Journal of Power Sources 195 (2010) 3529–3535

Contents lists available at ScienceDirect

Journal of Power Sources

journal homepage: [Link]/locate/jpowsour

Fabrication of metallic bipolar plate for proton exchange membrane fuel cells by
rubber pad forming
Yanxiong Liu, Lin Hua ∗
School of Automobile Engineering, Wuhan University of Technology, Luoshi Road 122#, Wuhan 430070, China

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

Article history: In this paper, the rubber pad forming process is used to fabricate the metallic bipolar plate for a proton
Received 10 December 2009 exchange membrane (PEM) fuel cell, which has multi-array micro-scale flow channels on its surface. The
Accepted 10 December 2009 rubber pad forming process has the following advantages: high surface quality and dimensional accuracy
Available online 21 December 2009
of the formed parts, low cost of the die because only one rigid die is required, and high efficiency. The
process control parameters (rubber hardness, internal and outer radii, draft angle) of the rubber pad
Keywords:
forming are analyzed by the finite element method using the commercial software Abaqus. After that,
Metallic bipolar plate
the rubber pad forming process is used to manufacture a metallic bipolar plate of SS304 stainless steel
Rubber pad forming
Micro-scale flow channel
with perfect flow micro-channels. The results of this effort indicated that the rubber pad forming process
Fuel cell is a feasible technique for fabricating the bipolar plates of PEM fuel cells.
Finite element simulation © 2009 Elsevier B.V. All rights reserved.

1. Introduction The primary difficulty in manufacturing a metallic bipolar plate

is the formation of the micro-channel. In the past few years, sev-
In recent years, there has been great interest from governments eral fabrication methods of the metallic bipolar plate have been
and research institutions in proton exchange membrane (PEM) developed. Koc and Mahabunphachai [8] developed a manufactur-
fuel cells due to their high efficiency, fast startup, and potential ing process involving internal pressure-assisted embossing of the
for energy conservation, safety, and environmental protection. The micro-channels with in-die mechanical bonding. Lee and Lee [9]
PEM fuel cell is the main candidate for replacing the internal com- fabricated micro-scaled flow channels by electrochemical micro-
bustion engine in transportation applications [1,2]. However, the machining. Yokoyama et al. [10] studied fabrication by hot pressing
current cost of a fuel cell is 4–10 times greater than an internal in the supercooled liquid state. However, the fabrication process
combustion engine (PEM fuel cell: $200–30 kW−1 , internal com- using internal pressure-assisted embossing of micro-channels with
bustion engine: $30–50 kW−1 ). Of all the different components of in-die mechanical bonding is not suitable when the anode and
the fuel cell, the bipolar plate is the largest bottleneck for the com- cathode plates are not symmetrical. In addition, hydroforming and
mercialization of the PEM fuel cells because its weight is 60–80% mechanical bonding are difficult to perform in a single produc-
of the stack weight and its cost is 30–45% of the stack cost [3,4]. tion set up, and experiments investigating the hydroforming and
The U.S. Department of Energy (DOE) has set a goal of lower- mechanical bonding were carried out separately as the authors pre-
ing the overall cost of the bipolar plate to $6 kW−1 by the year sented in their paper. The surface quality of the parts formed by
2010. electrochemical micro-machining is poor, and this method is ineffi-
Nowadays, there are three main types of bipolar plates: (1) cient. A die with high heat resistance is required for the hot pressing
graphite bipolar plates, (2) polymer–carbon composite bipolar method, and the productivity for this method is also low.
plates, and (3) metallic bipolar plates. Metallic bipolar plates have This paper presents a novel method for producing a metal-
received considerable attention due to their low cost, excellent lic bipolar plate for use in PEM fuel cells: the rubber pad
mechanical, electrical, and thermal properties, and their ease of forming process, which adopts a rubber pad contained in a con-
production [5–7]. Parallel to the development of a corrosion resis- tainer acting as a rigid die. The rubber pad forming process
tant metallic material, it is also necessary to develop an efficient has many advantages over the traditional sheet fabrication pro-
and cost effective fabrication process of the bipolar plate in order to cess using rigid die set [11]. The new method only utilizes one
make the production of metallic bipolar plates commercially viable. rigid die, and the lower die is a rubber pad. The contact sur-
face between the rigid die and the rubber is flexible, which
greatly improves the formability of the blank to be shaped. Only
∗ Corresponding author. Tel.: +86 27 87168391; fax: +86 27 87168391. one rigid die must be manufactured, and the rubber pad and
E-mail addresses: liuyx716@[Link], lhuasvs@[Link] (L. Hua). rigid die do not need to be assembled precisely. Therefore, the

0378-7753/$ – see front matter © 2009 Elsevier B.V. All rights reserved.
doi:10.1016/[Link].2009.12.046
3530 Y. Liu, L. Hua / Journal of Power Sources 195 (2010) 3529–3535

Fig. 2. Schematic diagram of the bipolar plate sample.

Fig. 1. Schematic diagram of the rubber pad forming process.

The equipment for the rubber pad forming is designed to man-
ufacture the metal bipolar plate as shown in Fig. 3. The equipment
is composed of five parts: the rigid die, the stainless steel sheet,
time and cost required for the forming equipment can be greatly
the rubber pad, the steel container enclosing the rubber pad, and
reduced.
the fastening plate fixing the container. Polyurethane rubber with
Former researchers have investigated the rubber pad forming
a Shore A hardness of 70 (HD70) is used as a pad, and a SS304
process. Browne and Battikha [12] described a rubber pad pro-
annealed stainless steel sheet with a thickness of 0.1 mm is applied.
cess for manufacturing an aircraft wing or tail flap. Dirikolu and
Because much time and effort are required to manufacture the rigid
Akdemir [13] investigated the significant parameters associated
die with various geometries, FE simulation is adopted as the major
with the rubber pad forming process using numerical method. They
tool to investigate the effect law of the process parameters in this
determined that the rubber hardness, advance, blank material type,
study.
contact friction, and die design are crucial parameters that require
adjustment before production. Thiruvarudchelvan [14] presented
several kinds of flexible processes to produce sheet metal ashtrays 3. Finite element modeling
and plates. Salau [15] optimized the process with numerical sim-
ulations and experiments. Ramezani et al. [16] studied the friction The bipolar plate is not symmetrical (Fig. 2) and would ideally
behavior of the rubber pad forming process. In these studies, atten- be simulated by a 3D FE model. However, numerical simulations
tion has been concentrated on the deformation of large parts, but of the rubber pad forming process are complicated because the
the forming process deforms differently during the fabrication of method involves coupling of the deformation of the blank and rub-
micro-parts such as micro-channels [17]. ber. Because of the large deformation of the rubber pad, a severe
In this paper, finite element analysis (FE, Abaqus/Standard soft- mesh distortion may occur, which would lead to loss of precision
ware) is used to analyze the rubber pad forming process. The main and even termination of the simulation. In order to save time and
process parameters such as the rubber hardness and the key geo- improve the computational precision, a 2D FE model is used. The
metric dimensions of the rigid die (the draft angle ˛, outer radius three micro-scaled flow channels that are made in the bipolar plate
R, and internal radius r) are explored with the FE model. A metallic sample (Fig. 2) are analyzed as one cycle in the 2D FE model (Fig. 4).
bipolar plate is also fabricated and tested by a 3D laser scanning
system. 3.1. Material model

The rigid die is defined as an analytical rigid body and no mate-

2. Principles of the rubber pad forming process rial properties need to be specified. Therefore, only two different
materials are included in the FE model: a hyper-elastic material to
The rubber pad forming process is ideal for forming micro- describe the rubber pad and an elastic–plastic model of the blank.
channels. Fig. 1 shows a schematic diagram of the forming process In this study, polyurethane rubber is used as the rubber pad. The
[11]. As the rigid die moves down, the rubber deforms elastically behavior of the nonlinear hyper-elastic and incompressible rubber-
and provides a counter-pressure. Then, because of the counter- like material is usually described by the Mooney–Rivlin model,
pressure, the rubber and the blank flow into the cavity of the rigid
die together. The fabrication process can be divided into two steps:
(1) drawing the blank to the bottom of the rigid die using the
counter-pressure of the rubber pad, and (2) filling the blank until it
fits the rigid die.
A bipolar plate sample is designed as shown in Fig. 2. The
thickness of the bipolar plate is 0.1 mm, and the shape of bipolar
plate channel is snake-liked. The dimensions of the bipolar plate
channels affect the reaction efficiency of the fuel cell and play an
important role in the formability of the blank. According to the
studies of Watkins et al. [18], Peng et al. [19], and others, the geo-
metric dimensions of the micro-channel include the outer fillet
radius (R), internal fillet radius (r), draft angle (˛), flow channel
depth (h = 0.5 mm), flow channel width (w = 0.8 mm), and rib width Fig. 3. Sketch of the equipment used for bipolar plate fabrication by the rubber pad
(s = 1.2 mm). forming.
 Y. Liu, L. Hua / Journal of Power Sources 195 (2010) 3529–3535 3531

Fig. 4. 2D finite element model.

which uses a strain energy function W. The derivative of W with

respect to a strain component determines the corresponding stress
component. The strain energy can be expressed by the following
function [20]:
∂W Fig. 5. (a) Dimensions of the annealed SS404 stainless steel specimen and (b) photos
ij = of the specimens before and after the tensile test.
∂εij

n
1
(1)
W= Ckm (I1 − 3)k + (I2 − 3)m + k(I3 − 1)2 3.3. Dynamic contact conditions
2
k+m=1
There are two different contact pairs in the rubber pad form-
where I1 , I2 , and I3 , are the strain invariants, k is the bulk modu- ing process: between the surface of the rigid die and the upper
lus, and Ckm is the constant of the Mooney–Rivlin material model. surface of the blank and between the lower surface of the blank
For incompressible materials, I3 = 1. In this study, two-parameters and the upper surface of the rubber pad. Both pairs are surface-to-
(C10 and C01 ) of the Mooney–Rivlin model are used to describe the surface contacts and allow a small amount of sliding between the
polyurethane rubber pad material behavior, and n = 1. surfaces. The frictional behavior of both contact pairs is assumed to
To investigate the effect of the rubber hardness on plate fabri- follow Coulomb’s model. Under lubricated conditions, the friction
cation, two kinds of rubber (HD55 and HD70) are used in the FE coefficient is 0.2 for the first pair and 0.1 for the second pair [20].
model. The mechanical properties of the rubber pads are listed in
Table 1.
3.4. Constraint and load
An SS304 annealed stainless steel sheet with a thickness of
0.1 mm is used as the blank. The stress–strain curve is obtained
The rubber pad is fixed in the container during the forming pro-
from tensile tests of the annealed SS304. Five tensile specimens
cess. In order to simplify the simulation model, the container is not
conforming to GB-T228-2002 are tested (Fig. 5). The mean value
considered in this model through adding constraint on the rubber
of the curves used in the FE model is shown in Fig. 6. The elas-
pad. The lower surface of the rubber is held fixed in all directions,
ticity module (E) and Poisson’s ratios () of the annealed SS304
and the sides of the rubber surface are fixed in the x-direction and
are 162.5 GPa and 0.3, respectively. And the constitutive relation of
are free in the y-direction. The rigid die is only allowed to move in
annealed SS304 in the tensile plastic deformation can be obtained
the y-direction. Translation of the rigid die can be controlled by a
by the method of curve fitting and linear regression as Eq. (2):
concentrated force, F, which is the load force in this model.
 = (799.9ε0.232 + 60.9) MPa (2) A detailed geometry of the rigid die and the specific process
conditions used in the FE model are summarized in Table 2.
where  and ε are the true stress and true strain, respectively.

3.2. Finite element mesh

In the rubber pad forming process, the rigid die is treated as an

analytical rigid body. Four-node bilinear plane strain with quadri-
lateral reduced and integration hourglass control elements (CPE4R)
are applied to model the stainless steel blank. Four-node bilinear
plane strain quadrilateral hybrid elements with reduced integra-
tion and hourglass control (CPE4RH) are applied to describe the
rubber pad. Hybrid element is incompressible and suitable to rep-
resent the rubber-liked material property [21].

Table 1
Mechanical properties of the rubber pads [20].

Polyurethane Hardness M–R M–R Poisson’s

rubber shore A constant C10 constant C01 ratio ()
(MPa) (MPa)

Type 1 55 0.382 0.096 0.4999

Type 2 70 0.736 0.184 0.4999 Fig. 6. Stress–strain curve of the annealed stainless steel SS304.
3532 Y. Liu, L. Hua / Journal of Power Sources 195 (2010) 3529–3535

Table 2
Dimensional information for the FE model and specifications of the fabrication process employed in this
study.

Dimensions of the FE model (mm)

Rigid die

H = 0.5
s = 1.2
W = 1.0
˛ = 5◦ /10◦ /15◦ /20◦
R = 0.1/0.2/0.3
r = 0.1/0.2 (R + r ≤ H)

Load force (kN) 100/250

3.5. Verification of the simulation model Fig. 7 shows the bipolar plate formed by the experiment and
the FE simulation. The plate is cut in the middle along the dot-
To verify the accuracy of the FE simulation, the fracture surface ted line. In order to observe the fracture, the profile of the plate
morphology of a bipolar plate formed using the rubber pad forming is projected and enlarged to 100 times using optical measurement
process is compared with the results of the FE simulation. Both in equipment (Fig. 7a). The plate is cracked at the corner of the rib of
the simulation and the experiment, the dimensions of the rigid die the micro-channels. Fig. 7b shows the surface of the plate predicted
R, r, and ˛ are 0.2, 0.2, and 20◦ , respectively, and the load force F is by the FE simulation. The stress is concentrated at the corners of
250 kN. the micro-channels, and the thickness in these locations is reduced,

Fig. 7. The bipolar plate formed by the experiment and the FE simulation.

Fig. 8. Simulated micro-channels with ˛ = 20◦ and different radii: (a) R = 0.1, r = 0.2; (b) R = 0.2, r = 0.2; (c) R = 0.3, r = 0.2.
 Y. Liu, L. Hua / Journal of Power Sources 195 (2010) 3529–3535 3533

leading to cracking. The FE simulation agreed well with experi-

mental results and could be used to investigate the effect of other
process parameters on deformation.

3.6. Simulation results

3.6.1. Effect of outer radius R and internal radius r

In the rubber pad forming process, the radii of the rigid die
played an important role in determining the formability of the
blank. A simulation is performed with a constant internal radius
(r = 0.2), a rubber pad of HD70, a load force F = 100 kN, and an outer
radius R that varied from 0.1 to 0.3 mm. As R increased, the filling
Fig. 9. Stress and strain state at the corner of a micro-channel. percentage of the blank increased (Fig. 8). When R is 0.2 or 0.1, the
channels might be cracked at the corners, but an outer radius of
0.3 mm appeared to lead to a stable plate. This can be explained by
Fig. 9.
Fig. 9 represents the stress and strain states at the corner of
the micro-channel during deformation. It shows that plate not only
suffers from tensile stress  1 in the radial direction, but also from
stress  3 in the tangential direction and the stress  2 due to the
blending pressure imposed by the rigid die [22]. The smaller the
value of R, the larger the blending pressure imposed by the rigid
die will be. The sheet metal may bend seriously and fail to fill the
cavity of the rigid die if the outer radius is too small.
To investigate the effect on plate deformation, r is changed from
0.1 mm to 0.2 mm, with R held constant at 0.3 mm (r + R ≤ 0.5).
Fig. 10 shows the formed plates at different fillet radii for the form-
ing load F = 250 kN. When r = 0.1 mm, the blank does not fill the
Fig. 10. Fabricated plates with ˛ = 20◦ and different internal radii: (a) R = 0.3, r = 0.1; rigid die completely, and the filling percentage is perfect r = 0.2 mm.
(b) R = 0.3, r = 0.2. Therefore, the smaller r is, the more difficult it is to fill the cavity of
the rigid die.

3.6.2. Effect of the draft angle ˛

To investigate the effect on the rubber pad forming process, the
draft angle ˛ is changed from 5◦ to 20◦ for R = 0.3 mm, r = 0.2 mm,
and F = 250 kN. In these models, all of the blanks filled the rigid die
completely, but the variation curves of these forming loads with the
time increment are different. Fig. 11 shows the forming load with
the time increment at various draft angles. The time required for
the blank to fill the cavity of the rigid die decreased as ˛ increased.
Therefore, larger values of ˛ is beneficial for the rubber pad forming
process.

3.6.3. Effect of rubber hardness

To analyze the effect of the rubber hardness on the deformation
of the plate, two rubber pads of different Shore A hardness (Table 1)
are used in the FE model. The key parameters are R = 0.3 mm,
r = 0.2 mm, ˛ = 20◦ , and F = 250 kN. As shown in Fig. 12, the filling
percentages of the micro-channels are perfect regardless of the
hardness of the rubber pads. Fig. 12 also shows that the Von-Mises
Fig. 11. Forming load with the time increment at various draft angles ˛ of the rigid
die. stress distributions of the formed plates are almost the same in the
stress concentrated region. The hardness of the rubber pad is not

Fig. 12. The formed parts at different rubber pad (250 kN): (a) rubber pad with Shore A hardness of 55; (b) rubber pad with Shore A hardness of 70.
3534 Y. Liu, L. Hua / Journal of Power Sources 195 (2010) 3529–3535

Fig. 13. Thickness variation of the formed plate.

Fig. 16. Profiles of micro-channels measured by the 3D laser scanning measurement

system at (a) location A and (b) location B.

tion of the three micro-channels is almost the same. Moreover, the

thickness distribution of one channel is symmetrical. In this study,
we are most concerned with the rupture region within half of one
channel. Therefore, one half of a channel of the formed bipolar plate
with a length of 1.7 mm is analyzed (Fig. 13). The curve of the thick-
ness variation as a function of the measure distance can be divided
into three regions. In region A, the thickness reduction is small.
In region B (the side of the micro-channel), the thickness of the
blank decreases rapidly, and the thinnest point of the plate is in
this region. This thin region is where the blank experienced more
resistant friction force, which blocked the flow of material during
the forming process. The thickness distribution is not uniform in
region C due to the friction between the surfaces of the rigid die
Fig. 14. Photographs of the forming equipment for the experiment. and the blank.

an important factor in the forming process because the rubber pad 4. Experimental setup
is enclosed in the container, incompressible during the deforma-
tion, and capable of transferring the same loading force regardless The simulation results are taken as the process parameters for
of hardness. the experiment. The rubber pad equipment is designed and man-
ufactured as shown in Fig. 14, and the key geometric dimensions
3.6.4. Thickness variation of the rigid die are R = 0.3 mm, r = 0.2 mm, ˛ = 20◦ , and F = 250 kN.
During the forming process, the thickness of the formed plate A polyurethane rubber with a Shore A hardness of 70 is used
will decrease. However, after deformation, the thickness distribu- for the rubber pad, and an SS304 annealed stainless steel sheet

Fig. 15. (a) Front and (b) back view of a metal bipolar plate fabricated by the rubber pad forming process.
 Y. Liu, L. Hua / Journal of Power Sources 195 (2010) 3529–3535 3535

with a thickness of 0.1 mm is applied. The dimensions of the scale. Therefore, this forming process can reduce the overall
polyurethane rubber and blank are 82 mm × 75 mm × 15 mm and cost of producing bipolar plates. The estimated cost of a double
82 mm × 75 mm × 0.1 mm, respectively. A hydraulic press with a bipolar plate manufactured by this method with a non-coated
capability of 1000 kN is applied. SS304 is $1.9. This cost could meet the $6 kW−1 target but will
be decided based on the surface treatment of the double bipolar
5. Results and discussion plates. The main drawback of the rubber pad forming process
is that the life of the rubber pad is not so long, and should be
The metal bipolar plate samples are fabricated by rubber pad replaced after the production of only about 100 plates. Future
forming. The metal bipolar plate with a high surface quality is not studies should focus on prolonging the life of the rubber pad by
cracked and winkled (Fig. 15). changing the structure of the rigid die and the container. Over-
To examine whether the blank filled the cavity of the rigid die all, we strongly believe that metallic bipolar plates fabricated
or not, the profiles of the formed bipolar plate and the rigid die are by the rubber pad forming process will successfully solve the
measured using a 3D laser scanning measurement system with a problem of high fuel cell stack cost and improve the commercial
measurement accuracy of 0.02 mm. A comparison of the profile at competitiveness of fuel cells.
location A (vertical to the micro-channel) and location B (horizontal
to the micro-channel) between the bipolar plate and rigid die is Acknowledgments
presented in Fig. 16. From this figure, it is known that the cavity
of the rigid die is filled with blank very well at both location A and The authors would like to thank the Natural Science Founda-
location B, but the load force is insufficient for the blank to fill the tion of China for Distinguished Young Scholars (No. 50725517) and
bottom corner of the die. The profiles of the bipolar plate and the Wuhan Academic Leader Project (No. 200750730314) for the sup-
rigid die are almost coincident, which means that the amount of port given to this research.
spring back of the plate is very small. Based on the above analysis,
it is known that the rubber pad forming process is a promising References
process for fabricating metallic bipolar plates.
[1] H. Tawfik, Y. Hung, D. Mahajan, J. Power Sources 163 (2007) 755–767.
[2] N.B. Huang, B.L. Yi, M. Hou, et al., Prog. Chem. 17 (2005) 963–969.
6. Summary and conclusions [3] K. Sopian, Renew. Energy 31 (2006) 719–727.
[4] T. Matsuura, M. Kato, M. Hori, J. Power Sources 161 (2006) 74–78.
(1) This study describes the detailed procedure for the formation [5] K.S. Weil, G. Xia, Z.G. Yang, J.Y. Kim, Int. J. Hydrogen Energy 32 (2007)
3724–3733.
of a bipolar plate using the rubber pad forming process. The [6] T. Besmann, J. Henry, J. Klett, Proc. Fuel Cell Semin. 30 (2003) 61–64.
key process parameters of the rubber pad method are investi- [7] L.M. Chun, T.G. Liu, J. TianJin Univ. 39 (2006) 1252–1257.
gated with FE simulations. As the outer radius and draft angle [8] M. Koc, S. Mahabunphachai, J. Power Sources 172 (2007) 725–733.
[9] S.J. Lee, C.Y. Lee, J. Power Sources 185 (2008) 1115–1121.
increased, the formability of the blanks increased. The smaller [10] M. Yokoyama, S. Yamaura, H. Kimura, et al., Int. J. Hydrogen Energy 33 (2008)
the internal radius, the harder it is to fill the cavity of the 5678–5685.
rigid die. The hardness of the rubber pad is not an important [11] X.M. Jiang, M. Tian, J.Y. Zhang, The Application of the Polyurethane Rubber in
the Field of Stamping Technology, Defense Industry Academic Press, Beijing,
parameter for the formation of a bipolar plate. The thickness
1989.
distribution of the formed bipolar plate is uneven, and the most [12] D.J. Browne, E. Battikha, J. Mater. Process. Technol. 55 (1995) 218–223.
dangerous position occurred at the side of the micro-channel. [13] M.H. Dirikolu, E. Akdemir, J. Mater. Process. Technol. 148 (2004) 376–381.
(2) A 0.1-mm thick bipolar plate of SS304 stainless steel is fab- [14] S. Thiruvarudchelvan, J. Mater. Process. Technol. 122 (2002) 293–300.
[15] G. Salau, Mater. Des. 22 (2001) 299–315.
ricated using rubber pad forming equipment and the optimal [16] M. Ramezani, Z.M. Ripin, R. Ahmad, J. Mater. Process. Technol. 209 (2009)
process parameters indicated by the FE simulation results. 4925–4934.
(3) The formed plate had a high surface quality, and the amount [17] U. Engel, R. Eckstein, J. Mater. Process. Technol. 125–126 (2002) 35–44.
[18] D.S. Watkins, K.W. Dircks, D.G. Epp, Novel fuel cell fluid flow field plate, US
of spring back of the plate is very small, indicating that the 4,988,583 [P] (1991).
dimension precision is very high. This rubber pad forming pro- [19] L.F. Peng, X.M. Lai, D.A. Liu, P. Hu, et al., J. Power Sources 178 (2008) 223–230.
cess could improve the quality of the product because of the [20] M.H. Dirikolu, E. Akdemir, J. Mater Process Technol. 148 (2004) 376–381.
[21] Abaqus Theory Manual, Version 6.3. Hibbitt, Karlsson and Sorensen Inc., Paw-
low stiffness of the die material and minimize the die cost. The tucket, RI, USA. 2002.
operation of this forming process is very simple. The produc- [22] Z.R. Wang, Fundamentals of Plastic Deformation Mechanics, Defense Industry
tivity is high, and this method can produce the plates on a large Academic Press, Beijing, 1989.