Electrochimica Acta 256 (2017) 172–184

Contents lists available at ScienceDirect

Electrochimica Acta
journal homepage: www.elsevier.com/locate/electacta

Research Paper

Multiphysics computational framework for cylindrical lithium-ion

batteries under mechanical abusive loading
Binghe Liua,b , Hui Zhaoc,d , Huili Yuc,d , Jie Lic,d , Jun Xua,b,e,*
a
Department of Automotive Engineering, School of Transportation Science and Engineering, Beihang University, Beijing, 100191, China
b
Advanced Vehicle Research Center (AVRC), Beihang University, Beijing, 100191, China
c
Chongqing Changan Automobile Co. Ltd., Chongqing, 400023, China
d
State Key Laboratory of Vehicle NVH and Safety Technology, Chongqing, 401120, China
e
State Key Laboratory of Explosion Science and Technology, Beijing Institute of Technology, Beijing, 100081, China

A R T I C L E I N F O A B S T R A C T

Article history:
Received 7 August 2017 Lithium-ion batteries (LIBs) are now widely applied to electric vehicles, such that the inevitable
Received in revised form 29 September 2017 mechanical abuse safety problem during possible vehicle accidents has become a prominent barrier. This
Accepted 6 October 2017 study initially proposes a multiphysics computational framework model that couples mechanical,
Available online 7 October 2017 thermal, and electrochemical models to describe the complete process for a single 18650 LIB cell
subjected to abusive mechanical loading from initial deformation to the final thermal runaway. The
designed experiments reveal the proposed model’s suitable agreement with the established multiphysics
model. Parametric studies in terms of governing factors, such as state of charge, loading speed, and
deformation displacement, are conducted and discussed. These studies reveal the underlying mechanism
for the mechanical abuse safety of LIBs. This model can lay a solid foundation to understand the
electrochemical and mechanical integrity of LIBs, as well as provide critical guidance for battery safety
designs.
© 2017 Elsevier Ltd. All rights reserved.

1. Introduction LIBs under compression. However, although these pioneering

Lithium-ion batteries (LIBs) have been regarded as one of the
most promising alternative energy sources in the automotive
industry [1–3]. Extreme conditions, such as mechanical abuse
[4,5], thermal abuse [6,7], and charging/discharging abuse, [8–10]
can lead to irreversible and catastrophic consequences (e.g., fires or
explosions) [11]. Thus, LIB safety has received extensive attention,
particularly in terms of hazardous consequences caused by
mechanical abusive loading during inevitable vehicle accidents
[12,13].
Given the complicated multiphysics nature of LIBs, modeling
LIB behavior under mechanical abusive loading is considered
difficult. Therefore, models for mechanical abuse conditions have
been gradually developed in the past several years. From the
mechanical aspect, macroscopically homogeneous and detailed
models are both employed to describe the mechanical behavior of
LIBs under mechanical loading. Greve and Fehrenbach [14] first
utilized the isotropic model to predict the mechanical behavior of
* Corresponding author at: Department of Automotive Engineering, School of
Transportation Science and Engineering, Beihang University, Beijing, 100191, China.
E-mail address: junxu@buaa.edu.cn (J. Xu).
models can predict compressive response correctly, they exhibit a
large discrepancy in bending and indentation conditions. Xu et al.
[15] developed an anisotropic model coupled with state of charge
(SOC) and dynamic effects to predict complex loading conditions
and overcome the said problem effectively. Wang et al. [16]
established a finite element model based on clay-like mechanical
properties. Zhang et al. [17] and Mehdi and Avdeev [18] built a
detailed model by modeling each component (e.g., cathode, anode,
and separator) separately. Moreover, Liu et al. [4] combined the
concept of homogeneous and detailed models to describe the
penetration damage in LIBs; they utilized the detailed model
within the vicinity of the penetrator and adopted the homoge-
neous model in other parts.
From the electrochemistry aspect, the 1D battery model was
first proposed by Newman et al. [19]. This model has since been
developed in recent years [20–22]. The electrochemical properties
of the anode [23–25], cathode [25–27], electrolyte [28], and
separator [24,25] have been comprehensively studied to improve
this battery model in recent years. The Arrhenius-type equation
was modeled to reflect the relationship between chemical reaction
rate and temperature [10,29]. The solid electrolyte interphase (SEI)
decomposition [6,30,31], anode–electrolyte reaction [6,30,31],

https://doi.org/10.1016/j.electacta.2017.10.045
0013-4686/© 2017 Elsevier Ltd. All rights reserved.
 B. Liu et al. / Electrochimica Acta 256 (2017) 172–184 173

qdf heat flux by radiation (W/m2)

Nomenclature
R radius (mm)
Rg gas constant
Rr resistance (V)
t+ transfer data
A Arrhenius constants (1)
T temperature (K)
Aa area (mm2)
T temperature field
b weighing coefficient
Tamb external temperature (K)
c concentration (mol/m3)
a conversion degree (1)
csei dimensionless amount of lithium-containing meta-
a0 initial value of the conversion degree (1)
stable species in the solid electrolyte interphase
am ratio of tension and compression (1)
(SEI) (1)
r0 initial density (kg/m3)
csei0 initial value of the dimensionless amount of
r density (kg/m3)
lithium-containing metastable species in the SEI (1)
s B–D Stefan–Boltzmann constant (5.67  108 W/(m2 
ca dimensionless amount of lithium intercalated with-
K4))
in carbon (1)
s0 yield stress (MPa)
ca0 initial value of the dimensionless amount of lithium
sp flow stress (MPa)
intercalated within carbon (1)
s1 maximum principal stress (MPa)
ce dimensionless concentration of electrolytes (1)
s2 middle principal stress (MPa)
ce0 initial value of the dimensionless concentration of
s3 minimum principal stress (MPa)
electrolytes (1)
s eq equivalent stress of unified strength theory (MPa)
Cp heat capacity (J/(kg
K)) Unif ied
ep plastic strain (1)
dz thickness of 2D model/LIB length (mm)
d thickness (mm)
D diffusion coefficient (m2/s)
e volume fraction
Deff modified diffusion coefficient (m2/s)
e strain (1)
E Young’s modulus (MPa)
eDf surface emissivity (1)
Ea experimental activation energy (J/mol)
k electrical conductivity (S/m)
F Faraday’s constant (9.64853  104 C/mol)
keff modified electrical conductivity (S/m)
FV body force (N)
n Poisson’s ratio
FL deformation gradient
f potential (V)
H volume-specific carbon content (g/m3) dlnf
1 þ dlnc molar activity coefficient
heat transfer coefficient (W/(m2
K))
l
h
V(t) area of the 2D model (LIB’s section) with time (mm2)
I current (A)
i current density (A/m2)
~ Subscripts
j current density (A/m2)
eff modified
k thermal conductivity (W/(m
K))
eq equilibrium
L length (mm)
S Piola–Kirchhoff stress tensor
Superscripts
Sst short-circuit area (mm2)
a anode
Sst normalized short-circuit area (mm2)
a–e anode–electrolyte reaction
SOC state of charge value
c–e cathode–electrolyte reaction
t time (s)
c cathode
t+ transfer data
Cc Current collector
u displacement field
e electrolyte decomposition
v velocity field
D diffusion
Wc volume-specific carbon content (g/m3)
LB lithium-ion battery
Wp volume-specific positive active content (g/m3)
l electrolyte
We volume-specific electrolyte content in the jellyroll
Me mechanical model
(g/m3)
nc negative collector
X material coordinate
ng negative electrode
x spatial coordinate
nl penetration nail
z dimensionless measure of the SEI layer thickness (1)
pc positive collector
z0 reference dimensionless measure of the SEI layer
ps positive electrode
thickness (1)
sei SEI decomposition reaction
Q heat sources (W/m3)
s electrode
Qs additional heat sources (W/m3)
se separator
Qa reaction heat (W/m3)
St short-circuit model
Qj joule heat (W/m3)
Tm thermal model
Qr resistance heat from current collectors (W/m3)
1D 1D battery model
Qi irreversible heat (W/m3)
Q1D heat for 1D battery model (W/m3)
Qst heat from the short-circuit model (W/m3)
Qtr heat from the thermal runaway model (W/m3) cathode–electrolyte reaction [6,30,31], electrolyte decomposition
QTOT total heat rate (W) reaction [6,30,31], and electrochemistry reaction [30] have been
q heat flux (W/m2) discussed extensively and concluded in the thermal runaway
qfl heat flux by conduction (W/m2) model. Ohm’s law is generally employed in the short-circuit model
174 B. Liu et al. / Electrochimica Acta 256 (2017) 172–184
[17,26,32] to represent basic electricity behavior. Equivalent
resistance is commonly set as a constant [32–34]. Zhang et al.
[17] and Xu et al. [4] recently utilized a varying short-circuit
resistance model related to several physical statuses, such as
separator failure and deformation. The core equation for the
thermal model is the thermal equilibrium equation. Heat sources
can be extracted from the battery model [24,35], short-circuit
model [17,26,32,34], and thermal runaway model [6,30,31,36].
Heat dissipation in the thermal model requires the consideration
of heat radiation [30,32], heat convection [30,32], and heat
dissipation from venting [30].
Pioneering efforts to couple several of the aforementioned
models have been recently observed. Xu et al. [15] established the
relationship between a mechanical model and SOC, and they
further suggested the combination of mechanical, 1D battery, and
thermal models to describe LIB behavior under nail penetration
[4]. The stress state [14,37], strain state [17], or material failure
criteria [4] are applied to couple a mechanical model with a short-
circuit model, whereas short-circuit, thermal runaway, and
thermal models are often coupled with temperature and heat
[30,35,36,38]. NREL initiated the topic of multiphysics modeling to
improve LIB safety [39]. The multiphysics model can provide a
reliable method to address issues, reduce the numbers of build–
test–break prototypes, and accelerate the development cycle of
generating products [39].
However, no computational model can describe the multi-
physics process of cylindrical LIB behavior under mechanical abuse
loading. The current study attempts to build a framework model by
incorporating five sub-models, namely, the mechanical, battery,
short-circuit, thermal runaway, and thermal models. The rest of
this paper is organized as follows. Section 2 introduces the LIB core
process under the large compression and modeling strategy.
Section 3 presents the validation of the established model with
designed experiments. Section 4 describes the parametric studies
conducted, including SOC, compression speed, and compression
displacement. Section 5 presents the conclusions.
2. Method
2.1. Underlying multiphysics behavior of LIB under large compression
The typical behavior of cylindrical LIB cells subjected to
compression mainly includes mechanical deformation, triggering
of an internal short circuit, and leading of thermal runaway in time
sequence (Fig. 1). The outer shell of the battery cell first bears
loading along with the jellyroll upon compression [15,40]. After the
buckling of the battery shell, the jellyroll starts to play a key role in
maintaining the load-bearing capability. The cathode, anode, or
Fig. 1. Diagram of the compression process.
separator can mechanically fail (e.g., triggering the internal short
circuit) when the deformation evolves.
Four short-circuit modes exist, namely, Ca–An (contact of
cathode and anode active materials), Al–An (contact of aluminum
foil and anode), Ca–Cu (contact of cathode and copper foil), and Al–
Cu (contact of aluminum and copper foils) short circuit [26,41].
Once short circuit occurs, joule heat is generated and leads to a
drastic temperature increase. At the same time, the voltage drops
sharply [40]. Once the LIB temperature reaches a critical value, the
thermal runaway is activated, at which point chemical reactions
are generally involved within the cell (e.g., SEI decomposition,
reaction between negative active material and electrolytes,
reaction between positive active material and electrolytes,
electrolyte decomposition, and reaction between negative active
material and binder) [38,42]. The first stage of the thermal
runaway is the decomposition of the SEI metastable components
[e.g., polymers, ROCO2Li, (CH2OCO2Li)2, and ROLi] at 90  C to 120  C
[38,42]. As temperature increases, lithium reaction with the
solvents occurs, typically starting at 100  C (or at 68  C for special
electrolytes) [38,42]. The separator starts to melt at approximately
130  C, thereby triggering short circuit between the electrodes. The
metal oxide cathode material then starts to break down at elevated
temperatures. Approximately 1,000 J/g heat is produced at this
stage [33–37]. CO2, PF5, and POF3 are then generated from 200  C to
240  C [43], followed by the decomposition of the electrolytes at
approximately 200  C to 300  C [44–46]. The pressure inside the
battery shell becomes extremely high after this stage and can cause
a possible explosion. Other reactions also occur during the thermal
runaway [47]; they include the reaction of lithiated carbon with
fluorinated binder. The reaction functions during thermal runaway
are summarized in Table 1.
2.2. Modeling strategy
Therefore, five sub-models (i.e., 1D battery, short-circuit,
thermal, thermal runaway, and mechanical models) are included
in the integrated model and introduced as follows.
(1) Mechanical model
The mechanical model describes the mechanical behavior of LIB
under mechanical loadings (e.g., structure deformation, stress
state, and strain state). An anisotropic mechanical model, including
jellyroll, winding nail, and shell, was built in the previous study
[15]. The homogenized method has been successfully used in the
mechanical modeling of batteries (e.g., Refs. [14,15,48]). However,
without the loss of accuracy compared with that in Refs.
[14,15,31,48], the further homogenization of LIBs is regarded as
complete in the present study. A simplified 2D mechanical model is
proposed to reach a rapid computational convergence and lower
the computation cost.
 B. Liu et al. / Electrochimica Acta 256 (2017) 172–184 175

Table 1
Reactions during thermal runaway.

Reaction Reaction equation Note

Decomposition of SEI metastable components ðCH2 OCO2 LiÞ2 ! Li2 CO2 þC2 H4 þCO2 þ12O2 This stage starts at 90  C–120  C.
[38,42]
Solvent reaction with lithium [38,42] 2Li þ C2 H4 O3 ðECÞ ! Li2 CO3 þC2 H4 This stage starts at approximately 100  C (or 68  C for
2Li þ C4 H6 O3 ðPCÞ ! Li2 CO3 þC2 H4 special electrolytes).
2Li þ C3 H6 O3 ðDMCÞ ! Li2 CO3 þC2 H6
Breakdown of metal oxide cathode material Lix CO2 ! xLiCoO2 þ13ð1  xÞCo3 O4 þ13ð1  xÞO2 Different temperatures for different cathodes.
(e.g., LiCoO2) [33–37] Co3 O4 ! 3CoO þ 0:5O2 CoO ! Coþ0:5O2
CO2, PF5, and POF3 formation [43] 2Li þ 2EC ! LI  O  ðCH2 Þ4 O  Li þ CO2 These reactions start at 200  C–240  C.
LiPF6 ! LiF þ PF5
Electrolyte decomposition [44,45] 2ECLi  O  ðCH2 Þ4 O  Li þ PF5 ! Li  O  ðCH2 Þ4 þ2LiF þ POF3 These reactions start at 200  C–300  C.
C2 H5 OCOOC2 H5 þPF5 ! C2 H5 OCOOPF4 þHF þ C2 H4
C2 H4 þHF ! C2 H5 F
C2 H5 OCOOPF4 ! PF3 O þ CO2 þC2 H4 þHF
C2 H5 OCOOCPF4 ! PF3 O þ CO2 þC2 H5 F
C2 H5 OCOOCPF4 þ HF ! PF3 O þ CO2 þC2 H5 F
Reaction of lithiated carbon with fluorinated CH2 CF2  þ Li ! LiF þ CH ¼ CF  þ12H2 These reactions occur when the temperature
binder [47] 2Li þ RF2 ! 2LiF þ 0:5R2 exceeds 260  C.

The main equation of the mechanical model is given by stress of LIB. The constitutive equations to describe the mechanical
Newton’s second law: behavior of a battery cell are established by fitting the results of the
experiment data from Ref. [15].
@2 u The comparison of the experiment and simulation results is
r ¼ rX ðF L SÞ þ FV ; ð1Þ
@t2 presented in Fig. 2(a), which shows a reasonably good agreement
where r is density, u is the displacement field, X is the material between the two sets of results. However, presenting homogenized
coordinate, FL is the deformation gradient, S is the Piola–Kirchhoff models to provide a perfect fitting is difficult due to the complex
stress tensor, and FV is the body force. winding structure within the battery. The model is sufficient in
The key problem of the mechanical model is the equivalent predicting the short-circuit position and time [14,37,49].
model of LIBs. The elastic–plastic model is selected for the current (2) Battery model
study to predict the mechanical properties of LIBs. Young’s The battery model is only used to describe the voltage change
modulus, yield stress, and hardening curve are related to SOC and heat generation under discharging conditions (i.e., short
and they can be calculated as follows: circuit). The classical 1D battery model was first established by
Newman et al. [19]. This model can accurately predict voltage
ELB ¼ 340e0:4856SOC ; ð2Þ change and heat generation under charging/discharging condi-
tions. Obviously, this model is sufficiently accurate for the current
model. At the same time, battery heat generation is trivial
s LB
0 ¼ 4:148e
0:4856SOC
; ð3Þ compared with short-circuit heat generation. Therefore, this
model is selected to represent the average electrochemistry and
thermal properties of LIBs. The mass balance and charge balance of
p ¼ 22000ep
s LB LB6:5 0:8SOC
þs LB electrolytes dominate this model, as expressed in [50,51]
e 0 ; ð4Þ
l
where ELB is the Young’s modulus of LIB, SOC is the value of SOC; s LB @cl i tþ
p el ¼ Dlef f rcl þ ; ð5Þ
is the flow stress of LIB; eLB LB @t F
p is the plastic strain of LIB and s 0 is yield

Fig. 2. (a) Comparison of simulation result through the equivalent model with the experiment result in [15]. (b) Comparison of the simulation result through the battery
model with the charge/discharge experiment.
176 B. Liu et al. / Electrochimica Acta 256 (2017) 172–184

Table 2
Main parameter values of models.


l l
2klef f Rg T dlnf
i ¼ klef f rf þ 1þ ð1  tþ Þrlncl ; ð6Þ Parameters Value Source
F dlncl A1Da 81,900 mm2 [4]
Aa–e 2.5e13 (s1) [6,31,36]
where e , c , l l
Dlef f , l
i, k l
ef f
, l
and f are the volume fraction, Li +
Ae 5.14e25 (s1) [6,31,36]
concentration, modified diffusion coefficient, current density, Asei 1.167e15 (s1) [6,31,36]
modified electrical conductivity, and potential of the electrolyte, Ace 6.667e13 (s1) [6,31,36]
dlnf ca0 0.75 [6,31,36]
respectively; t+ is the transfer data; 1 þ is the molar activity
dlncl ce0 1 [6,31,36]
coefficient; Rg is the gas constant; and F is Faraday’s constant. csei0 0.15 [6,31,36]
Detailed information about this model can be found in Ref. [19]. caeq 31507mol=m
3 [23–25]
Heat generation includes produced reaction, joules, resistance of cceq 56250mol=m
3 [27,53]
current collector, polarization (not considered in this study), and C LB 1347:4J=ðkg
KÞ [4]
irreversible heat [50,52], The average heat of the battery model can dz 65 mm measured
be expressed as follows [50]: Dc 5  1013 [27,53]

Z L Dse Function of cLiPF6 and T [28,34]

s l Da Function of T [23–25]
Qa ¼ ð iðf  f  Eeq ÞdxÞ=ðLps þ Lng þ Lse Þ; ð7Þ
0
hLB 5 W/(m2K) [30]
Hsei 257J=g [6,31,36]
Ha 1714J=g [6,31,36]
Z s l Hc 314J=g [6,31,36]
L
df df He 115J=g [6,31,36]
Qj ¼ ksef f ð Þ2 þ klef f ð Þ2
0 dx dx k
LB 13W=ðm
KÞ estimated
D dlncl dfl RLB 9 mm measured
þ kef f ð Þð Þdx=ðL þ Lng þ Lse Þ;
ps
ð8Þ
dx dx z0 0.033 [6,31,36]
Wc 6.14e5 (g/m3) [6,31,36]
WP 1.221e6 (g/m3) [6,31,36]
We 4.069e5 (g/m3) [6,31,36]
RCc
Q r ¼ I2 r
; ð9Þ ka 10S=m [26]
A1D
a ðL
ps
þ Lng þ Lse Þ kse Function of cLiPF6 and T [28,34]
ka 100S=m [23–25]
a0 0.04 [6,31,36]
Lps 0.073 mm [4]
Q i  0; ð10Þ Lng 0.073 mm [4]
Lse 0.018 mm [4]
where I is the current equal to iA1D
a , L is the length of the 1D battery eDf 0.8 [30]
model, fs is the potential of the electrode, and RCc
r is the resistance
of the current collector. The total heat of the system is as follows:
Q 1D ¼ Q a þ Q j þ Q r þ Q i ; ð11Þ

where Qr is the heat from the current collectors. This variable is

difficult to compute accurately with the 1D battery model because
the resistance of the current collector Rr is related to the short- q ¼krT; ð13Þ
circuit position and short-circuit area. The electrical conductivities where dz is the thickness of the 2D model/LIB length, v is the
of copper and aluminum are approximately 3e7 S/m to 6e7 S/m,
velocity field, rLB is the density of LIB, C LB
p is the heat capacity of LIB,
and the typical size of the collector is 10 um  58.5 mm  700 mm.
The heat from the current collector is sufficiently small to be and Qs contains the following additional heat sources:
ignored. Thus, positive and negative current collectors are ignored Q s ¼Q 1D þ Q st þ Q tr ; ð14Þ
in this study.
The 1D battery model utilizes half of the thickness of the where Q1D is the heat from the battery model, Qst is the heat from
cathode and anode, as well as the entire thickness of the separator. the short-circuit model, and Qtr is the heat from the thermal
The cross section of the model is twice that of the unfolded area of runaway model. q is the heat flux written as follows:
the jellyroll. The parameters of the 1D battery model are q ¼ qf l þ qdf ; ð15Þ
summarized in Table 2.
Charge/discharge experiments with charge/discharge rates of where qfl is the heat flux by conduction, and qdfis the heat flux by
0.3C and 1C are developed using BK6808AR rechargeable battery radiation. They are written as follows:
performance testing equipment (BLUE-KEY). The battery model LB
n
qf l ¼ h ðT amb  TÞ; ð16Þ
matches well with the experiment, as shown in Fig. 2(b).
(3) Thermal model
The thermal model calculates the heat transfer. The geometric
boundary of the thermal model is determined by the mechanical n
qdf ¼ eDf s BD ðT 4amb  T 4 Þ; ð17Þ
model. That is, the displacement field x(t) is computed through the
where hLB is the heat transfer coefficient, Tamb is the external
mechanical model.
temperature,eDf is surface emissivity, and s BD is the Stefan–
The main equation of the thermal model is as follows:
Boltzmann constant. The model parameters are also summarized
@T in Table 2.
dz rLB C LB
p þ dz rLB C LB
p vrT þ r
dz q ¼dz Q s ; ð12Þ
@t (4) Short-circuit model
The short-circuit model confirms and calculates the short-
circuit time, position, and heat. Mechanical abuse can cause
 B. Liu et al. / Electrochimica Acta 256 (2017) 172–184 177

internal short circuit. The main equations for the resistive heating area, which is calculated as follows:
model [4] are listed as follows: 
RSt
r ¼ zð1  Sst Þ; ð27Þ
~
j ¼ kSt f; ð18Þ
where parameter z is a fitting parameter. The parameterized
scanning of z has been performed in simulations. By comparing
experiment and simulation results, the value of z is confirmed as
~
E ¼ f; ð19Þ 0.5.
(5) Thermal runaway model
The thermal runaway model calculates the heat caused by
~
j
~
j thermal runaway, which is determined with an Arrhenius-type
Q st ¼ St
; ð20Þ
k equation. SEI decomposition is expressed as follows:

where ~ j is the current density, f is the potential, ~

E is the electrical Esei
field, and kSt is the electrical conductivity of the short-circuit part. Q sei ¼ Hsei W c Asei exp½ a
csei ; ð28Þ
Rg T
The short-circuit parts are equivalent to those of a resistor. The
trigger point of the short circuit is set through stress. A previous
study [37] utilized unified strength theory as the short-circuit
dcsei Esei
criterion, which is written as follows: ¼ Asei exp½ a
csei : ð29Þ
dt Rg T
8
> am s þa s
< s1  ðbs 2 þ s 3 Þ; s2  1 m 3 The anode–electrolyte reaction is described as follows:
eq
s Unif ied ¼ 1þb 1þa ; ð21Þ
: 1 ðs 1 þbs 2 Þ  am s 3 ; s 2 s 1 þam s 3
> z Eae
1þb 1þa Q ae ¼ Hae W c Aae exp½
exp½ a
ca ; ð30Þ
z0 Rg T
eq
where s Unif ied
is the equivalent stress of unified strength theory,
s k(k = 1, 2, 3) is the principle stress, b is the weighting coefficient,
and am is the ratio between tension and compression. The second dca z Eae
¼ Aae exp½
exp½ a
ca ; ð31Þ
stress and difference between tension and compression in the dt z0 Rg T
compression loading are neglected, such that b is selected as 0 and
am is selected as 1 in the current study. The criterion in Eq. (21) is
reduced as follows: dz z Eae
¼ Aae exp½
exp½ a
ca : ð32Þ
dt z0 Rg T
s eq
Unif ied
¼ s1  s3: ð22Þ
The cathode–electrolyte reaction is shown as follows:
The s eq
Unif ied
value is confirmed by comparing the experiments
Ece
and simulating the mechanical model, which is s eq
Unif ied
¼ 43MPa, Q ce ¼ Hce W p Ace ace ð1  ace Þexp½ a

; ð33Þ
Rg T
which can well predict the short-circuit time. Then, the short-
circuit position can be determined by the following equation:
xq ðtÞ ¼ xðtÞ
cðs 1 ðtÞ  s 3 ðtÞÞ; ð23Þ dace Ece
¼ Ace ace ð1  ace Þexp½ a
: ð34Þ
dt Rg T
where xq(t) is the spatial coordinate of the short-circuit part, x(t) is
the spatial coordinate, and c(s 1  s3) is the critical function, which The electrolyte decomposition reaction is written as follows:
is expressed as follows:
( Eea
Q e ¼ He W e Ae exp½
ce ; ð35Þ
0s 1  s 3 < s eq
Unif ied Rg T
cðs 1  s 3 Þ ¼ : ð24Þ
1s 1  s 3 s eq
Unif ied

The short-circuit area Sst can then be integrated as follows: dce Ee
ðð ¼ Ae exp½ a
ce ; ð36Þ
dt Rg T
Sst ¼ cðs 1  s 3 Þds; ð25Þ
VðtÞ where Qsei, Qa-e, Qce, and Qe denote heat generation; Hsei, Hae,
Hce, and He are heat transfer coefficients; Asei, Aae, Ace, and Ae
where V(t) is the deformed area of the 2D model during are Arrhenius constants; Esei ce ce e
a , Ea , Ea , and Ea are the experimental
compression (V(0) denotes the area of the circle with the radius activation energy values; ca is the dimensionless amount of lithium
of RLB, where RLB is the radius of LIB). It can further be normalized intercalated within carbon; ca0is the initial value of the dimen-
as follows: sionless amount of lithium intercalated within carbon; ce is the
Sst dimensionless concentration of electrolytes; ce0 is the initial value
Sst ¼ 2
: ð26Þ of the dimensionless concentration of electrolytes; Wc, Wp, and We
pRLB are the active content densities; and Rg is the gas constant.
Obtaining the theoretical value for short-circuit resistance is The aforementioned reactions are included in the thermal
difficult because of the various factors influencing the resistance runaway model [6,31,36]. The total heat in the thermal runaway
(e.g., separator shutdown near the melting point, changing model Qtr is the summation of that in former models.
temperature, and large deformation). These factors are also
Q tr ¼ Q sei þ Q ae þ Q ce þ Q e ð37Þ
difficult to quantify and measure. Therefore, an empirical function
is used to describe the short circuit resistance value. Short-circuit The parameters of LiCoO2 LIB were well calibrated in [6] and
resistance is then assumed to be linearly related to the short-circuit further applied in [31,36]. These parameter values are summarized
in Table 2.
178 B. Liu et al. / Electrochimica Acta 256 (2017) 172–184

Fig. 3. Coupling method and schematic algorithms for the models.

2.3. Coupling strategy same as the temperature field of the thermal model. The input
average temperature of the 1D battery model is supplied by the
The aforementioned five models are coupled with one another average temperature of the thermal model. The relationships are as
through the following relationship. The schematic is shown in follows:
Fig. 3.
(1) Mechanical and thermal models T Tm ðtÞ ¼ T Tr ðtÞ; ð40Þ
The mechanical model can compute LIB deformation, leading to
a change of heat transfer. At the same time, the temperature
computed by the thermal model can cause the material properties T Tm 1D
a ðtÞ ¼ T a ðtÞ; ð41Þ
of the mechanical model to change. Thus, these two models can be Tm Tr
where T (t) and T (t) are the temperature fields of the thermal
coupled through deformation and temperature as follows:
model and thermal runaway model, respectively; and T Tm
a ðtÞ and
vMe ðtÞ ¼ vTm ðtÞ; ð38Þ T 1D
a ðtÞ are the average temperatures of the thermal model and 1D
battery model, respectively.
However, the short-circuit and thermal models are not coupled
T Me ðtÞ ¼ T Tm ðtÞ; ð39Þ through temperature in this study. The influence of temperature on
Me Tm the short resistance can be ignored in this case [17].
where v (t) and v (t) are the velocity fields of the mechanical
The heat sources of the thermal model come from the 1D
and thermal models, respectively; and TMe(t) and TTm(t) are the
battery model, thermal runaway model, and short-circuit model.
temperature fields of the mechanical and thermal models,
The coupling equation is shown in Eq. (13).
respectively.
(4) 1D battery and short-circuit models
The thermal softening effect is not considered in the mechani-
The 1D battery model and short-circuit model are coupled
cal model in the present study because the compression duration is
through the cell voltage and current density of the battery model,
short.
as well as through the voltage and current density of the short-
(2) Mechanical and short-circuit models
circuit model.
These two models are one-way coupled through the stress state
of the mechanical model. The short-circuit position, resistance, and U 1D ¼ U St ; ð42Þ
area can be determined by the mechanical model, and variable
coupling is introduced in Eqs. (21) to (27) in Section 2.2.
(3) Short-circuit, thermal runaway, and 1D battery–thermal A1D i
1D
¼ ISt ; ð43Þ
models
1D St
The short-circuit, thermal runaway, and 1D battery models are where U is the voltage of the 1D battery model; U is the voltage
coupled with the thermal model through temperature and heat. of the short-circuit model; A1D is the section area of the 1D battery
The input temperature field of the thermal runaway model is the
 B. Liu et al. / Electrochimica Acta 256 (2017) 172–184 179

Fig. 4. Experimental setup, including the testing machine Instron 8801 (for mechanics), Agilent 34410A (for electrochemistry), and thermal couples (for temperature). A
single battery cell is placed within the two rigid fixtures.

model; i1D is the current density of the 1D battery model and ISt is 2.3 GHz were used in the simulation, and the total simulation time
the current of the short-circuit model. lasted for approximately 20 min.

2.4. Simulation method 2.5. Experimental setup

The framework model was built based on COMSOL Multiphysics
platform. For the mechanical model part, “Free Quad” elements
(1,119 in total) were adopted to mesh the geometry. Two rigid
bodies were put on both sides of the battery; one was fixed, and the
other one controlled the displacement boundary conditions. A
penalty-based contact was set for the contact pairs with a friction
coefficient of 0.1. The same geometry and mesh were used for the
thermal model, thermal runaway model, and short-circuit model.
These models were coupled together via variables in equations.
The time stepping method was a backward differentiation formula,
and the solver was set for the PARDISO method. Eight cores at
Commercial 18650 LIBs provided by SONY, with a capacity of
2,200 mAh, were used in this study. Compression experiments
were designed and conducted to calibrate the model. An INSTRON
8801 hydraulic servo testing machine was selected to provide a
satisfactory mechanical testing platform for the tests with a
maximum load of 100 kN and enhanced resolution of 50 N. The
loading speed was set to 2 mm/min to provide a quasi-static
mechanical loading condition. An Agilent 34410A digital voltmeter
was employed to measure the voltage with a frequency of
10,000 Hz and accuracy of 0.01 mV. A thermocouple was selected
for temperature measurement. The red and blue wires shown in

Fig. 5. (a) Comparison of simulation and experiment: voltage, increased temperature, and load. (b) Experiment photo, (c) Tresca stress, (d) short position, (e) temperature,
and (f) thermal runaway state of LIB at different times for low SOC LIBs. (g) Battery heat, short-circuit heat, and thermal runaway heat–time curves. (h) Comparison of the
simulation and experiment voltage curves for LIB with SOC = 0.1 and 0.4 (Ref. [40]).
180 B. Liu et al. / Electrochimica Acta 256 (2017) 172–184

Fig. 4 were devoted to the voltage and temperature measurements,
respectively.
Only the cells with low SOC = 0.1 (charging of 1 h with 0.1C)
were utilized for these experiments for safety reasons by
measuring force, temperature, and voltage in this study. This
experiment was designed to validate the mechanical, short-circuit,
and thermal models.
3. Results
Fig. 5(a) shows the typical result for a cell behavior under
compressive mechanical loading. The reaction force increased with
time at the start of the battery structure deformation. Force
dropped slightly at 200 s and suddenly at 250 s. The former drop of
force was caused by the buckling of the shell, and the latter one was
due to the fracture of the jellyroll. The same was reported in Refs.
[40,49]. The voltage was sustained at an almost constant rate,
except for a minor voltage increase, which is similar to that
reported in Ref. [40]. The voltage started to drop at approximately
211 s, and then it dropped to 0 gradually at about 270 s. The
temperature remained constant before 211 s, and the change in the
increased temperature (DT = T  Tamb)–time curve was observed
slightly later than that for the voltage–time curve because the heat
conduction may take up time.
In Fig. 5(a), the solid lines represent the computation results.
The general trend of the curves matched that in the experiment
and computation. However, some visible devastations were noted.
In particular, the first drop of the load did not appear in the
simulation result because the homogenized model of LIB cannot
describe the buckling of the shell, which is trivial for the entire
process; nevertheless, the overall physical deformation of the cell
was correctly described by the model. Thus, a satisfactory
agreement in computation and experiment results was observed
and it validated the accuracy of the suggested computation
framework model.
The computation model revealed other important information,
which cannot be measured directly or quantitatively (e.g., short-
circuit current, stress state, short-circuit heat positions, short-
circuit heat sources, and temperature state). The stress state and
short-circuit heat positions can be predicted by the mechanical
model. The short-circuit heat sources can be described by the
short-circuit model, and the temperature state can be revealed by
the thermal model. The coupling algorithm is shown in Fig. 3. The
cell was deformed under compression from 0 s to 211 s, and stress
was generated simultaneously. The maximum Tresca stress
reached the threshold value at approximately 211 s and 43 MPa,
which triggered internal short circuit. The short-circuit zone
appeared and gradually expanded after 211 s. Short-circuit heat
was then produced at this position, followed by an increase in
temperature. The temperature gradually dropped up to the
maximum increased temperature of 7 K. The battery capacity
dropped to a low value, and the short-circuit heat generated was
less than the heat dissipation. Thermal runaway behavior was not
triggered in such low SOC cells; thus, the thermal runaway heat
was not produced, as shown in Figs. 5(b)–(f). Fig. 5(g) shows the
summarization of the heat from the battery short circuit and
thermal runaway. The result indicates the significant contribution
of short-circuit heat in this entire process.
The framework model can be used to predict some extreme
conditions that obstruct experiment design. It can also guide the
safety design of single cells and even battery packs.

Fig. 6. Simulation results of (a) voltage–time curves for different SOC LIBs under compression; (b) increased temperature–time curves for different SOC LIBs under
compression; (c) Tresca stress; (d) short position; (e) temperature; (f) thermal runaway state of LIBs at different times for high SOC LIBs; (g) short-circuit–time curves for
different SOC LIBs under compression; (h) total heat rate from short circuit, thermal runaway, and battery for different SOC LIBs under compression; (i) relationship between
maximum temperature and SOC; and (j) relationship between the time of maximum temperature and SOC.
 B. Liu et al. / Electrochimica Acta 256 (2017) 172–184 181

4. Discussion
4.1. Simulation results for different SOCs
This model can be utilized to compute cell behavior with
different SOCs, which can be dangerous in reality. Fig. 6(a) shows
that the voltage drop point started earlier with high SOC cells due
to the high stress and easily reachable short-circuit triggering
point. The voltages of the low SOC cells dropped to 0 within a short
time, whereas the voltages of the high SOC cells remained
considerably longer after the short circuit, similar to the discovery
in a previous experimental study [40]. This result was further
proved by a compression test on LIB with SOC = 0.4 from Ref. [40]
(Fig. 5(h)).
The temperature rising behavior has two different modes due to
SOC differences [i.e., low SOC mode (LSM) and high SOC mode
(HSM)]. LSM occurs in cases with SOC < 0.9. After short circuit
starts in these cases, the temperature increases almost linearly
with time. The main heat sources are the short circuit and battery
heat. HSM occurs only when SOC > 0.9. A previous study [40]
designed the compression tests for cells with SOC from 0 to 0.7 and
did not observe any thermal runaway phenomenon. This outcome
provides further proof for the model validation in the present
study. After short circuit starts in this case, the temperature first
increases linearly with time. Once the temperature reaches a
critical value (i.e., approximately 180  C), then the temperature
increases sharply to the peak value. The main heat sources of the
linear stage are the short-circuit heat and battery heat. The main
heat source of the sharp increasing stage is the thermal runaway
heat. The maximum temperature (Tmax = DTmax + Tamb) and time of
the maximum temperature are extracted from the curve, as shown
in Fig. 6(i) and (j). The maximum temperature increases with SOC.
The increasing rate of the maximum temperature for LSM
increased with SOC because a high voltage produces a large
amount of heat. The maximum temperature for HSM is remarkably
larger than that for LSM due to additional thermal runaway heat.
The total heat rate from the thermal runaway model, battery
model, and short-circuit model is shown in Fig. 6(h). The time of
maximum temperature decreases with SOC from 0 to 0.4 and
increases from 0.4 to 0.8 for LSM. This scenario occurs because the
short-circuit starting time decreases with SOC, although high SOC
cells remain for a long time after a short circuit. The time of
maximum temperature for HSM decreases with SOC because it
reaches the thermal runaway trigger temperature early.
The stress and thermal statuses of LIB for LSM are described in
Section 3. The stress for HSM is easily accumulated once

Fig. 7. Simulation results of (a) voltage–time curves at different compression speeds; (b) increased temperature–time curves at different compression speeds; (c) short-
circuit current–time curves at different compression speeds; (d) total heat rate from short circuit, thermal runaway, and battery at different compression speeds; (e)
relationship between the maximum temperature and compression speed; and (f) relationship between the time of the maximum temperature and compression speed.
182 B. Liu et al. / Electrochimica Acta 256 (2017) 172–184

deformation starts. At a critical temperature (i.e., approximately
180  C), thermal runaway occurs. A large amount of heat is
generated within a short time, leading to a dramatic temperature
increase. The temperature reaches its peak value after some time
when the generated heat is balanced by the dissipated heat and
then decreases by heat dissipation due to the battery power drain.
These processes are shown in Fig. 6(c)–(f). An animation of the
entire process is also included in the supplementary material.
4.2. Simulation results for different compressive loading speeds
The compression speed from the quasi-static to the low-speed
domain is selected at this point (i.e., from 0.1 mm/s to 1 m/s), in
which the dynamic effect for the LIB mechanical properties can be
ignored [15]. The fully charged cells are within the compression
displacement of 8 mm.
Fig. 7(a) and (c) show the voltage and short-circuit current–
time results from simulation. The time axial is expressed in
logarithmic coordinates. The voltage drop time (short-circuit time)
is clearly related to the compression speeds; in particular, a fast
compression speed causes rapid deformation, which triggers an
early short-circuit time.
Fig. 7(b) presents the increased temperature–time results; the
time axial is expressed in logarithmic coordinates. The
temperature increasing point is consistent with the short-circuit
point. Fig. 7(d) shows the heat generation rate–time curve, which
indicates that the short-circuit heat and thermal runaway heat are
the main reasons that cause temperature increase. The relation-
ship between the maximum temperature and compression speeds
is shown in Fig. 7(e). The maximum temperature increases with
the compression speed from 0.1 mm/s to 10 mm/s and slightly
differs with the compression speeds after 10 mm/s. The time of
maximum temperature is also related to the compression speed
from 0.1 m/s to 10 mm/s and slightly differs with the compression
speeds after 10 mm/s, as shown in Fig. 7(f).
4.3. Simulation results for different compressive loading
displacements
Compression displacement is also an important indicator that
represents the severity of mechanical abuse loading. The fully
charged cells (SOC = 1) are utilized in this study with a compression
speed of 0.1 m/s.
Fig. 8(a), (b), and (c) show the voltage, temperature, and short-
circuit current–time curves, respectively. The short-circuit time at
this compression speed is 0.048 s, and the compression durations
are from 0.07 s to 0.082 s with compression displacement of 7 mm
to 8.2 mm. The voltage remains constant before the short circuit.

Fig. 8. Simulation results of (a) increased voltage–time curves at different compression displacements; (b) increased temperature–time curves at different compression
displacements; (c) short-circuit current–time curves at different compression displacements; (d) total heat rate from short circuit, thermal runaway, and battery at different
compression displacements; (e) relationship between the maximum temperature and compression displacement; and (f) relationship between the time of maximum
temperature and compression displacement.
 B. Liu et al. / Electrochimica Acta 256 (2017) 172–184
After short circuit occurs, the voltage and increased temperature
are the same from 0.048 s to 0.07 s. Relative to the entire process,
the deformation stage takes such a short time that it cannot be
distinguished in the figures. Once compression is finished, the
voltage drops and temperature increasing rates are large for a large
compression displacement. The main difference in compression
displacement is the short-circuit resistance and short-circuit areas.
The temperature increase is mainly caused by the short-circuit
model and thermal runaway model, and the compression
displacements affect the heat generation rates from the short-
circuit model, thermal runaway model, and battery model, as
shown in Fig. 8(d).
Fig. 8(e) shows the relationship between the maximum
temperature and compression displacement, whereas Fig. 8(f)
shows the relationship between the time of maximum tempera-
ture and compression displacement. The maximum temperature
increases with the compression displacement, whereas the time
decreases with it. The large compression displacement causes
minimal resistance, which causes rapid heat release. Thus, LIBs
with high compression displacements reach maximum tempera-
ture early and show low heat dissipation, which leads to a high
maximum temperature
5. Concluding remarks
LIB behavior under compressive mechanical abuse loading
involves multiphysics mechanisms. Given the complicated nature
of LIB behavior, computation modeling has become one of the most
promising methods to address the problem effectively. This study
considers major events, including deformation, short circuit, and
thermal runaway during compressive loading. The following five
sub-models are proposed to describe the entire process through
coupling: 1D battery, mechanical, thermal, thermal runaway, and
short-circuit models. The integrated computational model is
validated at SOC = 0.1 via experiment measurements to ensure
accuracy. Finally, governing factors for mechanical and electro-
chemical behaviors, namely, SOC value, compression speed, and
displacement, under mechanical abuse loading are discussed
based on the model to reveal the underlying mechanisms. The
major conclusions are summarized as follows.
 Multiphysics computational modeling can accurately describe
the mechanical deformation of cells, as well as the short-circuit
and thermal runaway behavior in an affordable and cost-
effective way.
 The temperature rising behavior has two different modes due to
the SOC difference, namely, LSM and HSM.
 A fast compression speed causes rapid deformation, which
triggers an early short-circuit time. The maximum temperature
and the time of maximum temperature increase with the
compression speed from 0.1 mm/s to 10 mm/s but remain almost
constant if the compression speed exceeds 10 mm/s.
 LIBs with large compression displacements reach the maximum
temperature early with minimal heat dissipation, leading to a
high maximum temperature.
 A simplified 2D model is used to predict the typical mechanical
behavior of a cell under compressive loading conditions instead
of a more complicated 3D model. However, a 3D mechanical
model should be applied in complicated loading conditions, in
which the stress status cannot be described in 2D form.
 The 1D battery model cannot compute the heat from the current
collectors accurately, and it is ignored for its relatively low value
in this paper. The computation method of the heat from the
current collectors would be further studied to gain a more
accurate prediction capability.
183
 The integrated model proposes an innovative framework model
of compression abuse conditions and sheds light on LIB safety
design in terms of mechanical integrity.
Acknowledgement
This work is financially supported by start-up funds of “The
Recruitment Program of Global Experts” awardee from Beihang
University (YWF-17-BJ-Y-28), Research Project of the State Key
Laboratory of Vehicle NVH and Safety Technology under Grant
NVHSKL-201610, Opening project of State Key Laboratory of
Explosion Science and Technology (Beijing Institute of Technology)
with project number of KFJJ17–13 M and Excellence Foundation of
BUAA for PhD Students.
Appendix A. Supplementary data
Supplementary data associated with this article can be found,
in the online version, at https://doi.org/10.1016/j.electacta.2017.
10.045.
References
[1] C. Wang, G. Zhang, S. Ge, T. Xu, Y. Ji, X. Yang, Y. Leng, Lithium-ion battery
structure that self-heats at low temperatures, Nature 529 (2016) 515.
[2] P. Ping, Q. Wang, P. Huang, J. Sun, C. Chen, Thermal behaviour analysis of
lithium-ion battery at elevated temperature using deconvolution method,
Applied Energy 129 (2014) 261–273.
[3] W. Waag, S. Kabitz, D.U. Sauer, Experimental investigation of the lithium-ion
battery impedance characteristic at various conditions and aging states and its
influence on the application, Applied Energy 102 (2013) 885–897.
[4] B. Liu, S. Yin, J. Xu, Integrated computation model of lithium-ion battery
subject to nail penetration, Applied Energy 183 (2016) 278–289.
[5] J. Chen, T. Sun, Y. Qi, X. Li, A Coupled Penetration-Tension Method for
Evaluating the Reliability of Battery Separators, Ecs Electrochemistry Letters 3
(2014) A41–A44.
[6] T.D. Hatchard, D.D. Macneil, A. Basu, J.R. Dahn, Thermal Model of Cylindrical
and Prismatic Lithium-Ion Cells, Journal of the Electrochemical Society 148
(2001) A755–A761.
[7] B. Coleman, J. Ostanek, J. Heinzel, Reducing cell-to-cell spacing for large-
format lithium ion battery modules with aluminum or PCM heat sinks under
failure conditions, Applied Energy 180 (2016) 14–26.
[8] J. Ye, H. Chen, Q. Wang, P. Huang, J. Sun, S. Lo, Thermal behavior and failure
mechanism of lithium ion cells during overcharge under adiabatic conditions,
Applied Energy 182 (2016) 464–474.
[9] A.W. Golubkov, S. Scheikl, R. Planteu, G. Voitic, H. Wiltsche, C. Stangl, G. Fauler,
A. Thaler, V. Hacker, Thermal runaway of commercial 18650 Li-ion batteries
with LFP and NCA cathodes—impact of state of charge and overcharge, Rsc Adv
5 (2015) 57171–57186.
[10] X. Feng, X. He, M. Ouyang, L. Lu, P. Wu, C. Kulp, S. Prasser, Thermal runaway
propagation model for designing a safer battery pack with 25 Ah
LiNixCoyMnzO2 large format lithium ion battery, Applied Energy 154 (2015)
74–91.
[11] C.Y. Jhu, Y.W. Wang, C.M. Shu, J.C. Chang, H.C. Wu, Thermal explosion hazards
on 18650 lithium ion batteries with a VSP2 adiabatic calorimeter, J. Hazard.
Mater. 192 (2011) 99–107.
[12] Y. Xia, T. Wierzbicki, E. Sahraei, X.W. Zhang, Damage of cells and battery packs
due to ground impact, J. Power Sources 267 (2014) 78–97.
[13] X. Feng, C. Weng, M. Ouyang, J. Sun, Online internal short circuit detection for a
large format lithium ion battery, Applied Energy 161 (2016) 168–180.
[14] L. Greve, C. Fehrenbach, Mechanical testing and macro-mechanical finite
element simulation of the deformation, fracture, and short circuit initiation of
cylindrical Lithium ion battery cells, J. Power Sources 214 (2012) 377–385.
[15] J. Xu, B. Liu, X. Wang, D. Hu, Computational model of 18650 lithium-ion battery
with coupled strain rate and SOC dependencies, Applied Energy 172 (2016)
180–189.
[16] W.W. Wang, S. Yang, C. Lin, Clay-like Mechanical Properties of Components for
the Jellyroll of Cylindrical Lithium-ion Cells, Energy Procedia 104 (2016) 56–61.
[17] C. Zhang, S. Santhanagopalan, M.A. Sprague, A.A. Pesaran, Coupled
mechanical-electrical-thermal modeling for short-circuit prediction in a
lithium-ion cell under mechanical abuse, J. Power Sources 290 (2015) 102–113.
[18] M. Gilaki, I. Avdeev, Impact modeling of cylindrical lithium-ion battery cells: a
heterogeneous approach, J. Power Sources 328 (2016) 443–451.
[19] C.M. Doyle, Design and simulation of lithium rechargeable batteries, Lawrence
Berkeley National Laboratory, University of California, 1995, pp. 370.
184 B. Liu et al. / Electrochimica Acta 256 (2017) 172–184
[20] W. Lai, F. Ciucci, Mathematical modeling of porous battery electrodes—Revisit
of Newman's model, Electrochimica Acta 56 (2011) 4369–4377.
[21] M. Xu, Z. Zhang, X. Wang, L. Jia, L. Yang, Two-dimensional electrochemical-
thermal coupled modeling of cylindrical LiFePO4 batteries, J. Power Sources
256 (2014) 233–243.
[22] Y. Ye, Y. Shi, N. Cai, J. Lee, X. He, Electro-thermal modeling and experimental
validation for lithium ion battery, J. Power Sources 199 (2012) 227–238.
[23] V. Srinivasan, J. Newman, Design and optimization of a natural graphite/iron
phosphate lithium-ion cell, Journal of the Electrochemical Society 151 (2004)
A1530–A1538.
[24] K. Kumaresan, G. Sikha, R.E. White, Thermal Model for a Li-Ion Cell, Journal of
the Electrochemical Society 155 (2008) A164–A171.
[25] R.E. Gerver, J.P. Meyers, Three-Dimensional Modeling of Electrochemical
Performance and Heat Generation of Lithium-Ion Batteries in Tabbed Planar
Configurations, Journal of the Electrochemical Society 158 (2011) A835–A843.
[26] W. Fang, P. Ramadass, Z. Zhang, Study of internal short in a Li-ion cell-II.
Numerical investigation using a 3D electrochemical-thermal model, J. Power
Sources 248 (2014) 1090–1098.
[27] D.E. Stephenson, E.M. Hartman, J.N. Harb, D.R. Wheeler, Modeling of particle-
particle interactions in porous cathodes for lithium-ion batteries, Journal of
the Electrochemical Society 154 (2007) A1146–A1155.
[28] A. Nyman, M. Behm, G. Lindbergh, Electrochemical characterisation and
modelling of the mass transport phenomena in LiPF6-EC-EMC electrolyte,
Electrochimica Acta 53 (2008) 6356–6365.
[29] C.-Y. Jhu, Y.-W. Wang, C.-Y. Wen, C.-M. Shu, Thermal runaway potential of
LiCoO2 and Li(Ni1/3Co1/3Mn1/3)O-2 batteries determined with adiabatic
calorimetry methodology, Applied Energy 100 (2012) 127–131.
[30] P.T. Coman, S. Rayman, R.E. White, A lumped model of venting during thermal
runaway in a cylindrical Lithium Cobalt Oxide lithium-ion cell, J. Power
Sources 307 (2016) 56–62.
[31] K.-C. Chiu, C.-H. Lin, S.-F. Yeh, Y.-H. Lin, K.-C. Chen, An electrochemical
modeling of lithium-ion battery nail penetration, J. Power Sources 251 (2014)
254–263.
[32] R. Zhao, J. Liu, J. Gu, Simulation and experimental study on lithium ion battery
short circuit, Applied Energy 173 (2016) 29–39.
[33] W. Zhao, G. Luo, C.Y. Wang, Modeling Internal Shorting Process in Large-
Format Li-Ion Cells, Journal of the Electrochemical Society 162 (2015) A1352–
A1364.
[34] T.G. Zavalis, M. Behm, G. Lindbergh, Investigation of Short-Circuit Scenarios in
a Lithium-Ion Battery Cell, Journal of the Electrochemical Society 159 (2012)
A848–A859.
[35] X. Zhang, Thermal analysis of a cylindrical lithium-ion battery, Electrochimica
Acta 56 (2011) 1246–1255.
[36] G.H. Kim, A. Pesaran, R. Spotnitz, A three-dimensional thermal abuse model for
lithium-ion cells, J. Power Sources 170 (2007) 476–489.
[37] J. Xu, B. Liu, L. Wang, S. Shang, Dynamic mechanical integrity of cylindrical
lithium-ion battery cell upon crushing, Eng. Fail. Anal. 53 (2015) 97–110.
[38] R. Spotnitz, J. Franklin, Abuse behavior of high-power, lithium-ion cells, J.
Power Sources 113 (2003) 81–100.
[39] A. Pesaran, G. Kim, S. Santhanagopalan, C. Yang, Multi-physics Modeling for
Improving Li-Ion Battery Safety, NREL (National Renewable Energy
Laboratory), 2015.
[40] J. Xu, B. Liu, D. Hu, State of Charge Dependent Mechanical Integrity Behavior of
18650 Lithium-ion Batteries, Scientific Reports 6 (2016) 21829.
[41] S. Santhanagopalan, P. Ramadass, J. Zhang, Analysis of internal short-circuit in
a lithium ion cell, J. Power Sources 194 (2009) 550–557.
[42] Q.S. Wang, P. Ping, X.J. Zhao, G.Q. Chu, J.H. Sun, C.H. Chen, Thermal runaway
caused fire and explosion of lithium ion battery, J. Power Sources 208 (2012)
210–224.
[43] H. Yang, X.D. Shen, Dynamic TGA-FTIR studies on the thermal stability of
lithium/graphite with electrolyte in lithium-ion cell, J. Power Sources 167
(2007) 515–519.
[44] Q.S. Wang, J.H. Sun, X.L. Yao, C.H. Chen, Thermal stability of LiPF6/EC + DEC
electrolyte with charged electrodes for lithium ion batteries, Thermochimica
Acta 437 (2005) 12–16.
[45] Y. Shigematsu, M. Ue, J. Yamaki, Thermal Behavior of Charged Graphite and
LixCoO2 in Electrolytes Containing Alkyl Phosphate for Lithium-Ion Cells,
Journal of the Electrochemical Society 156 (2009) A176–A180.
[46] J.W. Jiang, J.R. Dahn, Effects of solvents and salts on the thermal stability of
LiC6, Electrochimica Acta 49 (2004) 4599–4604.
[47] A.D. Pasquier, F. Disma, T. Bowmer, A.S. Gozdz, G. Amatucci, J.M. Tarascon,
Differential scanning calorimetry study of the reactivity of carbon anodes in
plastic Li-ion batteries, Journal of the Electrochemical Society 145 (1998) 472–
477.
[48] T. Wierzbicki, E. Sahraei, Homogenized mechanical properties for the jellyroll
of cylindrical Lithium-ion cells, J. Power Sources 241 (2013) 467–476.
[49] E. Sahraei, J. Campbell, T. Wierzbicki, Modeling and short circuit detection of
18650 Li-ion cells under mechanical abuse conditions, J. Power Sources 220
(2012) 360–372.
[50] J. Newman, K.E. Thomas-Alyea, Electrochemical systems, John Wiley & Sons,
2012.
[51] D. Miranda, C.M. Costa, A.M. Almeida, S. Lanceros-Mendez, Computer
simulations of the influence of geometry in the performance of
conventional and unconventional lithium-ion batteries, Applied Energy 165
(2016) 318–328.
[52] A. Nyman, T.G. Zavalis, R. Elger, M. Behm, G. Lindbergh, Analysis of the
Polarization in a Li-Ion Battery Cell by Numerical Simulations, Journal of the
Electrochemical Society 157 (2010) A1236–A1246.
[53] M. Tang, P. Albertus, J. Newman, Two-Dimensional Modeling of Lithium
Deposition during Cell Charging, Journal of the Electrochemical Society 156
(2009) A390–A399.