Applied Thermal Engineering 52 (2013) 97e108

Contents lists available at SciVerse ScienceDirect

Applied Thermal Engineering

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

Three-dimensional modeling and sensitivity analysis

of multi-tubular metal hydride reactors
Zewei Bao a, b, Zhen Wu a, Serge Nyallang Nyamsi a, Fusheng Yang a, Zaoxiao Zhang a, b, *
a
State Key Laboratory of Multiphase Flow in Power Engineering, Xi’an Jiaotong University, Xi’an 710049, PR China
b
Suzhou Research Institute, Xi’an Jiaotong University, Suzhou 215123, PR China

h i g h l i g h t s

< A novel three-dimensional model of a metal hydride reactor was proposed.

< In the model the effects of the flow of heat transfer fluid were considered.
< N was defined to assess the effects of neglecting the fluid temperature variation.
< The effect of the contact conductance was discussed in detail.
< Sensitivity analysis for parameters influencing the reactor performance was given.

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

Article history: In order to predict heat and mass transfer characteristics of metal hydride reactors accurately, a novel
Received 7 August 2011 three-dimensional multiphysics model was presented. In the newly established model, the velocity field
Accepted 19 November 2012 of the heat transfer fluid was obtained by solving the NaviereStokes equations or the k-ε turbulence
Available online 24 November 2012
model. The model was numerically solved using the commercial software package COMSOL Multiphysics
V3.5a. Two traditional models were also solved for the reactors under the same set of conditions. A
Keywords:
dimensionless parameter N was defined to assess the effects of neglecting the variation of heat transfer
Metal hydride
fluid temperature on the hydrogen absorption rate. The results from numerical simulation indicated that
Reactor
Parameter analysis
when N is greater than 0.01, the variation of heat transfer fluid temperature cannot be neglected. In this
Three-dimensional model case, the newly established model was valid while the other two models were not. Moreover, it was
Sensitivity analysis found that the effective thermal conductivity of the metal hydride, the flowrate of the heat transfer fluid
and the contact resistance were crucial factors for improving the performance of the metal hydride
reactors.
Ó 2012 Elsevier Ltd. All rights reserved.

1. Introduction Many researchers have established mathematical models to

Since some metal hydride materials, e.g. LaNi5, TiFe and Mg2Ni,
were successfully developed, the applications of metal hydride
have been immensely extended, such as hydrogen storage, heat
pump, thermal energy storage, hydrogen thermal compression,
isotope separation, etc [1,2]. The metal hydride reactors play an
important role in the above systems, in which the metal hydride
formation and decomposition occur. Identifying heat and mass
transfer characteristics of the metal hydride reactors is the basis for
the reactor design and optimization.
* Corresponding author. State Key Laboratory of Multiphase Flow in Power
Engineering, Xi’an Jiaotong University, Xi’an 710049, PR China. Tel./fax: þ86 29
82660689.
E-mail addresses: zhangzx@mail.xjtu.edu.cn, zaoxiaoz@hotmail.com (Z. Zhang).
analyse heat and mass transfer characteristics of metal hydride
reactors. El Osery [3] and Lucas et al. [4] used one-dimensional
mathematical model to describe the adsorption/desorption
process in the metal hydride reactor. Jemni and Ben Nasrallah [5]
formulated a mathematical model for the two-dimensional tran-
sient heat and mass transfer within a metal hydride reactor based
on the volume averaging method. Their results indicated the local
thermal equilibrium assumption was valid and the convective
heat transfer could be neglected. Kuznetsov and Vafai [6] estab-
lished an analytical criteria for the validity of local thermal equi-
librium, steady state and frontal model approximation. Aldas et al.
[7] extended the previous model to three-dimensional model and
studied heat and mass transfer in a metal hydride bed. They found
that hydrogen flow significantly influenced on the temperature
profile, but the overall hydride formation was not affected by

1359-4311/$ e see front matter Ó 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.applthermaleng.2012.11.023
98 Z. Bao et al. / Applied Thermal Engineering 52 (2013) 97e108

Nomenclature U overall heat transfer coefficient, W m2 K1

u X-axis velocity, m s1
A cross-sectional area, m2 V volume, m3
!
Ah heat transfer area, m2 V velocity vector, m s1
b hydride bed thickness, mm v Y-axis velocity, m s1
Ca reaction rate constant, s1 X reacted fraction
cp specific heat, J kg1 K1 w Z-axis velocity, m s1
dH diameter of the hydrogen injection tube, mm DH reaction heat, J mol1
di internal diameter of the heat exchanger tube, mm DX typical heat diffusion length, mm
do external diameter of the heat exchanger tube, mm
dp particle size, mm Greek symbols
Ea activation energy, J mol1 a modification coefficient
hex heat transfer coefficient between the hydride bed and ε volume fraction
the tube wall, W m2 K1 εt turbulence energy dissipation rate, m2 s3
K permeability, m2 h dynamic viscosity, kg m1 s1
k turbulent kinetic energy, m2 s2 l thermal conductivity, W m1 K1
l length, mm n kinematic viscosity, m2 s1
Mg molecular weight, g mol1 r density, kg m3
m _ hydrogen mass absorbed per unit volume and unit
time, kg m3 s1 Subscript
n normal vector 0 initial
Nu Nusselt number e effective value
p pressure, Pa eq equilibrium
Pr Prandtl number f heat transfer fluid
QV volumetric flowrate of heat transfer fluid, m3 s1 g hydrogen
q thermal flux, W m2 in inlet
rh heat transfer controlled reaction rate, mol m3 s1 m mean
Rex thermal contact resistance, mm2K W1 s metal hydride bed
Rg general gas constant, J mol1 K1 sat saturation
Re Reynolds number t turbulence
T temperature, K w heat exchanger tube wall
t time, s

hydrogen flow in the system. The effects of the radiative heat
transfer were studied by Askri et al. [8]. They found that radiative
effects on the sorption process were negligible in the case of the
LaNi5ehydrogen system, but very important for the MgeH2
system. Ha et al. [9] developed a two-dimensional model for
unsteady heat and mass transfer in a metal hydride bed in the
hydriding process. Their results showed that the higher thermal
conductivity, smaller bed diameter and the presence of fins in the
bed gave more enhanced heat transfer rate from the bed. Yang
et al. [10,11] defined two key parameters, heat transfer controlled
reaction rate and mass transfer controlled reaction rate, to analyse
the performance of metal hydride reactors, which established
a two-dimensional mathematical model to prove the validity of
the parameter analysis. Chaise et al. [12] presented a criterion for
quantifying the error which is made when the fluid flow is
neglected. Mellouli et al. [13] developed a two-dimensional
mathematical model to optimize the designs of the metal
hydride storage tanks for fuel cell vehicles. Pourpoint et al. [14]
investigated the thermal characteristics of the Ti1.1MnCr system
using the experimental data and a basic numerical model. Veer-
raju et al. [15] presented a two-dimensional transient model of
plate fin-and-elliptical tube type metal hydride reactors. Visaria
et al. [16] formulated and solved one-dimensional and two-
dimensional computational models for reactors packed with
a high-pressure metal hydride (Ti1.1CrMn). The one-dimensional
model was used for calculating the maximum thickness of metal
hydride layer in the initial design stage. The heat exchanger’s
thermal and kinetic response was obtained by the two-
dimensional computational model in the final design stage.
However, the models in the above literatures were formulated
neglecting the variation of heat transfer fluid temperature [3e16].
MacDonald et al. [17] and Chung et al. [18] used one-dimensional
energy equations to describe the temperature distribution of the
air in the heat exchanger pipe. Freni et al. [19] presented a three-
dimensional model to simulate the whole metal hydride-based
hydrogen storage tank, and the computational domain included
the heat transfer fluid. Krokos et al. [20] developed a detailed
three-dimensional Cartesian model for a multi-tubular metal
hydride tank, which included an energy balance equation for the
cooling fluid. However, the above models are developed based on
the assumption that the axial velocity for the heat transfer fluid is
uniform and constant, but usually this assumption doesn’t coin-
cide with the practical conditions.
In this paper, a novel three-dimensional multiphysics model for
metal hydride reactors was presented, in which the velocity field of
the heat transfer fluid was obtained by solving the NaviereStokes
equations or the k-ε turbulence model. The computational
domain was the whole reactor consisting of the metal hydride bed,
the heat exchanger tube wall and the heat transfer fluid. The
mathematical model was solved by the commercial software
package COMSOL Multiphysics V3.5a. The validity of two assump-
tions was investigated, i.e. the variation of heat transfer fluid
temperature is negligible, as well as that the axial velocity for the
heat transfer fluid is uniform and constant. A dimensionless
parameter N was defined to assess the effects of neglecting the
variation of heat transfer fluid temperature on the hydrogen
absorption processes. The three-dimensional distribution of the
reacted fraction, temperature and gas velocity in the reactors was
obtained, and the effects of some important parameters on the
reactor performance were discussed.
 Z. Bao et al. / Applied Thermal Engineering 52 (2013) 97e108 99

2. Mathematical model where ai are the polynomial coefficients and recommended by

Dhaou et al. [22] and Askri et al. [21].
The reactors under discussion are filled with LaNi5 alloy unless
stated otherwise (MmNi4.6Fe0.4). The following assumptions have 2.2. Heat exchanger devices
been made to formulate the models:
Energy equation for the heat exchanger tube wall:
a) The local thermal equilibrium is valid.
b) The radiative heat transfer among metal hydride particles is vTw
rw cpw ¼ V$ðlw VTw Þ (10)
negligible. vt
c) The thermo-physical properties of the metal hydride do not The physical properties of the heat transfer fluid have little
vary with temperature and concentration. changes during the absorption, so the heat transfer fluid flow is
d) The heat transfer fluid (water) is assumed to be incompressible. assumed to be in steady state. The flow in a circular pipe is laminar
e) The reactor is well insulated (no heat transfer takes place for Re < 2300 or turbulent flow for Re > 10,000 [23].
between the reactor and the surroundings). Continuity equation for the heat transfer fluid:
!
V$ V f ¼ 0 (11)
2.1. Metal hydride bed
The flow of the heat transfer fluid can be described by the
Mass equation for the bed: NaviereStokes equation when Re < 2300:

vrs ! ! h
!
! T i
ð1  εÞ _
¼ m (1) rf V f $V V f ¼ V$ hf V V f þ V V f  Vpf (12)
vt
Continuity equation for hydrogen gas: When Re > 10,000, the k-ε turbulence model is adopted to
describe the flow of the heat transfer fluid:
vrg
! 
ε þ V$ rg V g ¼ m
_ (2) n
h !
! T io
vt ! !
rf V f $V V f ¼ V$ hf þ ht VV f þ VV f  Vpf (13)
The hydrogen velocity vector is determined by the Darcy’s law:
The turbulent kinetic energy:
! K
V g ¼  Vpg (3)  
n !
hg ! ht
rf V f $Vk ¼ V$ hf þ Vk þ ht V V f
sk
KozenyeCarman’s equation is adopted to calculate the perme- h !
! T io
ability K: : VV f þ VV f  r f εt (14)

d2P ε3g The turbulence energy dissipation rate:

K ¼  2 (4)  
150 1  εg ! ht c1 εt ht n !
rf V f $Vεt ¼ V$ hf þ Vεt þ VV f
Energy equation for the bed:
sε k
h !
! T io ε2
vTs !   : VV f þ VV f  c2 rf t (15)
ðrcP Þe þ rg cpg V g $VTs ¼ V$ðle VTs Þ þ m
_ DH=Mg þ mT
_ s cpg  cps k
vt
Turbulent viscosity ht:
(5)
Effective volumetric heat capacity is expressed as follows: k2
ht ¼ rf Cm (16)
  X εt
rcp e
¼ εi ri cpi (6)
where c1, c2, sε, sk and Cm are constants.
Effective thermal conductivity is given below: Energy equation for the heat transfer fluid is as follows:
X
le ¼ εi li (7) vTf
! 

rf cpf þ V$ rf V f Tf ¼ V$ lf VTf (17)
vt
Reaction kinetic equation for hydrogen absorption of LaNi5 or
MmNi4.6Fe0.4 [5,21]:
2.3. Heat transfer coefficient
  !
 Ea pg
m ¼ Ca exp  ln ðrsat  rs Þ (8) The overall heat transfer coefficient comprises the heat transfer
Rg Ts peq
coefficient between the fluid and the heat exchanger tube wall, the
The equilibrium pressures are determined by the P-C-T conductance of the tube wall and the heat transfer coefficient
equation: between the hydride bed and the tube wall.
Heat transfer coefficient between the fluid and the heat transfer
X   
7
DH 1 1 tube wall can be set as:
peq ¼ ai X i exp  for LaNi5 (9a)
Rg Ts 303
i¼0 lf
hif ¼ Nuf (18)
   di
X
9
DH 1 1
peq ¼ ai X i exp  for MmNi4:6 Fe0:4 (9b) At the entrance of circular tubes, the Nuf is expressed for the
Rg Ts 298
i¼0 laminar flow condition as follows [23]:
100 Z. Bao et al. / Applied Thermal Engineering 52 (2013) 97e108

    Table 1
Ref Prf 1=3 hf 0:14 Some major parameters in the model equations [19,21,24e27].
Nuf ¼ 1:86 (19)
l=di hw Parameter Symbol Value
For the turbulent flow condition, Nuf is expressed as [23]: Initial temperature, K T0 293
Inlet temperature, K Tin 293
  2=3   Molecular weight of hydride, kg mol1 MgMH 0.423
di Prf 0:11
Nuf ¼ 0:012ðRe0:87
f  280ÞPr 0:4
f $ 1 þ Molecular weight of hydrogen MgH2 0.002
l Prw kg mol1
Heat transfer coefficient between the hex 1000
(20)
hydride bed and the tube wall,
Overall coefficient of heat transfer is thus calculated as: W m2 K1
LaNi5 MmNi
4.6Fe0.4
1 Initial hydride density, kg m3 r0
U ¼ (21) 8400 8000
1 do do do 1 Saturation hydride density, kg m3 rsat 8527 8110
þ ln þ
hif di 2lw di hex Hydride specific heat, J kg1 K1 cps 419 419
Hydride effective thermal conductivity, le 1.087 1.22
W m1 K1
Reaction heat, J mol1H2 DH 30100 21000
2.4. Initial and boundary conditions Porosity ε 0.5 0.5
Inlet pressure pg,in 0.8 MPa 4.5 MPa
2.4.1. Initial conditions
Initially, the hydrogen pressure, the temperature and the metal
hydride density in the reactor are assumed to be uniform.
transfer fluid is uniform and constant, is named Model 2. The model
Therefore,
consisting of Eqs. (1)e(9) and Eqs. (18)e(26), in which the variation
! of heat transfer fluid temperature is neglected, is named Model 3.
pg ¼ p0 ; Ts ¼ T0 ; V g ¼ 0; rs ¼ rs0 ; Tw ¼ T0 ; Tf ¼ T0
Model 2 and Model 3 are traditional mathematical models for
(22) metal hydrides reactors. Some major physical parameters in the
2.4.2. Boundary conditions model equations are listed in Table 1, and the initial reacted fraction
Adiabatic wall (or symmetry boundary): of metal hydride bed for absorption is 0.1. These three models were
solved using the commercial software package COMSOL Multi-
vTs vpg vTw vTf vpf physics V3.5a. The unstructured tetrahedral mesh was adopted. The
¼ 0; ¼ 0; ¼ 0; ¼ 0; ¼ 0 (23) relative error was 103 and the absolute error was 104.
vn vn vn vn vn
Jemni et al. [26] carried out some experimental studies on
Hydrogen inlet boundary: a cylindrical metal hydride reactor filled with LaNi5 in 1999, and
their results were taken for validation of our models in this paper.
vTs
pg ¼ pg;in ; ¼ 0 (24) In their experiments, the reactor was submerged in a constant
vn
temperature water bath, so the variation of heat transfer fluid
Outer wall of the heat exchanger tube: temperature could be neglected and Model 3 was adopted to
simulate the reactor. The simulated temperature histories are
vpg
¼ 0 (25) shown with the experimental results in Fig. 1. It is indicated that the
vn simulation results agree satisfactorily with the experimental

 results, thus the validity of Model 3 is proved. Compared with
vTs
le ¼ U Ts  Tf (26) Model 3, Model 1 adds NaviereStokes equations and the k-ε
vn turbulence model to simulate the heat transfer fluid flow, and
Model 1 and Model 2 add energy equations to simulate the heat
vTs vTw
le ¼ hex ðTs  Tw Þ ; lw ¼ hex ðTw  Ts Þ (27)
vn vn
Inner wall of the heat exchanger tube:

qw ¼ qf ; Tw ¼ Tf ; uf ¼ vf ¼ wf ¼ 0 (28)

Inlet boundary of the heat transfer fluid:

4QV
uf ¼ vf ¼ 0; wf ¼ uin ¼ ; Tf ¼ Tin (29)
pd2i
Outlet boundary of the heat transfer fluid:


pf ¼ 0; V$ lf VTf ¼ 0 (30)

2.5. Implementation and validation of the model

In this paper, the newly established model named Model 1

consists of Eqs. (1)e(17), Eqs. (22)e(25) and Eqs. (27)e(30). The
model consisting of Eqs. (1)e(10), Eq. (17), Eqs. (22)e(25) and
Eqs. (27)e(30), which assumes that the axial velocity for the heat Fig. 1. The comparison of simulation results and experimental data.
 Z. Bao et al. / Applied Thermal Engineering 52 (2013) 97e108 101

Fig. 2. (a) The schematic of the tank reactor equipped with a central heat exchanger tube; (b) The schematic of the multi-tubular reactor equipped with heat exchanger tubes and
hydrogen injection tubes; (c) The reactor geometry implemented in COMSOL Multiphysics simulation environment.

transfer processes of the heat exchanger tube wall and the heat absorption rate can be estimated by the heat transfer controlled
transfer fluid. Similar simulations for heat transfer and fluid flow in reaction rate defined by Yang et al. [11]:
the heat exchanger devices have been performed using COMSOL
Multiphysics and validated [28,29]. Therefore, the validity of the
models for the metal hydride beds, the heat exchanger tube walls


and the heat transfer fluid in Model 1 and Model 2 have been
Ts  Tf Ah
proved, so the validity of Model 1 and Model 2 is also proved. rh ¼ (31)
DX 1 DH$V
þ
3. Parameter analysis ale U

3.1. Parameter definition

In this study, rh0 is defined as the heat transfer controlled reac-
For a given metal hydride reactor, the reaction rate is generally tion rate when the temperature of the heat transfer fluid increases
controlled by heat transfer processes [2]. Therefore, the hydrogen to Tf0 , then:
102 Z. Bao et al. / Applied Thermal Engineering 52 (2013) 97e108

Fig. 3. Time evolution of the simulated average reacted fraction obtained by the three models for different values of N and different reactor geometries: (a) N ¼ 0.1, RAT; (b) N ¼ 0.01,
RAT; (c) N ¼ 0.001, RAT; (d) N ¼ 0.1, RMT; (e) N ¼ 0.01, RMT; (f) N ¼ 0.001, RMT.



Ts  Tf0 Ah Ts  Tf
Ah
rh0 ¼ (32) Tf0  Tf ¼   (36)
DX 1 DH$V DX 1
þ rf QV cpf þ
ale U ah le U
A dimensionless number N is defined to estimate the error when where Tf0  Tf quantifies the temperature variation of the heat
neglecting the variation of heat transfer fluid temperature, and can transfer fluid inside the heat exchanger channel.
be written: Substituting Eq. (36) into Eq. (34), we obtain:

A
rh  rh0 N ¼ h  (37)
N ¼ (33) DX 1
rh rf QV cpf þ
ah le U
substituting Eqs. (31) and (32) into Eq. (33), we get: When N/0, the variation of heat transfer fluid temperature can
be neglected. Otherwise, when N >> 0, the energy equation for the
heat transfer fluid should be incorporated into the mathematical
Tf0  Tf model.
N ¼ (34)
Ts  Tf
3.2. Results of the parameter analysis
According to energy conservation principle, we can obtain:
The validity of the dimensionless number N is tested on the tank

 reactor geometries, which are commonly used in various reactor
rf QV cpf Tf0  Tf ¼ rh V DH (35) studies [2]. Fig. 2a shows a cylindrical tank reactor equipped with
a central heat exchanger tube (RAT for short), in which hydrogen is
substituting Eq. (31) into Eq. (35), and the following equation is fed in the radial direction. A multi-tubular reactor equipped with
derived: five heat exchanger tubes and four hydrogen injection tubes (RMT

Fig. 4. Time evolution of the simulated average reacted fraction of RAT obtained by the three models for different values of N: (a) N ¼ 0.1; (b) N ¼ 0.01; (c) N ¼ 0.001.
 Z. Bao et al. / Applied Thermal Engineering 52 (2013) 97e108 103

Fig. 5. Time evolution of the simulated outlet mean temperature of the heat transfer fluid obtained by the three models for (a) RAT and (b) RMT.

for short) is shown in Fig. 2b. In RMT, the internal and external
diameters of the heat exchanger tube are 8 mm and 10 mm,
respectively, and the diameter of the hydrogen injection tube is
4 mm. For the reasons of symmetry, the computational domain of
RMT is one eighth of the whole reactor, which is depicted in Fig. 2c.
For each reactor, QV is gradually varied and the other
parameters are kept constant. The values of QV have been
selected to produce values of N equal to 0.1, 0.01 and 0.001,
respectively. Grid independence test was carried out by analysing
the effect of the grid number on the time evolution of the average
temperature of the metal hydride bed. The numbers of mesh
elements in the three models for RAT were 65,282, 65,282 and
7784, respectively, and they were respectively 15,810, 15,810 and
9402, for RMT.
The time evolution of the simulated average reacted fraction is
presented in Fig. 3. Fig. 3a and d show that the differences in the
simulated average reacted fraction obtained by the three models
are very evident when N is 0.1. Big differences in the simulated
average reacted fraction between Model 1 and Model 3 are
observed, which indicates that the variation of heat transfer fluid
temperature cannot be neglected in this case. The differences
between the simulation results of Model 1 and those of Model 2 are
more significant. This is because the boundary layer effect is
neglected in Model 2, thus the heat transfer efficiency between the
heat transfer fluid and the hydride bed is actually exaggerated. It
needs to be noticed that the differences in the average reacted
fraction between Model 1 and Model 2 are larger than those
between Model 1 and Model 3 although the temperature variation

Fig. 6. Distribution of the reacted fraction in the metal hydride bed: (a) t ¼ 100 s; (b) t ¼ 1000 s; (c) t ¼ 2000 s.
104 Z. Bao et al. / Applied Thermal Engineering 52 (2013) 97e108
Fig. 7. Distribution of the temperature in the metal hydride bed: (a) t ¼ 100 s; (b) t ¼ 1000 s; (c) t ¼ 2000 s; (d) t ¼ 3000 s.
of the heat transfer fluid is taken into account in Model 2. This
phenomenon indicates the errors caused from the assumption that
axial velocity for the heat transfer fluid is uniform and constant are
greater than the one that the temperature variation of heat transfer
fluid is neglected in the three-dimensional simulation. Fig. 3b and e
present the simulation results when N is 0.01. It can be found that
Model 1 and Model 3 almost give the same results. The differences
between Model 1 and Model 2 are less remarkable compared with
the previous case, but still can be observed. Fig. 3 c and f present the
simulation results when N is 0.001. For these two reactors, the three
models almost give the same results. Obviously, the differences
among the simulation results of the three models decrease as the
parameter N decreases. In the case that N is 0.01, the computational
time for simulating the RMT using Model 1, Model 2 and Model 3
was respectively about 410, 390 and 290 s on a personal computer
with 12 GB RAM and a quad-core 2.8 GHz CPU.
It can be found that when N is smaller than 0.01, no difference is
visible between the results obtained by Model 1 and those obtained
by Model 3, in which the temperature variation is neglected.
Therefore, 0.01 is a critical value of N to determine whether the
temperature variation of the heat transfer fluid can be neglected in
the mathematical model.
To check whether the critical value of the parameter N will
change significantly for different metal hydride materials or
different design parameters of reactors, we consider RAT packed
with MmNi4.6Fe0.4 at different values of N. The hydride bed thick-
ness b and the bed length l are changed to 100 mm and 5 mm,
respectively, and other design parameters are kept constant. The
values of QV have been selected to produce values of N equal to 0.1,
0.01 and 0.001, respectively. From Fig. 4, it can be found that the
critical value of N still remains around 0.01.
The mean temperature of the heat transfer fluid is as follows [23]:
Z
cPf rf Vf Tf dA
A
Tm ¼ Z (38)
cPf rf Vf dA
A
 Z. Bao et al. / Applied Thermal Engineering 52 (2013) 97e108 105

Fig. 8. Distribution of the velocity in the metal hydride bed: (a) t ¼ 100 s; (b) t ¼ 1000 s; (c) t ¼ 2000 s; (d) t ¼ 3000 s.

Fig. 9. Distribution of the temperature of the heat transfer fluid and the heat exchanger tube.
106 Z. Bao et al. / Applied Thermal Engineering 52 (2013) 97e108
Fig. 10. Simulated results at different effective thermal conductivities: (a) The average reacted fraction; (b) The outlet mean temperature of the heat transfer fluid.
The variations of the outlet mean temperature of the heat
transfer fluid when N is 0.01 are presented in Fig. 5. There is a sharp
increase in outlet mean temperature in the initial stage, and
afterwards the temperature increases gradually until reaching the
maximum mean temperature, because a lot of reaction heat is
released during the initial stage of the absorption reaction. Later,
the outlet mean temperature decreases due to decline of the
reaction rate. From Fig. 5, big differences between the simulated
mean temperatures from Model 1 and Model 2 are observed when
N is 0.1. The highest temperature value obtained by Model 1 is
lower than that obtained by Model 2 because the boundary layer
effect is neglected in Model 2. When N is 0.01 or 0.001, the corre-
sponding flowrate of the heat transfer fluid is very large, so the
temperature variation of heat transfer fluid in these two cases can
be neglected, which will not be presented in this paper. Model 2 is
not recommended to simulate the metal hydride reactors because it
cannot give more accurate results and needs longer computing
time compared with Model 3.
4. Results of the mathematical model
4.1. Reacted fraction, temperature and gas velocity distributions
The three-dimensional distribution of the reacted fraction,
temperature and gas velocity in RAT obtained by Model 1 is pre-
sented for the flowrate of 3.1  106 m3 s1. The pressure variations
in the reactor are insignificant and not presented. The maximum
pressure drop is around 0.001 MPa. Fig. 6 shows the three-
dimensional distribution of the reacted fraction in the metal
hydride bed at three representative moments, t ¼ 100, 1000 and
Fig. 11. Simulated results at different tube heat transfer coefficients: (a) The average reacted fraction; (b) The outlet mean temperature of the heat transfer fluid.
2000 s. The reacted fraction is higher near the heat exchanger tube
wall, where the bed temperature is lower. This indicates that the
heat transfer controls the absorption process. At 1000 s, the
absorption reaction near the tube wall is almost completed, but the
reacted fraction of metal hydride far away from the wall is still very
low. At 2000 s, the absorption reaction in the whole reactor is
almost completed.
Fig. 7 shows the three-dimensional distribution of the temper-
ature in the metal hydride bed at four representative moments,
t ¼ 100, 1000, 2000 and 3000 s. Due to the exothermic effect of the
hydrogen absorption process, the temperature of the hydride bed
increases rapidly. Moreover, it can be seen that the temperature of
the metal hydride bed near heat exchanger tubes is lower than that
of the rest regions, as far as the same cross section is concerned. It is
also found that the average bed temperature in a cross-section
increases along the flow direction, which is due to the descend-
ing cooling effect by the fluid.
Fig. 8 shows the three-dimensional distribution of the hydrogen
velocity in the metal hydride bed at four representative moments,
t ¼ 100, 1000, 2000 and 3000 s. At 100 s, the velocity of the upstream
region is higher than that of downstream region, because the reac-
tion is fast near the heat transfer fluid inlet due to better cooling. On
the contrary, at 3000 s, the velocity of the upstream region is lower
than that of downstream region. It is because that after the initial
rapid absorption near the fluid inlet, the metal hydride is almost
saturated there and the resulting hydrogen flow weakened.
Fig. 9 shows the three-dimensional distribution of the temper-
ature of the heat transfer fluid at four representative moments,
t ¼ 100, 1000, 2000 and 3000 s. The temperature rises along the
axial direction because the reaction heat is continually transferred
 Z. Bao et al. / Applied Thermal Engineering 52 (2013) 97e108 107

to the heat transfer fluid. At 100 s, the outlet temperature distri-

bution is not uniform and the region near the heat exchanger tube
wall is much warmer than the core region, which is due to the low
thermal conductivity of the heat transfer fluid (water). At 3000 s,
the absorption process almost ceases and the fluid temperature
hardly experiences any changes from inlet to outlet.

4.2. Process intensification

The effective thermal conductivity of the hydride bed is gener-

ally very low, but it can be improved by mixing metal hydride
powder with heat conduction matrices (the most popular one is Al
foam) or making metal hydride powder into compacts [30,31].
When the volume fractions of LaNi5 and Al foam are respectively
0.462 and 0.1 in the metal hydride bed, the effective thermal
conductivity of the hydride bed can reach 7.5 W m1 K1 and the
permeability will decrease to 5.8  1013 m2 [30]. Enhancement of Fig. 12. Sensitivity analysis of the main parameters influencing the absorption
heat transfer by increasing the effective thermal conductivity of the completion time.
metal hydride bed was examined about its effect, on both the
average reacted fraction and the outlet mean temperature of the
Fig. 12 shows the influence of the 4 parameters on the absorp-
heat transfer fluid. Fig. 10a presents the former effect. The increase
tion completion time. The absorption completion time is defined as
of the effective thermal conductivity improves the heat transfer
the time interval for completing 90% of the hydriding reaction.
performance and accelerates the absorption process. Fig. 10b
Fig. 12 shows that the absorption completion time is most sensitive
presents the latter effect. The outlet mean temperature of the heat
to the effective thermal conductivity of the metal hydride. When
transfer fluid increases sharply in the initial stage and reaches its
the effective thermal conductivity reaches 7.5 W m1 K1, the
maximum value. When the effective thermal conductivities are
absorption completion time is 1000 s. However, the measure for
1.087 and 7.5 W m1 K1, the maximum outlet mean temperature
enhancing the effective thermal conductivity reduces the hydrogen
values of the heat transfer fluid will be 302.0 and 303.9 K, respec-
storage capacity per unit mass and increases the total thermal
tively. Subsequently, the outlet temperature slowly decreases
capacity of the reactor. The flowrate of the heat transfer fluid is the
approaching the temperature of the inlet heat transfer fluid. When
second most effective mean to reduce the absorption completion
the effective thermal conductivity reaches 7.5 W m1 K1, the
time. Moreover, the flowrate of the heat transfer fluid is an oper-
outlet mean temperature of the heat transfer fluid returns to the
ating parameter, which is easy to control. Contact resistance is the
initial temperature quickly within 1960 s. However, it takes about
third parameter that has an appreciable influence on the absorp-
3270 s as the bed thermal conductivities decrease down to
tion completion time. Generally contact resistance is greater than
1.087 W m1 K1.
1000 mm2 K W1. From this figure, when contact resistance is
The thermal contact resistance between the hydride bed and the
2500 mm2 K W1, the absorption completion time is 1950 s.
heat exchanger tube wall, which is the reciprocal of hex, is another
In summary, the sensitivity sequence of the various parameters
critical factor to the reactor performance when the heat transfer
on the absorption completion time is: the effective thermal
fluid is water. Fig. 11 presents the effect of hex on the average reacted
conductivity of the metal hydride > the flowrate of the heat
fraction and the outlet mean temperature of the heat transfer fluid,
transfer fluid > contact resistance > the heat exchanger tube
respectively. The increase of hex accelerates the absorption process,
material.
as can be seen in Fig. 11a. When the tube heat transfer coefficients
are 500 mm2 K W1, 1000 mm2 K W1 and 2000 mm2 K W1, the
maximum mean temperature values are 308.2, 306.8 and 304.7 K,
5. Conclusions
respectively. It is observed from Fig.11b that when the contact
conductance attains a low value of 500 mm2 K W1, the heat
In this study, a novel three-dimensional model and two tradi-
transfer fluid reaches the peak temperature and returns to the
tional models for metal hydride reactors were solved using the
initial temperature more rapidly. This is because the heat transfer
commercial software package COMSOL Multiphysics V3.5a. The
coefficient between the hydride bed and the heat transfer fluid
performance of metal hydride reactors during the absorption
increases. Therefore, the increase of hex is favourable for the reactor
process was simulated by the three models. Moreover, a dimen-
performance.
sionless parameter N was defined to assess the effects of neglecting
the variation of heat transfer fluid temperature on the hydrogen
4.3. Sensitivity analysis absorption rate. The following conclusions can be drawn:

The effective thermal conductivity of the metal hydride and
contact resistance between the hydride bed and the tube wall are
two key properties of metal hydride reactors. The effective thermal
conductivity of Al-foams with metal hydride powder can reach
7.5 W m1 K1 [30]. According to research of Muthukumar and
Groll [24], contact resistance can decrease to 1000 mm2 K W1. The
increase of the flowrate of heat transfer fluid can improve the heat
transfer performance of the reactor. The heat exchanger tube of the
reactor under above discussion is made of stainless steel 316. When
it is made of brass, lr will increase up to 109 W m1 K1 (at 293 K).
1) The dimensionless parameter N can be used to estimate the
errors caused from the assumption that the variation of heat
transfer fluid temperature is neglected for the metal hydride
reactors. The differences among the simulation results of the
three models decrease as the parameter N decreases.
2) When N is greater than 0.01, the variation of the heat transfer
fluid temperature cannot be neglected. In this case, the newly
established model (Model 1) was valid while the two tradi-
tional models (Model 2 and Model 3) were not. Model 2 is not
recommended to simulate the metal hydride reactors because
108 Z. Bao et al. / Applied Thermal Engineering 52 (2013) 97e108
it cannot give more accurate results and needs longer compu-
tational time compared with Model 3,
3) The most effective methods of reducing the absorption
completion time were enhancing the effective thermal
conductivity of the hydride bed, increasing the flowrate of the
heat transfer fluid, and reducing the contact conductance
between the hydride bed and the heat exchanger tube wall.
Acknowledgements
The financial support from the National Natural Science Foun-
dation of China (No. 51106118), the Specialized Research Fund for
the Doctoral Program of Higher Education (No. 20100201110007)
and the Application Fundamentals Research Program of Suzhou city
(No. SYG201019) is greatly acknowledged. The authors would like
to thank Dr. Xiangyu Meng for his help in the writing.
References
[1] W. Wongsuwan, S. Kumar, P. Neveu, F. Meunier, A review of chemical heat
pump technology and applications, Appl. Therm. Eng. 21 (2001) 1489e1519.
[2] F.S. Yang, G.X. Wang, Z.X. Zhang, X.Y. Meng, V. Rudolph, Design of the metal
hydride reactors e a review on the key technical issues, Int. J. Hydrogen
Energy 35 (2010) 3832e3840.
[3] I.A. El Osery, Theory of the computer code ret1 for the calculation of space-
time dependent temperature and composition properties of metal hydride
hydrogen storage beds, Int. J. Hydrogen Energy 8 (1983) 191e198.
[4] G.G. Lucas, W.L. Richards, Mathematical modeling of hydrogen storage
systems, Int. J. Hydrogen Energy 9 (1984) 225e231.
[5] A. Jemni, S. Ben Nasrallah, Study of two-dimensional heat and mass transfer
during absorption in a metal-hydrogen reactor, Int. J. Hydrogen Energy 20
(1995) 43e52.
[6] A.V. Kuznetsov, K. Vafai, Analytical comparison and criteria for heat and mass
transfer models in metal hydride packed beds, Int. J. Heat Mass Tran. 38
(1995) 2873e2884.
[7] K. Aldas, M.D. Mat, Y.A. Kaplan, A three-dimensional mathematical model for
absorption in a metal hydride bed, Int. J. Hydrogen Energy 27 (2002) 1049e
1056.
[8] F. Askri, A. Jemni, S. Ben Nasrallah, Study of two-dimensional and dynamic
heat and mass transfer in a metalehydrogen reactor, Int. J. Hydrogen Energy
28 (2003) 537e557.
[9] M.Y. Ha, I.K. Kim, H.D. Song, S. Sung, D.H. Lee, A numerical study of thermo-
fluid phenomena in metal hydride beds in the hydriding process, Int. J. Heat
Mass Tran. 47 (2004) 2901e2912.
[10] F.S. Yang, X.Y. Meng, J.Q. Deng, Y.Q. Wang, Z.X. Zhang, Identifying heat and
mass transfer characteristics of metal hydride reactor during adsorption e
parameter analysis and numerical study, Int. J. Hydrogen Energy 33 (2008)
1014e1022.
[11] F.S. Yang, X.Y. Meng, J.Q. Deng, Y.Q. Wang, Z.X. Zhang, Identifying heat and
mass transfer characteristics of metal hydride reactor during adsorption e
improved formulation and further discussion about parameter analysis, Int. J.
Hydrogen Energy 34 (2009) 1852e1861.
[12] A. Chaise, P. Marty, P. De Rango, D. Fruchart, A simple criterion for estimating
the effect of pressure gradients during hydrogen absorption in a hydride
reactor, Int. J. Heat Mass Tran. 52 (2009) 4564e4572.
[13] S. Mellouli, F. Askri, H. Dhaou, A. Jemni, S. Ben Nasrallah, Numerical simulation
of heat and mass transfer in metal hydride hydrogen storage tanks for fuel cell
vehicles, Int. J. Hydrogen Energy 35 (2010) 1693e1705.
[14] T.L. Pourpoint, V. Velagapudi, I. Mudawar, Y. Zheng, T.S. Fisher, Active cooling
of a metal hydride system for hydrogen storage, Int. J. Heat Mass Tran. 53
(2010) 1326e1332.
[15] C. Veerraju, M. Ram Gopal, Heat and mass transfer studies on plate fin-and-
elliptical tube type metal hydride reactors, Appl. Therm. Eng. 30 (2010)
673e682.
[16] M. Visaria, I. Mudawa, T. Pourpoint, Enhanced heat exchanger design for
hydrogen storage using high-pressure metal hydride: part 1. Design meth-
odology and computational results, Int. J. Heat Mass Tran. 54 (2011) 413e423.
[17] B.D. MacDonald, A.M. Rowe, A thermally coupled metal hydride hydrogen
storage and fuel cell system, J. Power Sources 20 (2006) 346e355.
[18] C.A. Chung, C.S. Lin, Prediction of hydrogen desorption performance of Mg2Ni
hydride reactors, Int. J. Hydrogen Energy 34 (2009) 9409e9423.
[19] A. Freni, F. Cipitì, G. Cacciola, Finite element-based simulation of a metal
hydride-based hydrogen storage tank, Int. J. Hydrogen Energy 34 (2009)
8574e8582.
[20] C.A. Krokos, D. Nikolic, E.S. Kikkinides, M.C. Georgiadis, A.K. Stubos, Modeling
and optimization of multi-tubular metal hydride beds for efficient hydrogen
storage, Int. J. Hydrogen Energy 34 (2009) 9128e9140.
[21] F. Askri, M. Ben Salah, A. Jemni, S. Ben Nasrallah, Heat and mass transfer
studies on metal-hydrogen reactor filled with MmNi4.6Fe0.4, Int. J. Hydrogen
Energy 34 (2009) 43e52.
[22] H. Dhaou, F. Askri, M. Ben Salah, A. Jemni, S. Ben Nasrallah, J. Lamloumi,
Measurement and modeling of kinetics of hydrogen sorption by LaNi5 and
two related pseudobinary compounds, Int. J. Hydrogen Energy 32 (2007)
576e587.
[23] S.M. Yang, W.Q. Tao, Heat Transfer Theory, third ed., Chinese Higher Education
Press, Beijing, 1998.
[24] P. Muthukumar, M. Groll, Metal hydride based heating and cooling systems:
a review, Int. J. Hydrogen Energy 35 (2010) 3817e3831.
[25] B.D. MacDonald, A.M. Rowe, Impacts of external heat transfer enhancements
on metal hydride storage tanks, Int. J. Hydrogen Energy 31 (2006) 1721e1731.
[26] A. Jemni, S. Ben Nasrallah, J. Lamloumi, Experimental and theoretical study of
a metalehydrogen reactor, Int. J. Hydrogen Energy 24 (1999) 631e644.
[27] Y.Q. Wang, F.S. Yang, X.Y. Meng, Q.F. Guo, Z.X. Zhang, I.S. Park, S. Kim, K.J. Kim,
Simulation study on the reaction process based single stage metal hydride
thermal compressor, Int. J. Hydrogen Energy 35 (2010) 321e328.
[28] Z. Kolondzovski, A. Belancen, A. Akkio, Multiphysics thermal design of a high-
speed permanent-magnet machine, Appl. Therm. Eng. 29 (2009) 2693e2700.
[29] T. Dang, J.T. Teng, J.C. Chu, A study on the simulation and experiment of
a microchannel counter-flow heat exchanger, Appl. Therm. Eng. 30 (2010)
2163e2172.
[30] F.S. Yang, G.X. Wang, Z.X. Zhang, V. Rudolph, Investigation on the influences of
heat transfer enhancement measures in a thermally driven metal hydride
heat pump, Int. J. Hydrogen Energy 35 (2010) 9725e9735.
[31] K.J. Kim, G.M. Lloyd, K.T. Feldman Jr., A. Razani, Thermal analysis of the
Ca0.4Mm0.6Ni5 metalehydride reactor, Appl. Therm. Eng. 18 (1998) 1325e
1336.