Battery PHM: Life & Stress Prediction

Optimizing Battery Pack Designs Meant Building and Testing Prototypes for Months

Determining the size and configuration of battery modules is directly tied to production costs in electric vehicle battery pack design. However, because battery stress and State of Health (SOH) are difficult to predict in advance, engineers had to repeatedly build and test physical battery cells until failure to identify the optimal design. A full lifecycle test could take up to 180 days, significantly increasing both development time and validation costs before mass production.

Predicting Battery Health and Optimizing Designs Without Physical Testing

An equivalent mechanical model was developed to quantify changes in battery module stress, along with a Kalman filter-based SOH prediction model. The model represents battery health as a probability distribution by accounting for uncertainty, and its prediction accuracy improves as additional experimental data becomes available.

Based on predicted stress, expansion, and SOH values, a reliability-based design optimization approach was applied, enabling the optimal battery module configuration to be identified without extensive physical testing.

Reducing Production Costs by $59.8M and Cutting Development Time by 90 Days

Reliability-based design optimization reduced battery module volume by 8.3%. When applied to electric vehicle mass production, this translates into an estimated $59.8 million reduction in production costs.

Development time was reduced by 90 days. While conventional lifecycle testing required up to 180 days, the SOH prediction model enables the same design conclusions to be reached with testing of only about 500 charge-discharge cycles. This significantly shortens development time and supports faster design validation aligned with vehicle launch schedules.

The project also established a data-driven design framework, replacing the conventional approach of relying on repeated physical testing. By leveraging data and AI-based prediction models, battery pack designs can be validated more efficiently and rapidly adapted to new battery cells or module configurations.