16th International LS-DYNA® Users Conference Automotive

A Path Towards Including Batteries in Electric

or Hybrid Car Crash Simulations with LS-DYNA®
Pierre L’Eplattenier, Iñaki Çaldichoury
Livermore Software Technology, LLC

e
or
Abstract
Safety is an important functional requirement in the development of large-format, energy-dense, lithium-ion (Li-ion) batteries used in

Am
electrified vehicles. Many automakers have dealt with this issue by enclosing the batteries into protective cases to prevent any
penetration and deformation during the car crash. But with the range of electric vehicle increasing and thus the size of the batteries, a
more detailed understanding of a battery behavior under abuse becomes necessary.

N
Computer aided engineering (CAE) tools that predict the response of a Li-ion battery pack to various abusive conditions can support
analysis during the design phase and reduce the need for physical testing. In particular, simulations of the multi-physics response of
external or internal short circuits can lead to optimized system designs for automotive crash scenarios.
DY
The physics under such simulations is quite complex, though, coupling structural, thermal, electrical and electrochemical. Moreover, it
spans length scales with orders of magnitude differences between critical events such as internal shorts happening at the millimeter
level, triggering catastrophic events like the thermal runaway of the full battery. The time scales also are quite different between the car
crash happening in milliseconds and the discharge of the battery and temperature surge taking minutes to hours.
by

A so called “distributed Randles circuit” model was introduced in LS-DYNA in order to mimic the complex electrochemistry happening
in the electrodes and separator of lithium ion batteries [1][2][3]. This model is based on electrical circuits linking the positive and
negative current collectors reproducing the voltage jump, internal resistance and dumping effects occurring in the active materials.
ht

These circuits are coupled with the Electromagnetics (EM) resistive solver to solve for the potentials and current flow in the current
collectors and the rest of the conductors (connectors, busses, and so forth). The EM is coupled with the thermal solver to which the joule
rig

heating is sent as an extra heat source, and from which the EM gets back the temperature to adapt the electrical conductivity of the
conductors as well as the parameters of the Randles circuits [1]. One of the advantages of the Randles circuit model is the relative
easiness to introduce internal short circuits by just replacing the Randles circuits in the affected area by a short resistance [1][3]. The
Randles circuit model also is coupled with the mechanical solver of LS-DYNA where the deformations due to a battery crush allow the
py

definition of criteria to initiate internal shorts [1].

The Randles circuit model can be used either on a solid element mesh that include all the layers of a cell [1][2][3], or using composite
Co

Tshells [4][5]. In the second case, the mechanics is solved on the composite Tshell, but an underlying solid mesh with all the layers still
has to be built to solve the EM and the thermal. This implies very large meshes and hence simulation times when dealing with many
cells, let alone modules, packs or a full battery. This new Battery Macro (BatMac) model allows simulating a cell with very few layers
of elements (down to one). Two fields exist at each node of the mesh, representing the potential at the positive and negative current
collectors. These two fields are connected by a Randles circuit at each node. It still is possible to include external and internal shorts.
The internal shorts can be locally created depending on local values of different mechanical, thermal or EM parameters. The Joule
Heating generated by the current leaking through the short resistance is sent to the thermal solver.

In this paper, the BatMac model will be presented along with some examples.

June 10-11, 2020 1

16th International LS-DYNA® Users Conference Automotive

1-Introduction
Safety is an important functional requirement in the development of large-format, energy-dense, lithium-ion (Li-
ion) batteries used in electrified vehicles. Computer aided engineering (CAE) tools that predict the response of a
Li-ion battery pack to various abusive conditions can support analysis during the design phase and reduce the
need for physical testing. In particular, simulations of the multiphysics response of external or internal short
circuits can lead to optimized system designs for automotive crash scenarios.
Recently, a so called “distributed Randles circuit” model was introduced in LS-DYNA in order to mimic the
complex electrochemistry happening in the electrodes and separator of lithium ion batteries [1][2][3]. This model
is based on electrical circuits linking the positive and negative current collectors reproducing the voltage jump,

e
internal resistance and dumping effects occurring in the active materials. These circuits are coupled with the
Electromagnetics (EM) resistive solver to solve for the potentials and current flow in the current collectors and

or
the rest of the conductors (connectors, busses, and so forth). The EM is coupled with the thermal solver to which
the joule heating is sent as an extra heat source, and from which the EM gets back the temperature to adapt the

Am
electrical conductivity of the conductors as well as the parameters of the Randles circuits [1]. One of the
advantages of the Randles circuit model is the relative easiness to introduce internal short circuits by just replacing
the Randles circuits in the affected area by a short resistance [1][3]. The Randles circuit model also is coupled
with the mechanical solver of LS-DYNA where the deformations due to a battery crush allow the definition of

N
criteria to initiate internal shorts [1].
The Randles circuit model can be used either on a solid element mesh that include all the layers of a cell [1][2][3],
DY
or using composite Tshells [4][5]. In the second case, the mechanics is solved on the composite Tshell, but an
underlying solid mesh with all the layers still has to be built to solve the EM and the thermal. This implies very
large meshes and hence simulation times when dealing with many cells, let alone modules, packs or a full battery.
This new Battery Macro (BatMac) model allows simulating a cell with very few layers of elements (down to one).
by

Two fields exist at each node of the mesh, representing the potential at the positive and negative current collectors.
These two fields are connected by a Randles circuit at each node. It still is possible to include external and internal
shorts. The internal shorts can be locally created depending on local values of different mechanical, thermal or
ht

EM parameters. The Joule Heating generated by the current leaking through the short resistance is sent to the
thermal solver.
In this paper, this BatMac model will be presented along with some examples.
rig
py

2-Presentation of the model

2.1 General idea
Co

A battery cell is composed of many layers, which can be decomposed as a succession of a few tens of “unit” cell,
each composed of a positive current collector, a positive electrode, a separator, a negative electrode and a negative
current collector. Figure 1 shows the current flow from one tab to the other through the two current collectors,
with the current flowing in one direction in the positive current collector, and in the other one in the negative one.

June 10-11, 2020 2

16th International LS-DYNA® Users Conference Automotive

e
or
N Am
DY
Fig.1: Current flow from the positive tab to the negative tab from the current collectors and the Randles circuits R
(only 1 represented here).
by

In the previous “Randles circuit” models [1-5], the electrochemistry happening in the electrodes sandwiched
between the two current collectors was replaced by distributed Randles circuits and the current flow in the current
collectors was solved by a Finite Element Model (FEM). The succession of layers was still necessary to compute
ht

the correct current flow.

In this new BatMac model, instead of having two different parts, one for the positive current collector where the
rig

downstream current flows, and one for the negative current collector, where the upstream current flows; the flow
of the current is represented on one part only, where two fields co-exist. One field represents the potential on the
positive current collector (the gradient of which is the current on the positive current collector), and the other one
py

the potential on the negative one (its gradient being the current on the negative current collector). At each node,
these two fields are connected by a local Randles circuit.
In the BatMac model, we even replace all the unit cells of a cell by only one or a few solid layers with the above-
Co

mentioned 2 fields at each node. These fields represent the averages of the current flowing respectively on all the
positive and negative current collectors of the cell. This is shown on Figure 2.

June 10-11, 2020 3

16th International LS-DYNA® Users Conference Automotive

e
or
N Am
Fig.2: Going from the layered model (left) where all the positive and negative current collectors are meshed to the
BatMac model (right) where 2 fields coexist at each node of the mesh, one representing the potential (or current) in
DY
the positive current collectors of the cell, and the other one representing the potential (current) on the negative
current collectors

2.2 Model equations

by

A schematic of the EM in the BatMac model is shown on Figure 3.

ht
rig
py
Co

Fig.3: Schematics of the BatMac model.

June 10-11, 2020 4

16th International LS-DYNA® Users Conference Automotive

The equations solved in the BatMac cells are the following:

1 1
∇ ∙ �𝜎𝜎𝑝𝑝 ∇𝜑𝜑𝑝𝑝 � + (𝜑𝜑𝑝𝑝 − 𝜑𝜑𝑛𝑛 ) = (𝑢𝑢 − 𝑣𝑣𝑐𝑐 )
𝑅𝑅0 𝑅𝑅0
1 1
∇ ∙ (𝜎𝜎𝑛𝑛 ∇𝜑𝜑𝑛𝑛 ) − �𝜑𝜑𝑝𝑝 − 𝜑𝜑𝑛𝑛 � = − (𝑢𝑢 − 𝑣𝑣𝑐𝑐 )
𝑅𝑅0 𝑅𝑅0
𝑑𝑑𝑣𝑣𝑐𝑐 𝑖𝑖 𝑣𝑣𝑐𝑐
= −
𝑑𝑑𝑑𝑑 𝑐𝑐10 𝑟𝑟10 𝑐𝑐10

With the notations of a Randles circuit reminded in Figure 4, and where 𝑖𝑖 represents the transverse current.

e
or
N Am
DY
Fig.4: Notations for the components of a Randles circuit
by

From the potentials 𝜑𝜑𝑝𝑝 and 𝜑𝜑𝑛𝑛 , one gets the currents in the positive and negative current collectors respectively
by:
𝑖𝑖𝑝𝑝 = ∇𝜑𝜑𝑝𝑝
ht

𝑖𝑖𝑛𝑛 = ∇𝜑𝜑𝑛𝑛
rig

2.3 Thermal solver

The connection of BatMac with the thermal solver works the same way as when using solid elements or Tshells
py

1
[1], i.e. the Joule heating 2 𝑅𝑅0 𝑖𝑖 2 generated in the internal resistance 𝑅𝑅0 is added to the thermal solver at the node
connected to the Randles circuit. Conversely, the different parameters of each Randles circuit 𝑅𝑅0 , 𝑅𝑅10 , 𝐶𝐶10 can
Co

depend on the local temperature at the node.

2.4 External and internal shorts in the BatMac model

External shorts can be introduced in the BatMac model exactly the same way as with solid and Tshell battery
models [1][4], i.e. it is modeled by replacing some Randles circuits by short resistances. This switch is controlled
locally by the values of different mechanical, thermal and EM parameters at the node corresponding to the randles
circuit. This is done using a *DEFINE_FUNCTION which returns the value of the local short resistance, if any.
Several parameters are available in the define function such as node location, pressure, density, von mises stress,
electrical conductivity, temperature and effective strain. Several were added following users’ feedback and more
can be added if needed.

June 10-11, 2020 5

16th International LS-DYNA® Users Conference Automotive

3-Setting up a BatMac model in LS-DYNA

First of all, a BatMac run requires to turn on the Resistive heating solver in *EM_CONTROL

*EM_CONTROL
emsol
3

A BatMac cell or a set of BatMac cells is defined by the keyword *EM_RANDLES_BATMAC, which is based
on a part set Id. The rest of the card is similar to the already existing *EM_RANDLES_SOLID for cells defined

e
with solid elements [1] and *EM_RANDLES_TSHELL for cells based on composite Tshells [4].

or
*EM_RANDLES_BATMAC
cellId rdlOrder AreaType Psid

Am
1 1 1 1

As explained previously, the positive and negative potentials coexist at each node of a BatMac cell. In order to
define both the electrical conductivities (and possibly EOS’s) of the positive (𝜎𝜎𝑝𝑝 ) and negative (𝜎𝜎𝑛𝑛 ) current

N
collectors, the keyword *EM_MAT_006 has been added. DY
*EM_MAT_006
mid mtype sigmaP EosP sigmaN eosN
1 5 1.e6 3.e6
by

The connection between the BatMac cells and the tabs are done using *EM_ISOPOTENTIAL and
*EM_ISOPOTENTIAL_CONNECT as shown in Figure 5.
ht
rig
py
Co

Fig.5: Summary of a BatMac model setup showing in particular the electrical connections between the BatMac
parts and the tabs

June 10-11, 2020 6

16th International LS-DYNA® Users Conference Automotive

4-Examples
4.1 Sphere crash on a 10 cells module
We consider a module composed of 10 adjacent cells mounted in parallel. Each cell is about 20 cm long, 12 cm
wide and 3.5 mm thick. The mesh of each cell has 1 element through thickness (to represent the actual 89 layers)
and 12x19 in the other directions, thus making a mesh of only 2280 elements, and 5200 nodes, hence 5200 Randles
circuits. The case setup is presented on Figure 6.

e
or
N Am
DY
by

Fig.6: Mesh for the sphere impacting a 10 cells module. The mesh has 3275 solid elements, with 2280 for the
cells themselves
ht

This can be compared to the same case using composite Tshells [4] where the underlying mesh used for the EM
rig

and the thermal was made of 202,920 elements, and which had 55,440 Randles circuits. This makes the present
run about 20 times faster than the same one using composite Tshells.
py

In this case, the onset of internal short was triggered in the *EM_RANDLES_SHORT define function by a
criteria based on the local pressure. Figure 7, shows the evolution of the internal short.
Co

June 10-11, 2020 7

16th International LS-DYNA® Users Conference Automotive

e
or
N Am
DY
Fig.7: Impact of a sphere on a 10 cells module: current density (top) and temperature (bottom) at different times.
by

4.2 Crush of a 50 cells pack

Since the BatMac model only requires a small number of elements per cells, the next example represents a small
ht

pack with 50 cells, crushed by a plane on one of its corner, as shown in Figure 8.
rig
py
Co

Fig.8: 50 cells pack impacted by a moving plane

June 10-11, 2020 8

16th International LS-DYNA® Users Conference Automotive

The mesh contains about 12,000 elements and ran in about 30 mn on 4 CPU’s. The short was triggered by a
criteria on the strain at each node. Figure 9 shows the potential, current density and temperature at different time
of the impact.

e
or
N Am
DY
by
ht

Fig.9: 50 cells pack impacted by a moving plane: potential (top), current density (middle) and temperature (bottom)
at different times
rig

4.3 Nail Penetration

py

The erosion capability was added to the BatMac model, allowing the modeling of nail penetration in a cell or a
bunch of cells. During the penetration, the Randles circuits corresponding to the nodes that are eroded are removed
Co

from the model, and the one around the hole created by the nail are switched to internal short resistance. These
internal shorts create a high current concentration and thus Joule heating with a large temperature increase around
the holes. Since the nail penetration is very fast corresponding to the discharge of the cells through the internal
shorts and the temperature elevation, the simulation was done with mechanic explicit during the nail penetration;
and switched to mechanics implicit for the longer discharge and temperature elevation.

Figure 10 shows a nail penetrating into a module of 30 cells. The nail is sent with a given velocity and it
decelerates at each cell penetration in the erosion process. Its initial kinetic energy allowed it to penetrate around
10 cells, so we used the BatMac model for the first 10 cells and a meshless model for the other 20, in order to
save computation time while keeping the correct discharge time and energy.

June 10-11, 2020 9

16th International LS-DYNA® Users Conference Automotive

e
or
N Am
DY
Fig.10: 30 cells pack impacted by a nail

As said before, the nail penetration itself is very fast, taking about 10ms, but the discharge of the cells through
the internal shorts created by the nail takes much longer, about 100s in this case. However both processes were
by

simulated in the same simulation where we increased dramatically the mechanical, EM and thermal time steps
after the penetration.
ht

Figure 11 shows the penetration itself and the internal shorts that are created all around the eroded areas triggering
a high current density in these areas. Figure 12 shows the much longer cell discharge through the internal shorts
rig

generating some joule effect that increase the temperature of the cell.
py
Co

June 10-11, 2020 10

16th International LS-DYNA® Users Conference Automotive

e
or
N Am
DY
Fig.11: Current density evolution during the nail penetration at times (a) t=0.002s; (b) t=0.004s; (c) t=0.006s and
(d) t=0.008s
by
ht
rig
py
Co

Fig.12: Temperature rise after the nail has penetrated (a) t=1s; (b) t=20s; (c) t=40s and (d) t=80s

June 10-11, 2020 11

16th International LS-DYNA® Users Conference Automotive

4.4 Exothermal reaction and thermal runaway

During a short, localized temperature rise may trigger exothermic decomposition reactions leading to further
temperature rise and eventually, thermal runaway and fire. This translates into adding extra power source term in
the heat equation. In LS-DYNA, there are several options and models available. The first one is general, and gives
the user the choice of defining his own model using the keyword
*EM_RANDLES_EXOTHERMIC_REACTION which makes use of a *DEFINE_FUNCTION to specify a
heat source term function of Randles circuit parameters.

In parallel to this approach, thermal reaction models were added based on [9],[10] and [11]. Reaction kinetics are

e
used to model those decomposition reactions. Two models are currently implemented, the 1-equation model and
the 4-equation model (See *LOAD_HEAT_EXOTHERMIC_REACTION). In the 4-equation model for

or
example the total heat generation due to thermal abuse can be modeled as four exothermic reactions: SEI
decomposition reaction, Negative-Solvent reaction, Positive-Solvent reaction and Electrolyte decomposition

Am
reaction, where the R’s are reaction parameters and the c’s dimensionless concentrations, and the S’s are the
corresponding heat sources:
𝐸𝐸
SEI decomposition −� 𝑎𝑎,𝑠𝑠𝑠𝑠𝑠𝑠 �

N
𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚
reaction 𝑅𝑅𝑠𝑠𝑠𝑠𝑠𝑠 (𝑇𝑇, 𝑐𝑐𝑠𝑠𝑠𝑠𝑠𝑠 ) = 𝐴𝐴𝑠𝑠𝑠𝑠𝑠𝑠 𝑒𝑒 𝑅𝑅𝑢𝑢𝑇𝑇 𝑐𝑐𝑠𝑠𝑠𝑠𝑠𝑠
𝑆𝑆𝑠𝑠𝑠𝑠𝑠𝑠 = 𝐻𝐻𝑠𝑠𝑠𝑠𝑠𝑠 𝑊𝑊𝑐𝑐 𝑅𝑅𝑠𝑠𝑠𝑠𝑠𝑠
DY
𝑑𝑑𝑐𝑐𝑠𝑠𝑠𝑠𝑠𝑠
= −𝑅𝑅𝑠𝑠𝑠𝑠𝑠𝑠
𝑑𝑑𝑑𝑑
𝐴𝐴𝑠𝑠𝑠𝑠𝑠𝑠 , 𝐸𝐸𝑎𝑎,𝑠𝑠𝑠𝑠𝑠𝑠 , 𝐻𝐻𝑠𝑠𝑠𝑠𝑠𝑠 , 𝑊𝑊𝑐𝑐 , 𝑚𝑚𝑠𝑠𝑠𝑠𝑠𝑠 : model constants
by

𝐸𝐸𝑎𝑎,𝑛𝑛𝑛𝑛𝑛𝑛 𝑡𝑡𝑠𝑠𝑠𝑠𝑠𝑠
Negative-solvent −� � 𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 (−𝑡𝑡 )
reaction 𝑅𝑅𝑛𝑛𝑛𝑛𝑛𝑛 �𝑇𝑇, 𝑐𝑐𝑛𝑛𝑛𝑛𝑛𝑛 , 𝑡𝑡𝑠𝑠𝑠𝑠𝑠𝑠 � =
𝐴𝐴𝑛𝑛𝑛𝑛𝑛𝑛 𝑒𝑒 𝑅𝑅𝑢𝑢 𝑇𝑇 𝑐𝑐𝑛𝑛𝑛𝑛𝑛𝑛 𝑒𝑒 𝑟𝑟𝑟𝑟𝑟𝑟
𝑆𝑆𝑠𝑠𝑠𝑠𝑠𝑠 = 𝐻𝐻𝑛𝑛𝑛𝑛𝑛𝑛 𝑊𝑊𝑛𝑛 𝑅𝑅𝑛𝑛𝑛𝑛𝑛𝑛
ht

𝑑𝑑𝑐𝑐𝑛𝑛𝑛𝑛𝑛𝑛
= −𝑅𝑅𝑛𝑛𝑛𝑛𝑛𝑛
𝑑𝑑𝑑𝑑
rig

𝑑𝑑𝑡𝑡𝑠𝑠𝑠𝑠𝑠𝑠
= 𝑅𝑅𝑛𝑛𝑛𝑛𝑛𝑛
𝑑𝑑𝑑𝑑
𝐴𝐴𝑛𝑛𝑛𝑛𝑛𝑛 , 𝐸𝐸𝑎𝑎,𝑛𝑛𝑛𝑛𝑛𝑛 , 𝐻𝐻𝑛𝑛𝑛𝑛𝑛𝑛 , 𝑊𝑊𝑛𝑛 , 𝑚𝑚𝑛𝑛𝑛𝑛𝑛𝑛 , 𝑡𝑡𝑟𝑟𝑟𝑟𝑟𝑟 : model constants
py

Positive-solvent 𝐸𝐸𝑎𝑎,𝑝𝑝𝑝𝑝𝑝𝑝
−� �
reaction 𝑅𝑅𝑝𝑝𝑝𝑝𝑝𝑝 (𝑇𝑇, 𝛼𝛼) = 𝐴𝐴𝑝𝑝𝑝𝑝𝑝𝑝 𝑒𝑒 𝑅𝑅𝑢𝑢𝑇𝑇 𝛼𝛼 𝑚𝑚𝑝𝑝1 (1 − 𝛼𝛼)𝑚𝑚𝑝𝑝2
Co

𝑆𝑆𝑝𝑝𝑝𝑝𝑝𝑝 = 𝐻𝐻𝑝𝑝𝑝𝑝𝑝𝑝 𝑊𝑊𝑝𝑝 𝑅𝑅𝑝𝑝𝑝𝑝𝑝𝑝

𝑑𝑑𝑑𝑑
= 𝑅𝑅𝑝𝑝𝑝𝑝𝑝𝑝
𝑑𝑑𝑑𝑑
𝐴𝐴𝑝𝑝𝑝𝑝𝑝𝑝 , 𝐸𝐸𝑎𝑎,𝑝𝑝𝑝𝑝𝑝𝑝 , 𝐻𝐻𝑝𝑝𝑝𝑝𝑝𝑝 , 𝑊𝑊𝑝𝑝 , 𝑚𝑚𝑝𝑝1 , 𝑚𝑚𝑝𝑝2 : model constants

Electrolyte 𝐸𝐸
−� 𝑎𝑎,𝑒𝑒𝑒𝑒𝑒𝑒 � 𝑚𝑚𝑒𝑒
decomposition reaction 𝑅𝑅𝑒𝑒𝑒𝑒𝑒𝑒 (𝑇𝑇, 𝑐𝑐𝑒𝑒 ) =
𝐴𝐴𝑒𝑒𝑒𝑒𝑒𝑒 𝑒𝑒 𝑅𝑅𝑢𝑢𝑇𝑇 𝑐𝑐𝑒𝑒
𝑆𝑆𝑒𝑒𝑒𝑒𝑒𝑒 = 𝐻𝐻𝑒𝑒𝑒𝑒𝑒𝑒 𝑊𝑊𝑒𝑒 𝑅𝑅𝑒𝑒𝑒𝑒𝑒𝑒
𝑑𝑑𝑐𝑐𝑒𝑒
= 𝑅𝑅𝑒𝑒𝑒𝑒𝑒𝑒
𝑑𝑑𝑑𝑑
𝐴𝐴𝑒𝑒𝑒𝑒𝑒𝑒 , 𝐸𝐸𝑎𝑎,𝑒𝑒𝑒𝑒𝑒𝑒 , 𝐻𝐻𝑒𝑒𝑒𝑒𝑒𝑒 , 𝑊𝑊𝑒𝑒 , 𝑚𝑚𝑒𝑒 : model constants

Total heat source term 𝑆𝑆𝑡𝑡𝑡𝑡𝑡𝑡 = 𝑆𝑆𝑒𝑒𝑒𝑒𝑒𝑒 + 𝑆𝑆𝑝𝑝𝑝𝑝𝑝𝑝 + 𝑆𝑆𝑛𝑛𝑛𝑛𝑛𝑛 + 𝑆𝑆𝑠𝑠𝑠𝑠𝑠𝑠

June 10-11, 2020 12
16th International LS-DYNA® Users Conference Automotive

Not uncommonly, the main challenge for such models is how to define those constants which must be sought in
the literature and which are not available for each type of battery. Numerically, the kinetic reactions advancing in
time lead to a potential dependency of the results on the timestep, especially concerning the sharp slope those
reactions can have, so the user must ensure to use a ‘low enough’ value to reduce that dependency.

4.5 Thermal management using ICFD

The lithium ion batteries are widely used for electric vehicles due to high energy density and long cycle life. Since
the performance and life of lithium-ion batteries are very sensitive to temperature, it is important to maintain the
proper temperature range. Also, when the cell temperature goes above a certain limit, it will allow a series of

e
undesirable exothermic reactions to occur which will further increase the temperature. This chain type reaction will
continue and lead to an incident called thermal runaway.

or
Air is the most conventional way for cooling and has been used widely in various industries. Due to low heat capacity
and low thermal conductivity, air might not seem to be a good cooling medium. However, it is still an attractive cooling

Am
solution due to its simplicity and low cost. Toyota Prius and Nissan Leaf are two of the most famous examples.
Water is used in several industrial applications as one of the most efficient coolants. However, the main challenge with
directly cooling batteries with water is the short-circuit potential. Therefore, indirect methods are used to prevent
electrical conduction with the cells while maintaining high thermal conductivities.

N
We studied air and liquid cooling on a module of 96 cylindrical cells positioned in 6 rows of 16 cells, similar to a small
DY
Tesla model S module, as presented in [8].
by
ht
rig
py
Co

Fig.13: Presentation of the 2 cooling systems of the 96 cylindrical cell module: direct air cooling (bottom right), and
indirect water cooling through a water+glycol channel in the bottom plate (bottom right)

The simulations were done using the Batmac model for the 96 cylindrical cells, coupled with the thermal module and
the ICFD module for either the indirect liquid cooling or the direct air coiling. Figure 14 shows the difference between
the different cooling strategies and shows that both cooling strategies allow a very significant temperature reduction.

June 10-11, 2020 13

16th International LS-DYNA® Users Conference Automotive

e
or
N Am
DY
Fig.14: Comparison between the different cooling systems: (a): no cooling, final max temperature = 43 C.; (b):
indirect water glycol cooling, final max temperature = 24.4 C; (c): direct air cooling, final temperature = 25.8 C
by

Direct cooling, also known as immersion cooling, covers the entire surface of the cell and cools it uniformly. This
ht

mitigates hot/cold spots in the cell and improves the performance of the cell. The coolant for direct cooling should
be dielectric with low viscosity and high thermal conductivity and thermal capacity. One of the issues of this
rig

cooling system is the risk of external short circuits.

The new coupling between the EM and ICFD solvers, presented in another paper of this conference allows to
model such situations.
py

5-Conclusion
Co

The BatMac model was developed to handle battery modules, packs or even a full battery crush. The goal is to
be able to include a full battery in an electric or hybrid vehicle crash. Along with the already existing solid [1-3]
Tshell [4-5] and meshless battery models of LS-DYNA, it creates a suite of models for battery crash study [6][7],
where the user can use the more microscopic models (solid or Tshells) to study the behavior of one or a few cells
under abuse conditions, and then use the results of this study on the more macroscopic BatMac model. In
particular, the microscopic model should help setup the dependence of the internal short resistance on the different
local mechanical, thermal and EM parameters, dependence which then can be used in the
*DEFINE_FUNCTION of the *EM_RANDLES_SHORT card of the BatMac model. Table 1 shows a
summary of the different battery models in LS-DYNA along with their respective uses.
New coupling with the ICFD module of LS-DYNA allow the simulation of thermal management as well as
external shorts due to conducting liquid immersion.

June 10-11, 2020 14

16th International LS-DYNA® Users Conference Automotive

Solid Composite Macro model Meshless

Tshells
Cell geometry √/√/√ √/√/√ √/√/√ X
(pouch/cyl/prism)
Cell √ √ X X
(detailed study)
Module If small √ √ √
Pack – battery X X √ √
Connection to √ √ √ √
external circuit

e
Internal short √ √ √ X
External short √ √ √ √

or
Thermal √ √ √ (average) X
(Joule heating)

Am
Mechanics √ √ √ (average) X
(Deformations)
LS-PrePost® √ (pouch) Soon √ (pouch) X
Battery packaging

N
Table 1: Battery models in LS-DYNA with their different capabilities and usage
DY
References
by

[1] P. L’Eplattenier, I. Caldichoury, J. Marcicki, A. Bartlett, X. Yang, V. Mejia, M. Zhu, Y. Chen, “A distributed Randles circuit
model for battery abuse simulations using LS-DYNA”, Proceeding of the 14th International LS-DYNA User Conference, 2016.
[2] J. Marcicki, A. Bartlett, X. Yang, V. Mejia, M. Zhu, Y. Chen, P. L’Eplattenier, I. Caldichoury, “Battery abuse case study analysis
using LS-DYNA”, Proceeding of the 14th International LS-DYNA User Conference, 2016.
ht

[3] J. Marcicki, M. Zhu, A. Bartlett, X. Yang, Y. Chen, T. Miller, P. L’Eplattenier, I. Caldichoury, “A simulation framework for
battery cell impact safety modeling using LS-DYNA”, Journal of The Electrochemical Society, 164 (2017) A6440-A6448.
rig

[4] P. L’Eplattenier, S. Bateau-Meyer, I. Caldichoury, “Battery abuse simulations using LS-DYNA”, Proceeding of the 11th European
LS-DYNA User Conference, 2017.
[5] J. Deng, C. Bae, T. Miller, P. L'Eplattenier, S. Bateau-Meyer „Accelerate Battery Safety Simulations Using Composite Tshell
Elements” Journal of the Electrochemical Society 2018 165(13): A3067-A3076.
py

[6] J. Deng, C. Bae, T. Miller, P. L’eplattenier and I. Caldichoury, “Multiphysics Battery Simulations Across Length Scales”,
Journal of the Electrochemical Society, 2019 166(14): A3119-A3121.
[7] J. Deng, I. Smith, C. Bae, P. Rairigh, T. Miller, B. Surampudi, P. L’Eplattenier, and I. Caldichoury, “Impact Modeling And
Co

Testing on Pouch and Prismatic Cells”, to be published.

[8] SMH Moghaddam, „Designing battery thermal management systems (BTMS) for cylindrical Lithium-ion battery modules using
CFD“, MS Thesis, KTH Industrical Engineering and Management, 2019.
[9] D. D. MacNeil, J. R. Dahn, “Test of Reaction Kinetics Using Both Differential Scanning and Accelerating Rate Calorimetries as
Applied to the Reaction of LixCoO2 in Non-aqueous Electrolyte”, J. Phys Chem., 2001.
[10] T. D. Hatchard, D. D. MacNeil, A. Basu, and J. R. Dahn, “Thermal Model of Cylindrical and Prismatic Lithium-Ion Cells”, J. of
the Electrochemical Society, 2001.
[11] Gi-Heon Kim, Ahmad Pesaran, Robert Spotnitz, “A Three-dimensional thermal abuse model for lithium-ion cells”, J. of Power
Resources, 2007.

June 10-11, 2020 15