international journal of hydrogen energy 34 (2009) 3417–3424

Available at www.sciencedirect.com

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

Optimized heat transfer fin design for a metal-hydride

hydrogen storage container

M. Melnichuka,c,*, N. Silina,c, H.A. Perettib,c

a
CONICET, Concejo Nacional de Investigaciones Cientı́ficas y Técnicas, Argentina
b
C.N.E.A., Comisión Nacional de Energı́a Atómica, Argentina
c
Instituto Balseiro, Centro Atómico Bariloche (C.N.E.A.), Universidad Nacional de Cuyo, Av. E. Bustillo 9500, C.P. 8400, S.C. de Bariloche,
Rı́o Negro, Argentina

article info abstract

Article history: This paper presents a heat transfer fin optimization for a LaNi5 hydrogen storage container.
Received 30 December 2008 In this simplified approach, a one-dimension fin model is proposed in order to avoid
Received in revised form geometrical restrictions and constrains associated to a particular technological solution.
6 February 2009 Therefore, the presented model can be utilized as a general framework for the develop-
Accepted 9 February 2009 ment of containers with inner fins.
Available online 20 March 2009 The hydrogen absorption reaction was simulated using a developed finite difference
method. On this base, the heat transfer fin design was optimized for different absorption
Keywords: times and container diameters applying the gradient method. Numerical outcomes were
Hydrogen storage then confirmed by comparison with experimental measurements, showing good agree-
LaNi5 hydride container ment between the theoretical model and the experimental result.
Absorption simulation ª 2009 International Association for Hydrogen Energy. Published by Elsevier Ltd. All rights
Heat transfer fin reserved.
Design optimization

1. Introduction phenomena, fluid flow and chemical reaction involved during

the charge–discharge of hydrogen.
Hydrogen is the ideal candidate to replace fossil fuels in the The process of absorption/desorption of hydrogen usually
near future. Considering its massive usage, some technical causes a volumetric expansion/contraction up to 25% and
problems appear in relation to its physical properties. To store involves heat evolution, as the hydride formation/decompo-
it as a gas, high pressures must be achieved due to its low sition is an exothermic/endothermic process. After some
density (0.09 kg m3 at normal conditions), and as a liquid, it charge–discharge cycles the metal hydride, which is usually
requires the use of cryogenic technologies due to its low brittle, becomes a micrometric dust due to the fracture of its
boiling point (20.4 K). These problems might be addressed by grains during volume expansion. In this state, the thermal
the use of metal hydrides since they offer high hydrogen conductivity is very low, usually below 1 W m1 K1 [1].
storage capacity, cycling stability and safety. However, the Consequently in a hydrogen charge or discharge process
usage of hydrides as hydrogen storage media has been a temperature gradient field is generated inside the hydride
precluded by the complexity of the heat and mass transfer container.

* Corresponding author. Centro Atómico Bariloche (C.N.E.A.), Av. Bustillo 9500, C.P. 8400, S. C. de Bariloche, Rio Negro, Argentina.
Tel.: þ54 2944 445286; fax: þ54 2944 445299.
E-mail address: mmelnichuk@cab.cnea.gov.ar (M. Melnichuk).
0360-3199/$ – see front matter ª 2009 International Association for Hydrogen Energy. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijhydene.2009.02.040
3418 international journal of hydrogen energy 34 (2009) 3417–3424

Many efforts have been devoted to the numerical and

Nomenclature experimental characterization of hydride containers. Uchida
et al. [3] evaluated AB2 and AB5 alloys in two different
r radius [m]
container configurations: plate-fin type and divided-chamber
h height [m]
type, both with a heat exchange pipe system. From the results
B diameter
obtained with fin-type device, they concluded that for
k thermal conductivity [W m1 K1]
hydrides of similar plateau pressures, the charging rate
T temperature [K]
depends more on the container thermal performance, rather
r density [kg m3]
than on the characteristics of the hydride. Mosher et al. [4]
N number of calculus domain divisions
assessed a storage system prototype with NaAlH4, using an
V volume [m3]
open-cell aluminum foam to enhance the heat conduction.
t time [s]
They stated that the main disadvantage of the foam is its poor
S000 heat production per unit volume [W m3]
effective conductivity, which is lower in about 1/3 than the
cp heat capacity [J kg1 K1]
conductivity of the material the foam is built from (kfoam z 1/
F aluminum volumetric fraction
3$rrel$ksolid). This drawback is a consequence of the cell struts
A area between control volumes [m2]
random orientation with respect to the heat flow direction.
DH formation enthalpy [J mol1]
Therefore, they examined the optimum heat exchange design
Mw molecular weight [kg mol1]
simulating a finned tube design container by finite elements
3 porosity
method and assuming steady state. Based on these results,
x fraction of maximum capacity
they built a second prototype. Matsevity et al. [5] simulated
C capacity [gH g1
MH]
a storage device containing LaNi5 by the finite difference
R gas constant [R ¼ 8302.5 J mol1 K1]
method. They analyzed three different situations: no heat
E activation energy [J mol1]
exchanger, copper grains mixed with hydride particles, and
P pressure [bar]
copper finned plates. From evaluations of plates’ thickness
DS formation entropy [J mol1 K1]
and spacing, they established that for 0.1 mm fins’ thickness,
g plateau slope
plates spacing should be within 5 mm. Moreover, they con-
Q cost function of the optimization [moles]
structed a prototype with finned plates, and achieved good
Subscripts agreement between numerical and experimental results.
i ith control cell/volume Askri et al. [6] developed a two-dimensional model solved by
eff effective the control-volume-based finite difference method to inves-
H hydrogen tigate the heat and mass transient transport phenomena of
MH metal hydride hydrogen in a LaNi5 storage device. The proposed theoretical
Al aluminum approach showed good concordance with the experimental
Abs absorption measurements.
g gas supply As we can see there are many published numerical and
e equilibrium experimental studies for different configurations of metal
rel relative hydride storage systems, including systems using radial
m mth element of the vector of heuristic values internal fins. However, a systematic optimization of the fin
min minimum profile has not been exhaustively performed yet.
max maximum In this work, we carried out the numerical optimization of
a finned heat transfer system for a cylindrical LaNi5 container
Superscripts
by the gradient method, using the finite difference method in
j time step
one-dimension to simulate it. For this purpose, we considered
c related to heat of formation
the use of aluminum as heat conductive material for the fins
p related to sensible heat
owing to its high conductivity, low cost, low weight and its
þ outer limit of control cell/volume
ease to be machined or conformed to a variety of forms and
 inner limit of control cell/volume
shapes. The numerical optimizations were performed
considering two situations: (i) a container of B ¼ 120 mm at
different hydrogen charging times; (ii) a charging time of
In the absence of a heat managing system, the 3 min for different container diameters.
exothermal absorption process leads to a temperature In order to validate the first group of optimizations, exper-
increment, which raises the equilibrium pressure, so that the imental measurements were performed in a built hydride
absorption rate is reduced. On the other hand, the endo- device coupled with a home-made volumetric Sieverts type
thermal desorption process tends to lower the temperature apparatus. Experiments and numerical calculations were also
of the reaction. Thus, equilibrium pressure decreases and conducted for comparison in a honeycomb type extended
reaction rate is reduced. Since LaNi5 and its derived multi- surface, and with the hydride without heat exchange system.
component alloys show fast reaction kinetics, heat transfer The simplicity of the optimized heat transfer fin model and the
is the main limitation to its dynamical performance [2]. good agreement with the experimental measurements allow
Hence, in order to reduce reaction time, hydride containers the use of the model as a general reference for the design of
must have a heat transfer system. metal-hydride hydrogen storage containers.
 international journal of hydrogen energy 34 (2009) 3417–3424 3419

schematically the simulation domain and the fraction of

2. Numerical procedure aluminum of each cell Fi (0  Fi  1):
The method calculates the temperature value for each cell
2.1. Problem outline and each time step as follows:
 j!
The theoretical model was developed for a LaNi5 container with rcp eff;i Vi Ti j j
radial fins made of high thermal conductivity material. In the þ a þ 000c
i Ti1 þ ai Tiþ1 þ Si Vi
jþ1
Dt
first approach of the model, it was not possible to determine the Ti ¼ rcp j (2)
Vi 
þ 000r
optimal amount of heat conducting material for all the operative
eff ;i
Dt
þ a
i þ ai  Si Vi

conditions. A large amount of the conductive material would

where
enhance the reaction kinetics but at the same time would
restrain the volume fraction occupied by the hydride, leading to 1 keff;i Aþ
a reduction in the container final capacity. On the contrast, aþ
i ¼ R riþ1   z
i
(3)
1=keff;i AðrÞ dr Dr
ri
minimum or no heat exchange material would increase the total
capacity of the container, but would render a slow behavior 000p
S000 000c
i ¼ Si þ Si Ti (4)
performance. Thus, to carry out an optimization of a fin system,
it is necessary to define first the time of reaction and the diam-  
drH  DH  
eter of the container. We investigated fin profiles optimized for S000
i ¼   þ cpMH  cpH Ti (5)
dt i MwH
3 min and 5 min absorption times which are reasonable tenta-
tive charging times for hydrogen containers.
rH;i ¼ xH;i CMH rMH ð1  3Þ (6)
The numerical analysis was simplified by considering
a cylindrical container, with small diameter in comparison We defined a vector that represents the aluminum fraction in
with its length, so that the top and bottom heat losses can be each cell of the domain, named F.
neglected. The main parameters calculated as a function of
the radial position and time were: temperature, absorbed F ¼ ½F1 ; F2 ; .; Fi ; .; FN1 ; FN  (7)
charge, absorption kinetics and heat of reaction. In view of our Then, the effective thermal properties were defined for
system parameters, we applied the modeling equations used a distributed amount of aluminum along the radial direction:
in the work carried out by Laurencelle et al. [7], in which the
absorption and desorption reaction simulations were con- keff;i ¼ kMH ð1  Fi Þð1  3MH Þ þ kAl Fi (8)
ducted by the finite difference method. The authors compared

their numerical and experimental results, obtaining good rcp eff;i ¼ rAl cp;Al Fi þ rMH cp;MH ð1  3MH Þð1  Fi Þ (9)
agreement. The observed discrepancies were ascribed to
In the latter equation, rH cp;H is ignored for being comparatively
limitations in the model of the hydride kinetics, among other
small. We employed a simplified kinetics model as the one
model restrictions.
used by Laurencelle et al. [7], which depends on pressure,
Starting with a completely discharged hydride bed, the
temperature and state of hydrogen charge:
absorption process was simulated at a temperature of 50  C
and under 12.6 bar of hydrogen pressure. j !
Pg  
EAbs
dx RT
j
j
Once we validated the model predictions by comparing the  ¼ kAbs e i log j 1  xH;i (10)
dt i;Abs Pe;i
values obtained with those reported by Laurencelle et al., we
performed an optimization of the heat exchange system by where
the gradient method. For this purpose, a gradient vector was
DSAbs j
calculated by numerical differentiation of a benefit function,  DHj  R þgAbs ðxH;i 0:5Þ
j RT
Pe;i ¼ e i (11)
defined as the amount of absorbed hydrogen. Optimizations
were performed for different diameters and different total We considered that during the absorption process the
hydrogen charging times. container is supplied with hydrogen at a constant pressure Pg.
The hydrogen fraction charge for each cell of the domain as
2.2. Simulation method a function of time is given by Eq. (12):

Heat transfer calculations were performed by the finite

difference method in one-dimension. It was assumed isotropy Calculus domain
in the axial and angular direction, and heat losses through the
axial direction were disregarded. Hence, the heat transfer 1 2 ... i ... N-1 N
equation for cylindrical coordinates resulted in: B.C. 1 r- r+ B.C. 2
    A- A+
1 dT dT
dr kr þ S000 ¼ rcp (1)
r dr dt F1 F2 Fi FN-1 FN
cell
The simulation domain was represented by a vector with N rmin rmax
Heat flow
cells in the radial direction. Boundary conditions were fixed by r
an additional adiabatic cell (B. C. 1), and another additional
cell at constant temperature (B. C. 2). Fig. 1 shows Fig. 1 – Schema of the simulated domain.
3420 international journal of hydrogen energy 34 (2009) 3417–3424

j
j j1 dx Table 2 – Properties of the hydride used in the
xH;i ¼ xH;i þ Dt (12)
dt i;Abs simulations [7].
In Tables 1 and 2 are summarized the physical properties and Property LaNi5
parameters used for the calculations: 1
DHAbs [J mol )] 30,478
DS [J mol1 K1] 108
2.3. Optimization method EAbs [J mol1] 21,170
kAbs [seg1] 59.187
A simple gradient method was used as numerical procedure gAbs 0.13
CMH [wt%] 1.28
for optimization. The cost function defined, represented by
the amount of hydrogen absorbed in the calculus domain, is:

X
N X
N
mMH;i CMH xH;i The vector of heuristic values’ was defined so that the
Q¼ Qi ¼ (13) ~ Max was within the
MwH maximum recalculated cost function Q
i¼1 i¼1
~ ~ ~
uncertainty interval, i.e. Q Max sQ 1 ^Q m [8]. Then the aluminum
where fraction vector for the next optimization step was defined
from the element of vector F~ corresponding to Q ~ Max : F~Max ¼ Fjþ1
mMH;i ¼ rMH ð1  3MH ÞVi ð1  Fi Þ (14)

The elements of the gradient vector were defined as follows:

  3. Experimental procedure
vQ QðF1 ; F2 ; .; Fi þ d; .; FN1 ; FN Þ  Q0
VQi ¼ ¼ (15)
vF i d
3.1. Alloy preparation
where d represented a small increase in the amount of
aluminum (106) and Q0 the unperturbed cost function. In order to prepare LaNi5 alloy, lanthanum (from Beyond
With the purpose of speeding up the convergence, we Chemical Co., purity > 99.95%, impurities: other rare earths)
defined a vector F for the initial fraction of aluminum as and nickel (purity > 99.9%) were melted in an arc furnace
a linear function of the radial position. under an argon atmosphere (purity 99.998%). To ensure
As we mentioned in Section 2.1, there is an amount of
homogeneity, the melting products were turned and re-mel-
aluminum that maximizes the hydrogen absorption, given the
ted three times.
diameter of the container and the absorption time. In order to
determine this amount, we added to the vector F different
3.2. Experimental device
increments of aluminum, named Em.

F~m ¼ F þ Em (16) We built a device for the experimental measurements, con-

sisting of a cylindrical container made of aluminum
These increments were proportional to the normalized (B ¼ 60 mm, h ¼ 10 mm). The bottom and the top were ther-
gradient vector and a multiplying constant from a heuristic mally isolated with a Teflon disc (thickness ¼ 8 mm), so that
values vector (H ): radial heat flow prevailed, according to the model assump-
  tions (Fig. 2).
CFD
Em ¼ VQ Hm (17) The body of the device had an o-ring sealed lid that allowed
Q0
the use of different fin profiles. It also had a 0.5 mm stainless
being steel filter, an on-off valve and a threaded coupling. The
PN device was immersed in a thermostatic bath with forced
Fi
CFD ¼ i¼1
(18) circulation, commanded by a P.I.D. controller at 50  1  C.
N
All the device arrangements allowed good thermal contact
H ¼ ½H1 ; H2 ; .; Hm1 ; Hm  (19) between its body and the fin profile roots in order to accom-
plish the constant temperature border condition B.C.2,
Afterward, the cost function was recalculated for every F~m pointed at Section 2.2.
value:
h i 3.3. Fin profiles fabrication
~¼ Q
Q ~ 1; Q
~ 2 ; .; Q
~ m1 ; Q
~m (20)

We fabricated three different profiles: optimized for 3 min

absorption, optimized for 5 min absorption and honeycomb
type, hereafter called 3 M, 5 M and HC, respectively. Addi-
Table 1 – Physical properties of the materials used in the tionally we evaluated hydride reaction with no heat conduc-
simulation [7]. tion system, hereafter called NF from no fin.
Property LaNi5 Al H2 From the optimization procedure, we determined the
3
fraction of aluminum for each cell of the domain, considering
r [kg m ] 8310 2700 0.0899
that the radial fins were distributed homogenously within the
cp [J kg1 K1] 355 875 14,266
k [W m1 K1] 0.1 231 0.1897
hydride (see Eqs. (8) and (9)). This condition could be achieved
only if fins had an infinitesimal width, so that there were
 international journal of hydrogen energy 34 (2009) 3417–3424 3421

activated by 10 charge–discharge cycles, using z10 bar of high

B.C. 1 B.C. 2
purity hydrogen (99.998%) and vacuum (z0.03 mbar) at room
temperature. Further manipulations were always performed
Device body
8

Teflon inside a glove bag, under high purity argon. After changing
every fin profile, we made at least four charge–discharge
Heat flow
cycles before start measuring.
10

MH + Fins
Absorption capacity measurements were performed for
each fin profile at constant pressure and temperature
8

Teflon
Root (Pg ¼ 12.6  0.1 bar, T ¼ 50  1  C) for different absorption
times. They were quantified by desorption in a volumetric
Sieverts type apparatus.
30 Measurement of the hydride in the container with no fins
r could not be performed in the device described in Section 3.2
because the effective thermal conductivity of the hydride was
similar to the thermal conductivity of the isolating material of
Fig. 2 – Sectional view of the experimental device. the device, invalidating the assumed model conditions. For
this reason we utilized a different reactor for these measure-
ments, consisting of a stainless steel cylinder, with
a minimal distance between hydride and the closest fins. B ¼ 12.7 mm and h ¼ 168 mm, and 2 mm wall thickness, with
Therefore, the maximum number of fins ought to be deter- the same operative conditions as the other experiments.
mined on the basis of available fabrication technology. The porosity value for the simulation was estimated by
With the purpose of neglecting the effect of the distance volume and weight measurements after activating the alloy:
from the hydride bulk to the conductive material, we consid- 3MH ¼ 0.54. This value was consistent with the experimental
ered as a design criterion that the hydride particles inside the measurements of volume and hydride mass within 10%.
device should be at most within 2.3 mm to the closest
aluminum fin [7].
In order to fabricate the profiles properly and at the same 4. Results
time to achieve the distance criterion, we made the profiles
from 0.5 mm thickness aluminum foils, cut by laser Fig. 4 shows the aluminum fraction profiles obtained from the
(minimum width: 0.2  0.05 mm) and placed them staggered optimizations for 3 and 5 min, at a given diameter of 60 mm.
inside the experimental device (Fig. 3). By placing the fins Both results show a larger aluminum fraction near the
staggered, the maximum distances between the fins were external perimeter, r ¼ 30 mm, which is a logical result as
3.24 mm and 3.10 mm for the 3 M and 5 M fin profiles, the perimeter absorbs the heat driven by the temperature
respectively, and 3.24 mm for the HC profile, satisfying the gradients. Maximum aluminum fraction was approximately
distance criterion. 0.15 for the 3 M profile, and 0.1 for the 5 M profile.
Comparing the obtained results, the 3 M profile has
a greater amount of aluminum than the 5 M profile, which
3.4. Assembly and measurement procedure
allows a faster absorption at the beginning, but leads to
a smaller final capacity. This is illustrated in Fig. 5, where it
Cast alloy was sieved, obtaining particles sizes from 125 mm to
450 mm. The device was filled with the alloy and it was
0.16
3M
0.14 5M

0.12

0.10

0.08
F

0.06

0.04

0.02

0.00

0 5 10 15 20 25 30
radial position (mm)

Fig. 3 – Picture of the experimental device with the Fig. 4 – Simulation: aluminum fraction for 3 M and 5 M
staggered aluminum foils of 5 M fin profile. profiles (C [ 1.28 wt%).
3422 international journal of hydrogen energy 34 (2009) 3417–3424

0.7 0.35

0.6 0.30

0.25
0.5

0.20
0.4
n (moles)

F
step 1
0.15
step 4
0.3
0.10 step 8

0.2 step 12
0.05
3M step 20
0.1 0.00
5M
0 10 20 30 40 50 60

0.0 radial position (mm)

0 1 2 3 4 5 6 7 8 9 10
t (min) Fig. 7 – Simulation: convergence of aluminum fraction for
B [ 120 mm optimized at 3 min (C [ 1.28 wt%).
Fig. 5 – Simulation: absorption in time for 3 M and 5 M
(C [ 1.28 wt%).

The optimization results presented so far were calculated

taking into account the hydride capacity reported in Ref. [7],
can be seen that in 3 min the hydrogen absorption capacity which can differ depending on the reference considered. In
reached with the 3 M profile is larger than that achieved with our case, we measured an absorption capacity for LaNi5 of
the 5 M profile. However, after 5 min the situation changes, CMH ¼ 1.00  0.05 wt%. The subsequent simulations were per-
reaching a larger final hydrogen capacity with the 5 M profile. formed using this value to compare them with the experi-
Optimized fin profiles for containers of different diameters mental results.
at an optimization time of 3 min are shown in Fig. 6. It can be In Figs. 8 and 9 are shown the results of the numerical
noticed that a great fraction of the container must be occupied calculations and the experimental measurements of the
with aluminum when ø  60 mm. In the case of B ¼ 240 mm, number of absorbed hydrogen moles as a function of the time
we observed that the aluminum fraction for r > 70 mm did not for 3 M, 5 M and HC profiles. We can see that the fin profiles
vary with the optimizations steps, and that the influence of possess a better performance than HC profile during first
the aluminum fraction for r < 70 mm on the hydrogen stages of absorption as well as for its final capacity. This is due
capacity can be neglected. to the shorter path for the heat from the hydride bulk to the
Fig. 7 depicts the evolution of the vector F with the number container walls, as 3 M and 5 M profiles were built to facilitate
of optimization steps for B ¼ 120 mm. We considered that the the heat flow through the radial direction. The performance of
optimization process was fulfilled when the benefit function HC profile can be compared with the performance of the metal
augmentation was less than 0.01%. foam, since it has no preferential direction for the heat flow.

0.7
5M sim.

0.5 5M exp.
0.6

0.4 0.5
n (moles)

0.3 0.4
F

0.3
0.2 ø = 30 mm

ø = 60 mm
0.2
0.1 ø = 120 mm

ø = 240 mm 0.1
0.0
0 20 40 60 80 100 120 0.0
radial position (mm) 0 1 2 3 4 5 6 7 8 9 10
t (min)
Fig. 6 – Simulation: aluminum fraction for different radial
positions optimized at 3 min (C [ 1.28 wt%). Fig. 8 – Experiment and simulation for 5 M (C [ 1.00 wt%).
 international journal of hydrogen energy 34 (2009) 3417–3424 3423

0.7 3M sim.

3M exp.
0.6
HC sim.

0.5 HC exp.
n (moles)

0.4

0.3

0.2

0.1
Fig. 11 – Simulation: charge evolution of NF.
0.0
0 1 2 3 4 5 6 7 8 9 10
t (min) HC profile occurred mainly at a high hydrogen charge, as it can
be seen in Fig. 12. The reaction of 3 M and 5 M profiles showed
Fig. 9 – Experiment and simulation for 3 M and HC
the same behavior.
(C [ 1.00 wt%).

5. Discussion
The model undervalues the absorbed amount of hydrogen
at the beginning of the process and overvalues this amount at As we have seen, the simulation predictions showed very
the end. This fact is attributed to the model simplifications good agreement with experimental results. Differences
taken for the calculus of the equilibrium pressure, as we will between simulations and experiments were justified mainly
discuss in the following section. Nonetheless, these results due to limitations of the thermodynamic equilibrium pressure
show an excellent agreement between the simulations and model (Eq. (11)) that considers only the plateau region. In the
the experimental data. following a detailed discussion on this and other subjects is
In the case of the container with no fins, the simulation and presented:
the experimental outcomes are shown in Fig. 10. A clear
discrepancy between experiments and simulations can be  Eq. (10) represents the hydrogen absorption rate as a function
observed, because of the model limitations. Given the thermal of temperature, pressure and charge, i.e.: dx=dtfhðTÞf ðPÞgðxÞ,
conditions of this case, the state of charge tends to be close to where f(P) ¼ ln(Pg/Pe). Considering the LaNi5 isotherm as a plot
zero for most of the cells during the 10 min absorption reac- of ln(P) vs. x, Eq. (11) fits Pe with a straight line, which is a good
tion, as it can be seen in Fig. 11. In this situation, the model approximation for the absorption plateau. However, at the
does not predict accurately the behavior of the metal hydride. edges of the plateau, that is to say when the hydride was
On the other hand, the evolution of charge with time for the charged (b phase) or discharged (a phase), the simulated Pe
was, respectively, lower and higher than real Pe. The differ-
ence observed in Fig. 10 (NF) can be explained in terms of the
0.7 predominant low state of charge: this state of charge leads to
NF sim
higher simulated Pe, which is inversely proportional to
0.6 NF exp absorption kinetics. Then the model indicates a slower

0.5
n (moles)

0.4

0.3

0.2

0.1

0.0
0 1 2 3 4 5 6 7 8 9 10
t (min)

Fig. 10 – Experiment and simulation for NF (C [ 1.00 wt%). Fig. 12 – Simulation: charge evolution of HC.
3424 international journal of hydrogen energy 34 (2009) 3417–3424

charging rate than experiment when x z 0. The low absorp- Fin profiles have better performance than HC profile for the
tion corresponding to the first minute shown in Figs. 8 and 9 initial and final hydrogen absorption stages due to the shorter
confirm this behavior. path for the heat from the hydride bulk to the container walls.
Fig. 11 shows that absorption reaction was partially The performance of HC profile can be compared with the
complete after 10 min, so kinetics still had an important performance of the metal foam, since it has no preferential
influence. It can be expected that after a longer period of time, direction for the heat flow.
the simulation would give a result closer to the experiment. In this work we present results of optimized aluminum fin
At the other edge of the plateau, as the reaction was getting profiles for an idealized case. We consider that these results
complete (x z 1), the model predicted a greater absorption can be a useful reference for a hydride container design. Even
than the experiment, as it can be seen in Figs. 8 and 9 at 3 min though, for a prototype design a more detailed two-dimension
and 5 min. simulation is advisable.
 At the beginning of the absorption process, there was
a transient state of Pg caused by the connection of the gas
Acknowledgements
supply, at a regulated pressure, and the experimental device
in vacuum. We could also observe a decrease in the thermal
The authors want to thank the collaboration of Eng. Julián
bath temperature of about 1  C during the fist minutes of
Puszkiel, Lic. Roberto Ausas, and technical personnel of the
absorption, because we used a relatively small water
Centro Atómico Bariloche, so as financial support of CONICET,
volume as thermal ballast. Although these effects had
C.N.E.A., U. N. Cuyo (Proyect 06/C260) and A.N.P.C.yT. (Proyect
a slight influence, we remark them as experimental condi-
P.I.C.T.-R 188 B.I.D. 1728 C/AR).
tions that differed from model assumptions.
 We observed that as the number of charge–discharge cycles
augmented during the course of the experiments, the references
capacity decreased, due to cycling instabilities of the alloy
[9]. At the end of the experimental procedure, we have made
about 80 charge–discharge cycles, so disproportioning [1] Hahne E, Kallweit J. Thermal conductivity of metal hydride
effects had probably occurred. Manipulation effect had materials for storage of hydrogen: experimental
greater influence on 5 M profile results than on the others, investigation. Int J Hydrogen Energy 1998;23(2):107–14.
since these measurements were the last to be performed. [2] Goodell PD, Sandrock GD, Huston L. Kinetics and dynamic
 The numerical model did not consider that the effective aspects of rechargeable metal hydrides. J Less-Common
Metals 1980;73:135–42.
conductivity of the hydride changed during charging due to
[3] Uchida H, Oyakawa K, Watanabe S. Development of a metal
the hydride volumetric expansion [1]. In the case of LaNi5, hydride tank for quick charge. In: Proceedings 14th World
volumetric expansion reaches 25% [10]. Hydrogen Energy Conference 2002, Montreal, Canada.
[4] Mosher DA, Arsenault MS, Tang X, Anton DL. Design,
fabrication and testing of NaAlH4 based hydrogen storage
6. Conclusions systems. J Alloys Comp 2007;446–447:707–12.
[5] Matsevity Yu M, Solovey VV, Chernaya NA, Podgorny AN.
Development and creation of an effective design of a metal-
We optimized the aluminum fraction of heat transfer fins for
hydride hydrogen accumulator. In: Proceedings World
a LaNi5 hydrogen storage container, at different absorption
Hydrogen Technologies Convention 2007, Montecatini
times and container diameters. The absorption capacity of Terme, Italy.
a container of B ¼ 120 mm was evaluated numerically and [6] Askri F, Jemni A, Ben Nasrallah S. Dynamic behavior of
experimentally at three different situations: optimizing at metal-hydrogen reactor during hydriding process. Int J
hydrogen uptake times of 3 min (3 M) and 5 min (5 M), using Hydrogen Energy 2004;29:635–47.
a honeycomb type expanded surface (HC), and utilizing the [7] Laurencelle F, Goyette J. Simulation of heat transfer in
a metal hydride reactor with aluminiun foam. Int J Hydrogen
hydride device without heat exchanger system (NF). A very
Energy 2007;32:2957–64.
good agreement between the simulations and the experi- [8] Taha Hamady A. Investigación de Operaciones. 5ta ed.
mental results were obtained with 3 M, 5 M and HC profiles. Alfaomega; 2004. Section 20.1.1–20.1.2.
However, considerable discrepancies between the theoretical [9] Goodell PD. Stability of rechargeable hydriding alloys during
model and experimental outcomes were noticed in the case of extended cycling. J Less-Common Met 1984;99:1–14.
the hydride device without heat exchange system, based on [10] Ao BY, Chen SX, Jiang GQ. A study on wall stresses induced
by LaNi5 alloy hydrogen absorption–desorption cycles. J
the limitations of the numerical model. These discrepancies
Alloys Comp 2005;390:122–6.
could have been improved by a better adjustment of the
[11] Dhaou H, Askri F, Ben Salah M, Jemni A, Ben Nasrallah S,
kinetics equation, such as a polynomial fitting for Pe [11], and Lamloumi J. Measurement and modeling of kinetics of
taking into account cycling instabilities and effective hydrogen sorption by LaNi5 and two related pseudobinary
conductivity variations during charge. compounds. Int J Hydrogen Energy 2007;32:576–87.