Modeling and Analysis of the Battery Packs and Modules in A123 SystemsBinshan Ye & Shawn Zhang

A123 Systems, Inc.

Outline

• Overview of CAE capacity in A123

• CAE Modeling and Analysis Examples

• Random vibration fatigue analysis with HWPA program (DesignLife)

• Cell material properties characterizing with HyperStudy

• Concluding Remarks

A123 Engineering Simulation Capability

A123 Systems

Engineering Simulation

CFD and Thermal Management

Cooling Concept Development and

Validation

Battery Life Analysis

Battery Life Estimation Software Development

Finite Element Analysis

Linear Statics and Modal Frequency

Validation

Module and Pack Level Thermal and Flow Analysis

Thermal/Electrical Coupling (Joule Heating)

Thermal Analysis for Electronics

Battery Life Analysis

Battery Electrical Performance Simulation

Random Vibration and Fatigue

Mechanical Shock and Drop Analysis

Nonlinear Statics

Cell R&D and External Supplier

Pack and Module Level FEA Analysis

• Linear Statics and modal frequency analysis• Modal frequency analysis

• Foot/knee load, handle load, and lifting assistance analysis

• Topology/topography/shape/gauge optimization

• Random vibration stress and fatigue analysis• RMS stress calculation

• Fatigue life calculation for metal parts

• Mechanical shock, pothole, and drop analysis

• Nonlinear and contact analysis • Snap-in/pull-out force estimation

• Jack loading analysis

• Bolt assembly, module pressure plate, etc.

FEA Tools Used in A123 Systems

• Altair HyperWorks Suite• Radioss/Bulk

• Radioss/Block

• OptiStruct

• HyperStudy

• LS-DYNA3D

• ABAQUS (Implicit/Explicit)

• Access to other software through HyperWorks Partner Alliance License• nCode DesignLife

• Key to Metals

• Others

• Altair PBS Pro

Random Vibration Analysis Random Vibration Analysis

on Battery Pack

Battery Pack Vibration Analysis

• A123 conducts random vibration stress and fatigue a nalysis according

to customer specifications or industrial standards

• Approach• Use Radioss/Bulk to calculate RMS stresses from the PSD profiles

• Estimate fatigue life using nCode DesignLife if necessary

SAE J2380 PSD Profiles

Example – Random Vibration Analysis

• A prototype battery pack had a test failure on moun ting brackets during

random vibration test

• The analysis team was involved to identify the root causes of the

failures and find the solution in a limited time fr ame

Challenges

• A few locations on mounting brackets showed fatigue cracks

• Initial random vibration stress analysis showed the failure locations

have high RMS stress during vibration events, but i t cannot

accurately quantify the fatigue life

• The fatigue properties for the metal components wer e unknown• The fatigue properties for the metal components wer e unknown

• Project timing and budget won’t allow performing ma terial test to

obtain the fatigue properties

Correlating the Fatigue Properties

• nCode DesignLife was used to evaluate the fatigue l ife of metal

components:• The stress-life properties were estimated in DesignLife based on material specs

• Random vibration fatigue engine was used to estimate the fatigue life of the metal

components

• The fatigue properties and analysis parameters were then adjusted to correlate

the analysis results with test resultsthe analysis results with test results

Vibration Fatigue Analysis EngineMaterial Stress Life Curve

Result Comparisons

• With correlated fatigue properties, the analysis id entified all test failure

locations:• The failure locations have relatively high RMS stresses comparing to material specs

• The fatigue lives in these locations are lower than the requirement

� 3 RMS stress: 72% of material σuts

� Fatigue life: 10% of required life� 3 RMS stress : 55% of material σuts

� Fatigue life: 90% of the required life

10% of required life90% of required life

Improve the Design Through Analysis

• Based on the analysis results, new design concepts were proposed:• Change the shape of the components

• Add reinforcement brackets

• Change welding patterns

• New pack design passed the random vibration fatigue analysis

• These design changes were implemented and the new p ack went

through random vibration test without fatigue issue

Infinite 65 lives

Prismatic Cell material Prismatic Cell material

Property Characterization

Challenges for Battery Module Modeling

• Modal frequency is critical for battery

pack design, and battery modules

play a significant role

• Cell property largely unknown

• Ideally, we would like to use a simple

homogenized model to represent the homogenized model to represent the

complex structure of the module

(cells, heat sinks, and bands)

• The first few modal frequencies of the

module model should meet the test

results

Two Module Modeling Approaches

• Homogenized model

• Cell, heat sink, compliance pad are

homogenized into blocks

• End plate is modeled with shell

elements as one plane sheet

• Module bolt is modeled with beam

elements

• Detailed model

• Each component is modeled in detail with

corresponding material properties

• End plate is modeled in detail with shell

elements

• Module bolt is modeled with beam

elements

• All materials are isotropic

• Pros and Cons:

• Can be quick modeled and use very

little CPU time

• Accuracy is compromised due to

simplification

• Pro and Cons:

• Can better predict module dynamic

behavior

• Long modeling time due to complexity of

the module

• High CPU and memory costs

Hybrid Module Modeling Approach

Y

Z

X

Z

• Endplate modeled in detail by shell element

• Bolt was modeled by beam element with rod section

• Cell, heat sink, cell compliance pad, band were hom ogenized into a 3-d

orthotropic material

• Local coordinate system was used for the orthotropi c material modeling

Characterizing the Material

• Three modules were tested with free-free and fixed BC• Large size module, medium size module, and small size module

• For free-free boundary condition, the first 3 modes from test were used

for FEA model correlation

• For fixed boundary condition, the first 5 modes fro m test were used in

FEA model correlationFEA model correlation

• Homogenized orthotropic material was formulated usin g the following

engineering constants

Characterizing the Material

• Goal was to adjust E1, E2, E3, G12, G13, G23 to cor relate both the

mode shapes and frequencies with test results.

• Observations during initial evaluation: • Some Eii, Gij, and vij have strong influence to long and medium size modules’

modal frequencies;

• Other Eii, Gij, and vij have significant effect to small module modal frequencies• Other Eii, Gij, and vij have significant effect to small module modal frequencies

• The remaining Eii, Gij, and vij have little effect to the first 3 modal frequencies at

all. In that case, they are assigned to zero, leading to a simple material matrix

• Material parameters were first manually adjusted to match modal

shapes in order.

• Then HyperStudy was used to match first 3 frequenci es more closely

Modal Correlation

• HyperStudy

Modal Correlation

• HyperStudy

Modal Correlation

• HyperStudy

Results Correlations

Free-free BC 1st Mode 2 nd Mode 3 rd Mode

Large module 0.6% .48% 0.%

Medium module 3.3% 2.7% 1.5%

Small module .27% 5.9% 2.8%

Table-1: Relative Deviations of Estimated Modal Frequencies from Test Results under free-free condition

Fixed BC 1 st Mode 2nd Mode 3 rd Mode 4th Mode 5th Mode

Large module 4.5% 10.7% 1.6% 2.1% 0.3 %

Medium module 1.5% 9.5% 8.1% 17.5% 24.2%

Small module 0.1% 7% 8.1% -- --

Table-2: Relative Deviations of Estimated Modal Frequencies from Test Results under fixed condition

Illustration of Typical Module Mode

Summary of Hybrid Module Model

• This hybrid module model was a compromise among all 3 size modules,

with deviation within 5% in free-free boundary cond ition

• The hybrid module model was more skewed to large si ze modules

because for small size modules, the first frequency is very high already,

making them less sensitive to external vibration.

• By using such approach, a battery module for pack a nalysis can be • By using such approach, a battery module for pack a nalysis can be

quickly modeled and still achieve good analytical r esults

Concluding Remarks

• A123 has a broad range of engineering simulation ca pabilities to

support battery pack/module development activities

• Altair’s HyperWorks Suite and HWPA are the best cos t-effective tools to

match A123’s FEA simulation requirements