

0.6
for radial fins. Its over-prediction in the present case obviously
results from the fact that both the temperature drop along the fin 0.5
and the resistance to rotation, exerted by the shell, are not
accounted for. The numerical model predicts that the temperature 0.4 ΔT=10, Analytical
at the fin tip is about 7 degrees lower than the HTF temperature at 0.3 ΔT=10, Numerical
the early stages of the process, whereas later on this difference ΔT=20, Analytical
diminishes to about 3 degrees. Indeed, Fig. 8 shows that towards 0.2 ΔT=20, Numerical
the end of the melting process, the effect of fin temperature ΔT=30, Analytical
0.1
distribution weakens. The effect of friction at the envelope appears ΔT=30, Numerical
to be of a magnitude similar to that of the fin temperature 0
0 200 400 600 800 1000 1200 1400 1600
distribution. However, this effect becomes more pronounced at
the advanced stages of the process, because the force that pushes Time, s
the remaining solid to the envelope becomes more aligned with (a)
1

1 0.9

0.8

0.8
0.7
Melt fraction
Melt fraction

0.6
0.6
0.5

0.4 ΔT=10, Analytical

0.4
Analytical 0.3 ΔT=10, Numerical
Case 1 ΔT=20, Analytical
0.2 ΔT=20, Numerical
0.2 Case 2
Case 3 0.1 ΔT=30, Analytical
Experimental ΔT=30, Numerical
0 0
0 200 400 600 800 1000 1200 0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05
Time, s FoSte3/4
Fig. 8. Numerical predictions vs. experiment and effect of various physical (b)
phenomena, for the temperature difference of 20 °C and / = 2p/3: Case 1 –
L = 248 kJ/kg, with temperature distribution in the fin. Case 2 – L = 248 kJ/kg, with Fig. 9. Melting progress in the lower part of the unit with / = 2p/3 for various HTF–
temperature distribution in the fin and friction at the envelope. Case 3 – PCM temperature differences: (a) analytical and numerical predictions and (b)
Leff = 272 kJ/kg, with temperature distribution in the fin and friction at the envelope. generalization vs. FoSte3/4.
474 T. Rozenfeld et al. / International Journal of Heat and Mass Transfer 86 (2015) 465–477

melting rate, with the maximum deviation of less than 10%. This values for these cases are taken as 260, 272 and 284 kJ/kg. For the
indicates that some other mechanisms, not accounted for in the sake of comparison, analytical predictions are shown for all three
model, might develop here. For instance, rather than going straight cases. One can see that the analytical model always over-predicts
to the solid PCM by conduction, some heat from the vertical fin is the melting rate, because it includes the highest possible fin tem-
convected into the upper part of the inter-fin space where the hot perature while discarding friction resistance at the envelope. In
fluid approaches the remaining solid from above. This sort of pro- Fig. 9b, the same cases are shown vs. dimensionless time, FoSte3/4,
cesses is analyzed in [58]. following Eqs. (33) and (34). Naturally, the analytical predictions
It is worth to note that the total sinking towards the envelope, as are represented by a single curve. It is remarkable, however, that
predicted by the model, is about 1 mm even for a relatively big mass the numerical results, obtained for different temperatures, coin-
of PCM which corresponds to / = 2p/3, i.e. melting at the envelope cide now and form a single curve, based on the parameter revealed
does not contribute markedly to the overall phase change. This con- by the analytical model. Furthermore, Fig. 10(a) presents the
firms one of our major assumptions, namely, that melting at the results obtained for the fin angle of / = p/3, i.e. the case of a storage
envelope serves to detach the solid PCM from the shell and thus unit with six longitudinal plate fins, which form an ‘‘asterisk’’ con-
to allow its motion toward the heated surface of a vertical fin. figuration. As in the previous case, Y-configuration, here the same
generalization works perfectly, see Fig. 10(b). Two additional
4.2. Parametric study and generalization observations may be made from the results of Figs. 9 and 10: (1)
the difference between the analytical and numerical predictions
The validated full numerical model is now extended to some is bigger for / = p/3 than for / = 2p/3; and (2) the model predicts
representative geometrical and thermal cases. Fig. 9(a) shows the that for p/3 the close-contact melting stops at the melt fraction
melt fraction as a function of time, for the temperature differences, above 0.85. The both effects are due to the relation between the
Tw0  Tm, of 10, 20 and 30 °C. Accordingly, the effective latent heat moments of rotation and friction, as discussed above. We note that

1 1

0.9 0.9

0.8 0.8

0.7 0.7
Melt fraction

0.6
Melt fraction

0.6

0.5 0.5

0.4 ΔT=10, Analytical 0.4

0.3 ΔT=10, Numerical

0.3
ΔT=20, Analytical φ =π /4
0.2 ΔT=20, Numerical 0.2 φ =π /3
0.1 ΔT=30, Analytical φ =π /2
0.1
ΔT=30, Numerical φ =2 π /3
0 0
0 100 200 300 400 500 600 700 800 900 1000 0 100 200 300 400 500 600 700 800 900 1000
Time, s Time, s
(a) (a)
1 1

0.9 0.9

0.8 0.8

0.7 0.7
Melt fraction

Melt fraction

0.6 0.6

0.5 0.5

0.4 ΔT=10, Analytical 0.4

0.3 ΔT=10, Numerical 0.3 φ =π /4

ΔT=20, Analytical φ =π /3
0.2 ΔT=20, Numerical 0.2
φ =π /2
0.1 ΔT=30, Analytical 0.1 φ =2 π /3
ΔT=30, Numerical Eq. (40)
0 0
0 0.005 0.01 0.015 0.02 0.025 0.03 0 0.005 0.01 0.015 0.02 0.025 0.03 0.035
FoSte3/4 FoSte3/4φ -1/2
(b) (b)
Fig. 10. Melting progress in the lower part of the unit with / = p/3 for various HTF– Fig. 11. Generalization of the results for the melting progress in the lower part of
PCM temperature differences: (a) analytical and numerical predictions and (b) the unit: (a) vs. time for the temperature difference of 20 °C and different angles
generalization vs. FoSte3/4. between the fins and (b) completely generalized for practically significant angles.
 T. Rozenfeld et al. / International Journal of Heat and Mass Transfer 86 (2015) 465–477 475

the remaining solid is exactly the same in both cases, as, say, 7% of Fig. 12a presents the numerical results for the local Nusselt
2p/3 are equal to 14% of p/3, meaning that CCM ceases at the same number, defined according to Eq. (29), plotted vs. FoSte3/4 for three
remaining solid phase. different locations along the fin, namely (r  Ri)/(Ro  Ri) = 0.25,
Fig. 11 further explores the effect of the fin angle, /, including 0.5, and 0.75. It is obvious from the figure that at any given instant,
/ = 2p/3 (Y-configuration), / = p/2 (+ configuration), / = p/3 the local Nusselt number increases when r/Ro increases, whereas at
(⁄ configuration) and / = p/4 (eight fins). It appears from any given location along the fin the local Nusselt number decreases
Fig. 11(a) that the curves for different values of / are quite similar. with time. These results are consistent with the behavior of the liq-
Fig. 11(b) shows an additional generalization for the entire range uid layer thickness, dr, which is related to the Nusselt number in a
of /. One can see that the predicted melt fraction is completely manner similar to that predicted by the analytical model, see Eqs.
generalized when the group FoSte3/4/1/2 is used as the indepen- (27), (29) and (39). Namely, dr decreases when r increases, for any
dent variable. Obviously, in a practical 3-fin unit, the angle / instant, but the layer thickness increases with time and thus sup-
should be equal to or greater than p/2, otherwise no close-contact presses the heat transfer. We note that the predicted values of dr
melting could be achieved in the upper V-part, due to the shape of at various locations do not exceed a few tenths of millimeter
the shell which would constrain the solid motion. However, as throughout the most part of the process, which is consistent with
mentioned above, smaller angles, like / = p/3 and / = p/4, are the thin-layer approximation assumed in this study.
highly practical when the number of fins is planned to exceed Fig. 12(b) presents an attempt to generalize the local Nusselt
four. The generalized results for the melt fraction in Fig. 11b may number. As per Eq. (39), two additional dimensionless groups are
be successfully approximated by the following simple relation, also used, Ste1/4 and r/Ro. It should be noted that the analytical model pre-
plotted in the figure: dicts a linear dependence on r/Ro. The real picture is obviously more
 1:9 complicated, in particular because the temperature distribution
MF ¼ 1  1  23:5FoSte3=4 /1=2 ð40Þ

160 90
(r-Ri)/(Ro-Ri)=0.25 φ =π /4
140 (r-Ri)/(Ro-Ri)=0.5 80
φ =π /3
(r-Ri)/(Ro-Ri)=0.75 φ =π /2
120 70
φ =2 π /3
60
100

50
Nuavg
Nu

80
40
60
30

40
20

20 10

0 0
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045

FoSte 3/4 FoSte3/4

(a) (a)
160
(r-Ri)/(Ro-Ri)=0.25 φ =π /4
60
140 (r-Ri)/(Ro-Ri)=0.5 φ =π /3
(r-Ri)/(Ro-Ri)=0.75 φ =π /2
120 50
φ =2 π /3
-7/5

100 Eq. (41)

40
NuSte (r/Ro)

-4/9

80
Nuavgφ
1/4

30
60

20
40

20 10

0 0
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0 0.005 0.01 0.015 0.02 0.025 0.03
3/4
FoSte FoSte3/4φ -3/5
(b) (b)
Fig. 12. Local Nusselt numbers vs. FoSte3/4 in the lower part of the unit with Fig. 13. Time-dependent, spatially-averaged Nusselt numbers, Nuavg vs. FoSte3/4: (a)
/ = 2p/3: (a) at three different locations along the fin and (b) generalized for any for different angles between the fins and (b) completely generalized for practically
relevant r/Ro. significant angles.
476 T. Rozenfeld et al. / International Journal of Heat and Mass Transfer 86 (2015) 465–477

along the fin also depends on r/Ro. One can see that (r/Ro)7/5 allows management systems that utilize fins for heat transfer enhance-
to obtain a fairly successful generalization for the local Nusselt ment. It is demonstrated that the fins should not be considered just
number. as extended heat transfer surfaces: when properly designed and
Fig. 13 analyzes time-dependent, spatially-averaged Nusselt oriented, they cause the material to melt in their proximity
numbers, corresponding to the melt fractions presented in Fig. 11. throughout the entire melting process, thus accelerating the latter
Fig. 13(a) shows Nuavg vs. FoSte3/4 for various fin angles, /. The very significantly. It is our belief that the general trends revealed in
observed trend is the same as for the local Nusselt number: the this work, and especially the dimensionless parameters and rela-
average Nu decreases with time in all cases. We note that the tions established in it, will provide a proper guidance for any future
dependence on / is rather complex and cannot be deduced explic- analysis of similar systems.
itly from the analytical expressions. However, Fig. 13(b) shows that
a further generalization using the angle between the fins is possible, Conflict of interest
and the curves for various cases practically coincide when plotted
as Nuavg/4/9 vs. FoSte3/4/3/5. The result may be approximated by None declared.
a linear dependence, leading to a relation which expresses the
average Nusselt number in terms of the parameters explored in Acknowledgements
the present study:
3=4 This research was supported by a grant from the Ministry of
Nuavg ¼ a/4=9  bFoSte /7=45 ð41Þ
Science & Technology, Israel, in the framework of French-Israeli
where the constants are a = 62 and b = 1450 under the present con- cooperation in renewable energy.
ditions. Although the Nusselt number decreases with time in the This research was also sponsored by the European Union, under
present case, its reduction is rather moderate in comparison with Partnership Agreement INNOSTORAGE-PIRSES-GA-2013-610692
regular melting, where the growing layer of molten material (Use of innovative thermal energy storage for marked energy
impedes heat transfer and reduces the rate of melting markedly. savings and significant lowering of CO2 emissions).

References
5. Conclusion
_ Dinçer, M.A. Rosen, Thermal Energy Storage: Systems and Applications,
[1] I.
In the present study, melting in a horizontal double-pipe second ed., John Wiley & Sons, Chichester, 2011.
[2] B. Zalba, J.M. Marin, L.F. Cabeza, H. Mehling, Review on thermal energy storage
concentric enclosure with a longitudinally finned inner tube has
with phase change: materials, heat transfer analysis and applications, Appl.
been explored experimentally and via modeling. The investigation Therm. Eng. 23 (2003) 251–283.
has focused on close-contact melting (CCM) – an additional, [3] M. Kenisarin, K. Mahkamov, Solar energy storage using phase change
usually overlooked, mechanism that can contribute significantly materials, Renew. Sustainable Energy Rev. 11 (2007) 1913–1965.
[4] M. Lacroix, Modeling of latent heat storage systems, in: I. _ Dinçer, M.A. Rosen
to the process. CCM was achieved experimentally in a specially (Eds.), Thermal Energy Storage Systems and Applications, John Wiley & Sons,
designed and constructed transparent unit. In the upper part of Chichester, 2002. Chapter 7.
the unit, the solid phase translated vertically, and intensive melt- [5] K.A.R. Ismail, Heat transfer in phase change in simple and complex geometries,
_ Dinçer, M.A. Rosen (Eds.), Thermal Energy Storage Systems and
in: I.
ing occurred on the inclined upward-facing fin surfaces. In the Applications, John Wiley & Sons, Chichester, 2002. Chapter 8.
lower parts of the unit, the solid phase apparently slid on the shell [6] N. Sharifi, T.L. Bergman, A. Faghri, Enhancement of PCM melting in enclosures
and approached the vertical lower fin by rotation. Thus, close- with horizontally-finned internal surfaces, Int. J. Heat Mass Transfer 54 (2011)
4182–4192.
contact melting occurred on a vertical surface – a novel effect [7] Y. Kozak, B. Abramzon, G. Ziskind, Experimental and numerical investigation of
not reported in the literature. It has been demonstrated that CCM a hybrid PCM-air heat sink, Appl. Therm. Eng. 59 (2013) 142–152.
enhances significantly the heat transfer rate, shortening the melt- [8] H. Shabgard, T.L. Bergman, N. Sharifi, A. Faghri, High temperature latent heat
thermal energy storage using heat pipes, Int. J. Heat Mass Transfer 53 (2010)
ing time by about 2.5 times in the specific laboratory-scale device 2979–2988.
explored in the present study. [9] C.W. Robak, T.L. Bergman, A. Faghri, Enhancement of latent heat energy storage
Following the experimental results, a highly novel model was unit using embedded heat pipes, Int. J. Heat Mass Transfer 54 (2011) 3476–
3484.
developed for melting on a vertical surface, approached by a
[10] N. Sharifi, S. Wang, T.L. Bergman, A. Faghri, Heat pipe-assisted melting of a
rotating solid material. The model, which accounts for temperature phase change material, Int. J. Heat Mass Transfer 55 (2012) 3458–3469.
distribution along the fin and includes frictional resistance to [11] K. Nithyanandam, R. Pitchumani, Analysis and optimization of a latent thermal
sliding at the shell, was solved numerically, showing a good agree- energy storage system with embedded heat pipes, Int. J. Heat Mass Transfer 54
(2011) 4596–4610.
ment with the experimental results. A simplified version of the [12] K. Nithyanandam, R. Pitchumani, Thermal energy storage with heat transfer
model was solved analytically, yielding explicit and semi-explicit augmentation using thermosyphons, Int. J. Heat Mass Transfer 67 (2013) 281–
expressions for the time-dependent melt fraction, heat transfer 294.
[13] V. Ho-Kon-Tiat, E. Palomo del Barrio, Recent patents on phase change
rate and molten layer thickness, and revealing the governing materials and systems for latent heat thermal energy storage, Recent Patents
dimensionless groups. It has been found that the melt fraction Mech. Eng. 4 (2011) 16–28.
depends on the Fourier and Stefan numbers combined as FoSte3/4, [14] Y. Kozak, T. Rozenfeld, G. Ziskind, Close-contact melting in vertical annular
enclosures with a non-isothermal base: theoretical modeling and application
whereas the Nusselt number and the normalized layer thickness to thermal storage, Int. J. Heat Mass Transfer 72 (2014) 114–127.
both depend also on the same additional group, Ste1/4, and on the [15] J.C. Choi, S.D. Kim, Heat transfer characteristics of a latent heat storage system
normalized location along the fin, r/Ro. This result, consistent with using MgCl26H2O, Energy 17 (1992) 1153–1164.
[16] K.A.R. Ismail, F.A.M. Lino, Fins and turbulence promoters for heat transfer
our previous findings concerning CCM in an enclosure with radial enhancement in latent heat storage systems, Exp. Therm. Fluid Sci. 35 (2011)
fins [14], was used in a dimensional analysis applied to the results 1010–1018.
of a full numerical model, and then further generalized for different [17] M. Lacroix, Study of the heat transfer behavior of a latent heat thermal energy
storage unit with a finned tube, Int. J. Heat Mass Transfer 36 (1993) 2083–
angles between the inclined fins, obtaining a practically important
2092.
correlation. A similar analysis has been done for the local and [18] A. Erek, Z. Ilken, M.A. Acar, Experimental and numerical investigation of
average time-dependent Nusselt numbers. thermal energy storage with a finned tube, Int. J. Energy Res. 29 (2005) 283–
The present paper provides an additional, experimentally- 301.
[19] W. Ogoh, D. Groulx, Effects of the number and distribution of fins on the
supported evidence of the special role that close-contact melting storage characteristics of a cylindrical heat energy storage system: a numerical
can play in latent-heat thermal energy storage and thermal study, Heat Mass Transfer 48 (2012) 1825–1835.
 T. Rozenfeld et al. / International Journal of Heat and Mass Transfer 86 (2015) 465–477 477

[20] N.H.S. Tay, F. Bruno, M. Belusko, A. Castell, L.F. Cabeza, Experimental validation [38] A.A. Al-Abidi, S. Mat, K. Sopian, M.Y. Sulaiman, A. Mohammad, Internal and
of a CFD model on a vertical finned tube heat exchanger phase change thermal external fin heat transfer enhancement technique for latent heat thermal
energy storage system, in: Proceedings of InnoSTOCK 2012 – the 12th energy storage in triplex tube heat exchangers, Appl. Therm. Eng. 53 (2013)
International Conference on Energy Storage, Lleida, Spain, May 16–18, 2012. 147–156.
[21] J.N.W. Chiu, V. Martin, Submerged finned heat exchanger latent heat storage [39] A.A. Al-Abidi, S. Mat, K. Sopian, M.Y. Sulaiman, A. Mohammad, Experimental
design and its experimental verification, Appl. Energy 93 (2012) 507–516. study of melting and solidification of PCM in a triplex tube heat exchanger
[22] P. Johansson, J.N. Chiu, V. Martin, Impact of convective heat transfer with fins, Energy Buildings 68 (2014) 33–41.
mechanism in latent heat storage modeling, in: Proceedings of InnoSTOCK [40] C. Liu, D. Groulx, Experimental study of the phase change heat transfer inside a
2012 – the 12th International Conference on Energy Storage, Lleida, Spain, horizontal cylindrical latent heat energy storage system, Int. J. Therm. Sci. 82
May 16–18, 2012. (2014) 100–110.
[23] N.H.S. Tay, F. Bruno, M. Belusko, Comparison of pinned and finned tubes in a [41] R.E. Murray, D. Groulx, Experimental study of the phase change and energy
phase change thermal energy storage system using CFD, Appl. Energy 104 characteristics inside a cylindrical latent heat energy storage system: part 1
(2013) 79–86. consecutive charging and discharging, Renew. Energy 62 (2014) 571–581.
[24] N.H.S. Tay, M. Belusko, A. Castell, L.F. Cabeza, F. Bruno, An effectiveness-NTU [42] R.E. Murray, D. Groulx, Experimental study of the phase change and energy
technique for characterising a finned tubes PCM system using a CFD model, characteristics inside a cylindrical latent heat energy storage system: part 2
Appl. Energy 131 (2014) 377–385. simultaneous charging and discharging, Renew. Energy 63 (2014) 724–734.
[25] E. Guelpa, A. Sciacovelli, V. Verda, Entropy generation analysis for the [43] A. Sciacovelli, F. Gagliardi, V. Verda, Maximization of performance of a PCM
design improvement of a latent heat storage system, Energy 53 (2013) 128– latent heat storage system with innovative fins, Appl. Energy 137 (2014) 707–
138. 715.
[26] E.M. Sparrow, E.D. Larson, J.W. Ramsey, Freezing on a finned tube for either [44] V. Shatikian, G. Ziskind, R. Letan, Numerical investigation of a PCM-based heat
conduction-controlled or natural-convection-controlled heat transfer, Int. J. sink with internal fins, Int. J. Heat Mass Transfer 48 (2005) 3689–3706.
Heat Mass Transfer 24 (1981) 273–284. [45] M. Bareiss, H. Beer, An analytical solution of the heat transfer process during
[27] A.G. Bathelt, R. Viskanta, Heat transfer and interface motion during melting melting of an unfixed solid phase change material inside a horizontal tube, Int.
and solidification around a finned heat source/sink, J. Heat Transfer 103 (1981) J. Heat Mass Transfer 27 (1984) 739–746.
720–726. [46] E.M. Sparrow, G.T. Geiger, Melting in a horizontal tube with the solid either
[28] T. Betzel, H. Beer, Experimental investigation of heat transfer during melting constrained or free to fall under gravity, Int. J. Heat Mass Transfer 29 (1986)
around a horizontal tube with and without axial fins, Int. Commun. Heat Mass 1007–1019.
Transfer 13 (1986) 639–649. [47] P.A. Bahrami, T.G. Wang, Analysis of gravity and conduction-driven melting in
[29] P.V. Padmanabhan, M.V. Krishna, Murthy, Outward phase change in a a sphere, J. Heat Transfer 109 (1987) 806–809.
cylindrical annulus with axial fins on the inner tube, Int. J. Heat Mass [48] S.K. Roy, S. Segupta, The melting process within spherical enclosures, J. Heat
Transfer 29 (1986) 1855–1868. Transfer 109 (1987) 460–462.
[30] S.S. Al-jandal, A.A.M. Sayigh, Thermal performance characteristics of STC [49] S.A. Fomin, T.S. Saitoh, Melting of unfixed material in spherical capsule with
system with phase change storage, Renewable Energy 5 (1994) 390–399. non-isothermal wall, Int. J. Heat Mass Transfer 42 (1999) 4197–4205.
[31] K.A.R. Ismail, C.L.F. Alves, M.S. Modesto, Numerical and experimental study of [50] S.H. Emerman, D.L. Turcotte, Stokes problem with melting, Int. J. Heat Mass
the solidification of PCM around a vertical axially finned isothermal cylinder, Transfer 26 (1983) 1625–1630.
Appl. Therm. Eng. 21 (2001) 53–77. [51] M.K. Moallemi, R. Viskanta, Melting around a migrating heat source, J. Heat
[32] G.R. Solomon, R. Velraj, Analysis of the heat transfer mechanisms during Transfer 107 (1985) 451–458.
energy storage in a phase change material filled vertical finned cylindrical unit [52] S.A. Fomin, P.S. Wei, Contact melting by a non-isothermal heating surface of
for free cooling application, Energy Convers. Manage. 75 (2013) 466–473. arbitrary shape, Int. J. Heat Mass Transfer 38 (1995) 3275–3284.
[33] F. Agyenim, P. Eames, M. Smyth, A comparison of heat transfer enhancement [53] T. Hirata, Y. Makino, Analysis of close-contact melting for octadecane and ice
in a medium temperature thermal energy storage heat exchanger using fins, inside isothermally heated horizontal rectangular capsule, Int. J. Heat Mass
Sol. Energy 83 (2009) 1509–1520. Transfer 34 (1991) 3097–3106.
[34] F. Agyenim, N. Hewitt, The development of a finned phase change material [54] M.K. Moallemi, B.W. Webb, R. Viskanta, An experimental and analytical study
(PCM) storage system to take advantage of off-peak electricity tariff for of close-contact melting, J. Heat Transfer 108 (1986) 894–899.
improvement in cost of heat pump operation, Energy Buildings 42 (2010) [55] A. Bejan, Single correlation for theoretical contact melting results in various
1552–1560. geometries, Int. Commun. Heat Mass Transfer 19 (1992) 473–483.
[35] F. Agyenim, P. Eames, M. Smyth, Experimental study on the melting and [56] H. Kumano, A. Saito, S. Okawa, Y. Yamada, Direct contact melting with
solidification behaviour of a medium temperature phase change storage asymmetric load, Int. J. Heat Mass Transfer 48 (2005) 3221–3230.
material (Erythritol) system augmented with fins to power a LiBr/H2O [57] T. Rozenfeld, Y. Kozak, G. Ziskind, Enhanced melting in geometries suitable for
absorption cooling system, Renew. Energy 36 (2011) 108–117. thermal energy storage, in: Proceedings of the 15th International Heat
[36] I. Hamdani, T.M.I. Mahlia, Investigation of melting heat transfer characteristics Transfer Conference, Kyoto, Japan, August 10–15, 2014.
of latent heat thermal storage unit with finned tube, Proc. Eng. 50 (2012) 122– [58] H. Shmueli, G. Ziskind, R. Letan, Melting in a vertical cylindrical tube:
128. numerical investigation and comparison with experiments, Int. J. Heat Mass
[37] S. Mat, A.A. Al-Abidi, K. Sopian, M.Y. Sulaiman, A. Mohammad, Enhance heat Transfer 53 (2010) 4082–4091.
transfer for PCM melting in triplex tube with internal–external fins, Energy [59] E. Jahnke, F. Emde, F. Lösch, Tables of Higher Functions, sixth ed., McGraw-Hill,
Convers. Manage. 74 (2013) 223–236. New York, 1960.