Journal of Power Sources 280 (2015) 47e56

Contents lists available at ScienceDirect

Journal of Power Sources

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

Characterization of plasticity and fracture of shell casing of lithium-ion

cylindrical battery
Xiaowei Zhang*, Tomasz Wierzbicki
Impact and Crashworthiness Lab, Massachusetts Institute of Technology, 77 Massachusetts Ave., Room 5-218, Cambridge, MA 02139, USA

h i g h l i g h t s

A comprehensive set of mechanical tests on shell casing of 18650 cells was performed.

Plastic and fracture models of shell casing were provided along with calibration procedure.

Jellyroll was described by means of a crushable foam model.

The presence of shell casing significantly improves the cell's resistance to mechanical abuse loading.

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

Article history: Most of the literature on lithium-ion battery cells is concerned with modeling of jellyroll with little
Received 23 December 2014 attention to properties of shell casing. However, shell casing provides substantial strength and fracture
Received in revised form resistance under mechanical loading and therefore must be an important part of modeling of lithium-ion
6 January 2015
batteries. The paper reports on a comprehensive test program on commercially available empty shell
Accepted 13 January 2015
Available online 14 January 2015
casing of 18650 lithium-ion cylindrical cells. Part of the tests was used to determine plastic and fracture
properties from sub-size specimens cut from lateral part of the cans. The other part served to validate
plasticity and fracture models under various loading conditions. The associated flow rule was used to
Keywords:
Mechanical loading
simulate plasticity behavior and Modified Mohr-Coulomb (MMC) fracture model was adopted to predict
Lithium-ion cells crack initiation and propagation of shell casing. Simulation results confirmed that present plasticity and
Battery shell casing fracture models could predict global plastic behavior of the cells under different loading conditions. The
Plasticity jellyroll model with volumetric hardening was introduced to compare the performance of empty shell
Fracture casing, bare jellyroll and complete battery cell. It was shown that in many loading situations, for example,
three point bending of the cylindrical cells, the metallic shell casing provides most of mechanical
resistance.
© 2015 Elsevier B.V. All rights reserved.

1. Introduction limit of cells in terms of resistance to mechanical loading and

Shell casing of lithium-ion batteries provides the first level of
thermal and mechanical protection to the jellyroll. It has to perform
well under verity of abuse loading, and it must be light and easy to
manufacture. The casings are often made from extruded aluminum
tubes with laser welded endcaps. More commonly, a multi-stage
deep drawing technology is used for both steel and aluminum
thin sheets. The supplier of shell casing provided minimum or no
information on the mechanical properties of raw material used or
the final product. Such information is critical to assess performance
* Corresponding author.
E-mail address: zhangxv@mit.edu (X. Zhang).
fracture during accidental load and/or compliance with safety
standards such as SAE J2464 [1].
There are a number of requirements that lithium-ion batteries
must satisfy regarding electro-chemical, thermal and mechanical
properties. Most of the literature on the modeling of lithium-ion
cells is devoted to thermal management [2,3]. Recently, the
importance of the laboratory mechanical test and numerical
simulation has been recognized by the community. Sufficient evi-
dence has now been accumulated proving that fracture of shell
casing is a valid and serious failure mode, exposing the partially
damage jellyroll to the environment. Sahraei and Hill [4] performed
tests on axially compressed cylindrical cells during which a longi-
tudinal crack was developed from a local bulge. Fracture on the
tensile side of shell casing was observed in three-point bending test

http://dx.doi.org/10.1016/j.jpowsour.2015.01.077
0378-7753/© 2015 Elsevier B.V. All rights reserved.
48 X. Zhang, T. Wierzbicki / Journal of Power Sources 280 (2015) 47e56
[5,6]. Sahraei et al. [7] described a dynamic test on large format
cylindrical cell subjected to axial compression. The cell exploded,
tearing the shell casing into small pieces. Recently, Xia et al. [8]
performed numerical simulation of the 18650 cell under symmet-
ric and eccentric load by a sharp punch, parallel to the longitudinal
direction. The symmetrically loaded shell casing ruptured under
6 mm indentation depth. Fracture of eccentrically loaded cell
occurred much earlier, when the punch travelled only 1.6 mm.
Considering the existing evidence and potential hazard associated
with spreading gases, spill of the electrolyte and possible fire from
the bursting cell to the neighboring ones, a proper attention should
be given to this problem by the battery safety community.
The objective of the present paper is to determine the contri-
bution of the strength of the shell casing to the overall response of
the cells to the mechanical loading. To this end, plastic and fracture
properties of the shell casing material were determined by using
the most modern technology developed in the Impact and Crash-
worthiness Lab of MIT [9e16]. This effort consisted of an extensive
experimental program, model calibration and component valida-
tion. Almost 100 empty shell casings were tested to complete the
stated task. Specimens were cut from the actual shell casing, which
restricted their size and shape to a certain degree. The shell casing
material had undergone relatively large strain during the deep
drawing process. These strains were distributed non-uniformly
along the length of the cell. The anisotropic properties arising
first from the rolling process of the sheets and then deep drawing
process are substantial and were to be taken in to account. There is
a considerable variation in thickness of the can from 0.253 mm to
0.29 mm at the ends. The endcap on the other hand was not sub-
jected to any strain histories and represents the virgin material.
However, the size of the endcap is too small to cut any specimens.
All the above features made the testing and calibration procedure
much more difficult as compared to previous work in the lab on
thin automotive sheets. From this point of view, the present paper
is indeed innovative in terms of both test procedures and numerical
simulations.
In the last section of the paper, numerical simulations were
performed on bare jellyroll, empty shell casing and complete cells.
It was conclusively proven that in some loading situation, such as
three point bending, the shell casing provided most of the me-
chanical resistance of the cell. Distinguished feature of the present
paper is that not only it provides new information about the
important modeling problem, but also shows in some details how
the information was obtained.
Fig. 1. Non-uniform thickness distribution of the shell casing.
2. Plasticity modeling for the battery shell casing
2.1. Material
The cylindrical cells are widely used as batteries for laptop and
portable power. Such cells are also powering the Tesla Model S. The
shell casing is made by deep drawing, which can reduce the cost
compared with the extruding one with welded endcap. The present
paper is studying the properties of the commercially available
deep-drawn shell casing of the 18650 cell made from low carbon
steel, which is approximately 18 mm in diameter and 65 mm in
length.
Generally, a typical deep drawing process consists of several
steps from a flat circular sheet to the final shape. The deep drawing
process results in considerable accumulated plastic strain, leading
to the deformation induced anisotropic of the casing that can be
determined experimentally. Therefore, due to the anisotropic
properties of the virgin rolled sheet and deformation induced
anisotropy, the thickness of the shell casing varies along both axial
direction and hoop direction of the casing, as shown in Fig. 1. The
area with maximum thickness is around the open end of the can
and minimum thickness is approximately in the middle. Such a
distribution of thickness was also reported in the literature for deep
drawn shell casing [17]. The average wall thickness is 0.26 mm. The
chemical composition is summarized in Table 1. All specimens were
extracted from the same package of the battery cell (100 pieces)
and can be assumed to have the same mechanical properties.
2.2. Preparation of the specimens and description of the test
method
In this section, a rather detailed procedure on the preparation
and test method is given so that similar test could be performed
around other labs in the world. Two methods were used to prepare
tensile specimens for plasticity calibration. In both methods, the
endcap was removed. In the first method, the shell casing was cut
lengthwise and flattened, as shown in Fig. 2. Dogbone-shaped
specimens were cut from three different flattened-sheet orienta-
tions (axial or deep drawing direction, hoop direction and diagonal
direction) using a milling machine. The specimen geometry
(Fig. 3a) is different from that in ASTM-E8M standard (sheet type),
due to the limited length and diameter of the shell casing. Since the
dimension of the flattened sheet of the shell casing is 65 mm in
length and 56 mm in width (hoop direction), the geometry of
Table 1
Chemical composition of the shell casing.
Element Mn P Ni Cr Cu Mo V Ti Co
Percentage 0.16 0.008 1.36 0.056 0.012 0.002 0.003 0.003 0.006
 X. Zhang, T. Wierzbicki / Journal of Power Sources 280 (2015) 47e56
Fig. 2. Specimen preparation from the flattened shell casing in different material orientations (0 , 45 and 90 to the deep drawing direction).
Fig. 3. a) Geometry of the dogbone specimen (Dimension in mm); b) specimens before test.
dogbone along hoop direction is 10 mm shorter than in axial and
diagonal direction. Fig. 3b shows the specimens after cutting.
Relatively large spring back was observed for specimens along di-
agonal and hoop direction.
The flattening process may bring a certain amount of pre-
damage to the final specimens. To study the flattening effect, an
alternative uniaxial tensile test was conducted. Window-shaped
specimens were cut directly from the shell casing without flat-
tening process, as shown in Fig. 4 (left). The geometry of the gauge
section of the specimen is approximately the same as the above
dogbone specimen. Fig. 4 (right) shows the setup of the uniaxial
tensile test of the window-shaped specimen. To calculate the en-
gineering stressestrain curve of the specimen, exact area of the
initial gauge section was obtained from geometric model. Digital
Image Correlation (DIC) was used to measure the local strain and
Fig. 4. Geometry of the window-shaped specimen (left) and uniaxial tensile test setup
(right).
49
displacement (Vic 2D, version 2009). All the specimens were tested
on a table-top uniaxial loading machine (Instron Model 5940 with
load capability 2 kN) with speed of 0.2 mm min-1.
Fig. 5 shows the engineering stressestrain curves in three di-
rections from 6 tests of flattened shell casing (2 in each direction).
The displacement for calculating engineering strain was measured
from positions closed to the end of the gauge section (gauge length
was 8 mm for specimens along hoop direction and 18 mm for the
other two). The cross section was assumed as 0.26 mm when
calculating the engineering stress. The repeatability of the test is
relatively good, even though the thickness of the flattened sheet
varies from specimen to specimen and even varies from region to
region within specimen. The nominal Young's modulus is 160 GPa
(due to non-uniform thickness) and the average flow stress is
500 MPa. The comparison of window-shaped specimen and dog-
bone specimen is shown in Fig. 6. The results show that there was
only a slight hardening effect due to the flattening process for this
material.
2.3. Determination of hardening curve before and after necking
For tensile test, the true stressestrain curve is only valid up to
the point of the diffused necking. To determine the post-necking
stressestrain curve, an inverse method must be used. Following
the procedure developed by Dunand [12], a one-parameter linear-
combination of Swift-Voce law fitting function was adopted
Kopt ¼ aKs þ ð1  aÞKv (1)
where Swift law Ks ¼ A(εp þ ε0)n and Voce law
Kv ¼ BC*exp(D*εp). The parameters in Swift and Voce law were
fitted from uniaxial tensile test data while a was optimized from
the inverse method. All the calibrated parameters are listed in
Table 2.
50 X. Zhang, T. Wierzbicki / Journal of Power Sources 280 (2015) 47e56

Fig. 5. Engineering stressestrain curves (left) and fractured dogbone specimens (right).

Fig. 6. Comparison of stressestrain curves of uniaxial tension (left) and fractured window-shape specimen (right).

The yield surface of the shell casing material is also crucial since Table 2
the material of shell casing has considerable anisotropy. Hill'48 Combined Swift and Voce hardening law parameters.

yield condition was used and related parameters are given in A ε0 n B C D a

Appendix A. 800 0 0.084 356 194.5 360 0.5

3. Fracture modeling for the battery shell casing

where Dðεf Þ is the damage indicator, εp is the equivalent plastic
In the previous work of the investigating team, it was shown ε f is fracture locus, h is the stress triaxiality and q is the
strain, b
that the Modified Mohr-Coulomb (MMC) fracture model was able normalized Lode angle [10].
to predict correctly the initiation and propagation of the cracks. Another concept is the fracture locus b ε f ðh; qÞ for proportional
This fracture model was adopted for the present paper. There are loading. A specific form of the fracture locus was derived from the
two concepts in the MMC model. One is the damage accumulation classical Mohr-Coulomb fracture model [10]. Transforming the local
rule. fracture criterion on the fracture plane into the space of (h; q; εf ), the
following fracture criterion was obtained
ε

 Zp dεp
D εf ¼   (2) where A ¼ 680 MPa and n ¼ 0.047 are the parameters from refitted
b
ε f h; q
0 optimized hardening curve by Swift law, and c1, c2, c3 are three
parameters of MMC model needed to be calibrated from fracture

8 " 2sffiffiffiffiffiffiffiffiffiffiffiffiffiffi 39 1
pffiffiffi    #      =n
  <A 3 qp 1 þ c21 qp 1 qp
b
ε f h; q ¼ c þ pffiffiffi ð1  c3 Þ sec 1 4 cos þ c1 h þ sin 5 (3)
:c2 3 2  3 6 3 6 3 6 ;
 X. Zhang, T. Wierzbicki / Journal of Power Sources 280 (2015) 47e56 51

tests with different combinations of stress triaxiality h and
normalized Lode angle q.
3.1. Calibration of the MMC fracture model
Calibration procedure, to determine the three unknown pa-
rameters (c1, c2, c3) followed the work by Luo [12]. This involves
extracting five different specimens from the shell casing material
and performing tests in which measured were loadedisplacement
curves including displacements to fracture. The specimens in
Fig. 7aed were cut from flattened sheets of shell casing and the
shear specimens in Fig. 7e were cut directly from casings without
flattening process because of the buckling and spring back prob-
lems for flattened shear specimens. The fracture parameters
(c1 ¼ 0.20, c2 ¼ 425 MPa, c3 ¼ 0.90) were determined by using the
inverse method which required finite element simulation of each
test. Interested readers are referred to the above mentioned papers
for details. Here only the final results are given. Fig. 7 presents
comparison of measured and calculated load displacement curves.
In all cases, good agreements were observed.

Fig. 7. Comparison of forceedisplacement curves between test and simulation: (a) notched specimen; (b) specimen with central hole; (c) butterfly tension specimen; (d) punch
specimen; (e) simple shear specimen.
52 X. Zhang, T. Wierzbicki / Journal of Power Sources 280 (2015) 47e56

4. Validation of the plasticity and fracture model on the

“structural” level

To validate the plasticity and fracture model built in Sect. 3, four

types of experiments on empty shell casing were performed. All of
these types of tests are common loading cases of mechanical abuse
of battery cells. The endcap in all tests was not removed.

4.1. Axial compression test

Several tests on axial compression of the can were performed.

The measured loadedisplacement curves are shown in Fig. 8.
During these tests, no fracture of shell casing was observed.
Therefore, only the plasticity model can be validated at this stage. A
relatively accurate FE model of the shell casing including the end-
cap was built with thickness varying along axial direction, ac-
cording to the geometry measured in Fig. 1. The geometry of
endcap, which is not flat (Fig. 1), was carefully measured. In the FE Fig. 9. Force-displacement of lateral compresssion of shell casing.
model, the body of the shell casing was taken as anisotropic along
axial and hoop direction, while the endcap was assumed to be
defining the boundary condition conveniently. From Fig. 9, the force
isotropic. Two models were constructed, one with shell elements
level of the shell casing with endcap removed was greatly reduced,
(element type: S4R and mesh size 0.2 mm  0.2 mm) and the other
and the endcap played a major role in resisting the lateral
with solid elements (element type: C3D8R and mesh size:
compression of shell casing.
0.2 mm  0.2 mm  0.032 mm) and 8 elements through thickness.
Fig. 8 shows the comparison of the experiment and FE simulation
results. The numerical simulation predicted higher peak force, 4.3. Three point bending test
which is normal for a geometrically perfect axisymmetric model.
For the second and third peak, both the shell element and solid Three point bending test is also a typical loading case for battery
element models agree with experiment well. cell [5,6]. The test setup is shown in Fig. 11 (left), where the radius of
the cylindrical indenter is 10 mm, and the radii of supporting cyl-
inders are 24 mm. The distance between the two supporting cyl-
4.2. Lateral compression test
inders is 58 mm. No fracture was observed even when the
displacement of indenter reached 20 mm, 2 mm larger than the
The shell casing was also subjected to lateral compression be-
diameter of the casing. The model with shell element in Sect. 4.1
tween two flat plates. The forceedisplacement curves are plotted in
was used to simulation the test. The friction coefficient was set to
Fig. 9. No fracture was observed. The previous FE model with shell
0.05. The comparison of forceedisplacement curves of test and
elements (element type: S4R) was used to simulate the compres-
simulation is shown in Fig. 11 (right). The agreement between
sion process. The main contribution for the displacement mismatch
simulation and test is good. This also validated the plasticity model
in Fig. 9 is due to the modelling of the endcap, which has a
of shell casing only.
complicated geometry. Two additional simulations were run by
using the same model but one with endcap removed (Model A) and
the other with endcap only (Model B) as shown in Fig. 10. The 4.4. Internal pressure test
model B that with endcap only was cut from the position of 6 mm
from the edge of endcap rather than the edge with the purposes of A hydraulic bulge test using water was introduced, and the

Fig. 8. Concertina folding of shell casing (left) and forceedisplacement curves of axial compression of shell casing (right).
 X. Zhang, T. Wierzbicki / Journal of Power Sources 280 (2015) 47e56 53

Fig. 10. Models with or without endcap.

Fig. 11. Test setup of three point bending test and forceedisplacement curve.

Fig. 13. Force-displacement curve of hydraulic bulge test.

Fig. 12. Setup of hydraulic bulge test.

setup is shown in Fig. 12. The shell casing was filled with so called
de-aired water, which is commonly used in rock and soil tests and is
assumed to be incompressible. A piston with a steel seal plug to
release the spare water was assembled in the open end. The piston
sat on a flat plate and the side with endcap was subjected to
compression by another flat plate. The O-ring was designed to
prevent water from leaking out during the test. The clearance be-
tween the edges of shell casing and piston prevents possible sup-
ports from piston. The experimental forceedisplacement curves are
shown in Fig. 13, where the displacements were measured from the
flat plate at the endcap side during the test. The test was performed
up to the point of bursting fracture. Fig. 14. Comparison of fracture in test and simulation (contour: plastic equivalent
strain).
To simulation this test, a fine FE model of the shell casing with
varied thickness distribution along axial direction (Fig. 1 average
54 X. Zhang, T. Wierzbicki / Journal of Power Sources 280 (2015) 47e56

curve) was used and the calibrated MMC fracture model was

 

implemented. In the model, the loading was applied as an inside pl pl 1 1 Pt

pressure, which was calculated from the force in Fig. 13 over the
area of the piston (diameter: 17.45 mm) by neglecting the small
friction force between O-ring and shell casing. From the simulation,
it was observed that slant fracture occurred when the pressure
reached maximum value from test, see Fig. 14. It is worth noting
that this kind fracture can seldom happen in practice because the
battery cell has cap valve to release internal pressure (typically less
than 3 MPa). However, this type of test provides important vali-
dation of the present fracture model. The fracture of shell casing
happened when the pressure load increased to 15.6 MPa in nu-
merical simulation, which indicated the accuracy of MMC fracture
model.
5. Modeling of jellyroll
In order to demonstrate the contribution of the shell casing to
the bending strength of an individual cylindrical cell, a computa-
tional model of the jellyroll must be established. Such a model,
along with a detailed calibration procedure was developed in a
series of publications [4,5] in the LS-DYNA environment. An alter-
native compressible foam material in Abaqus was proposed in
Ref. [8] to model deformation of 18,650 cylindrical cell. This model
was adopted in the present paper. The crushable foam model is an
extension of the one developed by Deshpande and Fleck [18]. Be-
sides the original feature of isotropic hardening law in the Desh-
pande and Fleck model, it incorporates a new option of a volumetric
hardening law and the corresponding yield function to address the
difference of yield evolution between hydrostatic compression and
tension states [19]. There are three concepts involved, the yield
surface, the hardening function and the flow potential. The pres-
sure dependent yield function is assumed in the form.
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi



u 0 12
u Pc εpl  Pt Pc εpl þ Pt
u vol vol
F≡tq2 þ a@p  A a
2 2
where q is Mises stress, p is pressure, pc is hydrostatic compression
yield strength and pt is tension yield strength which is a constant
throughout any plastic deformation process. s0c and p0c are initial
yield strengthpin uniaxial compression and hydrostatic compres-
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
sion. a ¼ 3k= ð3kt þ kÞð3  kÞ, where k ¼ s0c =p0c , kt ¼ pt =p0c . The
hardening function Pc ðεpl vol
Þ has the following form

 sc εaxial sc εaxial a2 þ 9 þ 3
Pc εpl ¼ (5)
vol sc ðεpl
axial Þ
Pt þ 3
where εpl and εpl are volumetric and uniaxial plastic strain and
vol axial
they are equal in uniaxial compression when the Poison's ratio is
assume to be very small (highly compressible).
The flow potential is defined as
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
9
G¼ q2 þ p2 (6)
2
The model involved four independent parameters, E, y, k, kt. The
stressestrain curve in Fig. 15 was calculated from the test data from
lateral compression of pouch cell in the author's laboratory [4]. The
tests were performed in room temperature. It was observed that
the temperature raise of jellyroll happens only after the onset of the
electric short circuit. The present model is valid up to the onset of
the electric short circuit. The values of Young's modulus and Pois-
son's ratio were obtained from Ref. [4]. The parameter k is set to 1.1
according to the test data [8] and kt is optimized from the simu-
lation. From the parameters in Table 3, all other parameters defined
above could be calculated.
6. Modeling of full battery
Having determined separately the constitutive behavior of the
shell casing and the jellyroll, we can now move to the prediction of
the strength of the complete battery. This will demonstrate the
relative contribution of the individual components to the strength
of the structural systems. Two loading cases were considered: three
point bending and lateral indentation under cylindrical punch
loading.
The computational model of the shell casing was the same as the
¼0 (4) one with solid elements described in previous section. It should be
noted that even though the shell casing is thin, the solid element
representation with 8 elements through thickness was used. There
were 190,000 solid elements in the shell casing model. The jellyroll
was modeled by means of the solid element with type C3D8R in
Abaqus [19], with a total of 670,000 elements. All together, the
complete model of the cell consisted of 860,000 elements. It would
take approximately 10 h on 6 CPUs workstation to complete one
loading case. General contact method was defined between the
interacting components with 0.05 friction coefficient.
Results of the three point bending are shown in the Fig. 16. Three
different cases are compared in Fig. 16a, the bare jellyroll (dash
double dot line), the empty shell casing (dash line) and the com-
plete battery (dash dot line). Also the experimental measured
loadedisplacement curves are shown by the solid lines. The bare
jellyroll developed a negligible resistance force, not larger than
10 N. This was a quite unexpected results but simple hand calcu-
lations of the bending deformation of an elastic beam confirmed
the above results. The strength of the empty shell casing was more
than an order of magnitude higher and reached almost 500 N. The
predicted loadedisplacement relation of the complete battery
(jellyroll inside the shell casing) followed closely the experimental

Table 3
Parameters of jellyroll model.

E(MPa) n k kt s ε

500 0.01 1.1 10 Piecewise

data
Fig. 15. Stressestrain curve of jellyroll.
 X. Zhang, T. Wierzbicki / Journal of Power Sources 280 (2015) 47e56 55

Fig. 16. Comparison of three-point bending test and simulation results and corresponding contour of deformation of cell components.

curve and reached 2.5 kN before failure. The above example clearly
shows that the strength of the battery is not a sum of the strength of
the individual components loaded separately. This result should
serve as a justification that modeling of the shell casing, which is
the main subject of the present paper, is indispensable for accu-
rately predicting the response of the lithium-ion batteries.
The second example is the local indentation of the cell by a rigid
cylinder. It is seen from the simulation that the sum resistance of
the components is much closer to that of the entire battery. The
contribution of the shell casing is not negligible and it is actually
quite important in the initial phase. This is illustrated in Fig. 17.
7. Conclusion
In this paper, the plasticity and fracture behavior of shell casing
of cylindrical lithium-ion battery was studied. Due to the rolling
process of sheets and deep drawing process, considerable aniso-
tropic properties of the shell casing were detected in the tests.
Those properties were well captured by introducing Hill'48 aniso-
tropic yield function with the associate flow rule. Another feature of
the present method was a relatively exact hardening curve valid for
the post necking range which was incorporated in the FE simula-
tion. Modified Mohr-Coulomb (MMC) fracture model was used to
predict the onset of fracture and an extrusive experimental pro-
gram was conducted to determine fracture parameters.
Four validation studies were performed on empty or water-
filled shell casing. Those included axial and lateral compression
tests, three point bending test and hydraulic bugle test. In each
case, a very good correlation between test and simulation results
was observed.
Only after the calibration and complete validation of plasticity
and fracture models of the shell casing, we were able to

Fig. 17. Comparison of Indentation test and simulation results and corresponding contour of deformation of cell components.
56 X. Zhang, T. Wierzbicki / Journal of Power Sources 280 (2015) 47e56

demonstrate that the shell casing provided the most of the resis-
Table A-1
tance of the cell in some loading situation. These properties were
Hill'48 coefficients determined from Lankford coefficient
proven by comparing bending strength of the bare jellyroll, empty
shell casing and the complete cell. It was shown that the jellyroll F G H L M N
itself provided only a small percentage of the strength of the 0.65 0.56 0.43 1.5 1.5 1.37
complete battery. The final conclusion is that mechanical behavior
of the shell casing must be an integral part of characterizing the
lithium-ion cell from the mechanical loading. 1077.

Acknowledgement
Table A-2
Lankford coefficients determined from uniaxial tensile tests
The authors would like to acknowledge the financial support of
MIT Battery Consortium. We would also like to thank our col- a ¼ 0 a ¼ 45 a ¼ 90 ravg
leagues, Yong Xia, Elham Sahraei, Stephane Marcadet, Kai Wang r 0.77 0.62 0.66 0.683
and Keunhwan Pack for valuable discussions. Thanks are due to
Pierre-Olivier Santacreu of Aperam for determining the chemical
composition of the metal casing. [2] H.S. Hamut, I. Dincer, G.F. Naterer, Exergy analysis of a TMS (thermal man-
agement system) for range-extended EVs (electric vehicles), Energy 46 (2012)
117e125.
Appendix A. Calibration of Hill's 48 anisotropic plasticity [3] H.S. Hamut, I. Dincer, G.F. Naterer, Performance assessment of thermal man-
model agement systems for electric and hybrid electric vehicles, Int. J. Energy Res. 37
(2013) 1e12.
[4] E. Sahraei, R. Hill, T. Wierzbicki, Calibration and finite element simulation of
Hill'48 anisotropic yield surface with associated flow rule and pouch Li-ion batteries for mechanical integrity, J. Power Sources 201 (2012)
isotropic hardening law was used to model the shell casing 307e321.
material. [5] E. Sahraei, J. Campbell, T. Wierzbicki, Modeling and short circuit detection of
short circuit in 18,650 Li-ion cells under mechanical abuse conditions, J. Power
    Sources 220 (2012) 360e372.
f s; εp ¼ shill ðsÞ  k εp ¼ 0 (A-1) [6] L. Greve, C. Fehrenbach, Mechanical testing and macro-mechanical finite

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2  2 ffi
shill ¼ F sy  sz þ Gðsz  sx Þ2 þ H sx  sy þ 2Lt2yz þ 2Mt2zx þ 2Nt2xy (A-2)

element simulation of the deformation, fracture, and short circuit initiation of

cylindrical lithium ion battery cells, J. Power Sources 214 (2012) 377e385.
[7] Sahraei E, Wierzbicki T, Hill R, Luo M. Crash safety of lithium-ion battery to-
wards development of a computational model. SAE International. Paper
number 2010-01-1078.
shill denotes the Hill'48 equivalent stress, εp is the corresponding [8] Y. Xia, T. Wierzbicki, E. Sahraei, X. Zhang, Damage of cells and battery pack
work-conjugate equivalent plastic strain, s is Cauchy stress tensor, due to ground impact, J. Power Sources 267 (2014) 78e97.
and k is deformation resistance, which is a function of εp . x, y and z [9] Y. Bao, T. Wierzbicki, A comparative study on various ductile crack formation
criteria, J. Eng. Mater. Technol. 126 (2004) 11.
is the direction along axial, transverse and thickness direction of
[10] Y. Bai, T. Wierzbicki, A new model of metal plasticity and fracture with
the flattened sheet. The Hill'48 parameters, F ~ H, were calculated pressure and lode dependence, Int. J. Plast. 24 (2008) 1071e1096.
from Lankford ratio by assuming the associated flow rule and [11] Y. Bai, T. Wierzbicki, Application of extended MohreCoulomb criterion to
ductile fracture, Int. J. Fract. 161 (2010) 1e20.
planar anisotropy, as shown in Table A.1. The normality condition
[12] M. Luo, M. Dunand, D. Mohr, Experiments and modeling of anisotropic
for Hill’48 is written as aluminum extrusions under multi-axial loadingdpart II: ductile fracture, Int.
J. Plast. (2012) 36e58.
vf [13] M. Dunand, D. Mohr, Hybrid experimental-numerical analysis of basic ductile
dεp ¼ dεp (A-3) fracture experiments for sheet metals, Int. J. Solids Struct. 47 (2010)
vs 1130e1143.
Lankford ratio is defined by r ¼ dεpwidth =dεpthickness. The logarith- [14] K. Wang, T. Wierzbicki, L. Greve, FE simulation on edge fracture considering
pre-damage from punching process, Int. J. Solids Struct. (2015) (under
mic plastic strain along tensile and transverse direction can be review).
determined by using DIC, Poisson's ratio and Young's modulus. [15] K. Wang, M. Luo, T. Wierzbicki, Experiments and modeling of edge fracture for
Assuming that the plastic incompressibility, the plastic strain along an AHSS sheet, Int. J. Fract. 187 (2014) 245e268.
[16] K. Pack, M. Luo, T. Wierzbicki, Sandia fracture challenge: blind prediction and
thickness was then obtained. The summary of the Lankford ratio is full calibration to enhance fracture predictability, Int. J. Fract. 186 (2014)
listed in Table A-2. 155e175.
[17] S.H. Kim, S.H. Kim, H. Huh, Tool design in a multi-stage drawing and ironing
process of a rectangular cup with a large aspect ratio using finite element
analysis, Int. J. Mach. Tool. Manuf. 42 (2002) 863e875.
References [18] V.S. Deshpande, N.A. Fleck, Isotropic constitutive models for metallic foams,
J. Mech. Phys. Solids 48 (2000) 1253e1283.
[1] D. Doughty, SAE J2464 “EV & HEV Rechargeable Energy Storage System (RESS) [19] Abaqus, Reference Manual v6.10, Dassault Systems Simulia Corp, 2010.
Safety and Abuse Testing Procedure”, 2010. SAE Technical Paper 2010-01-