Finite Element Analysis and Machine Learning Guided Design of Carbon Fiber Organosheet-based Battery Enclosures for Crashworthiness

TL;DR

This paper combines finite element analysis and machine learning to optimize the design of carbon fiber battery enclosures for electric vehicles, enhancing crashworthiness prediction accuracy and reducing development time.

Contribution

It introduces a high throughput FEA-based simulation framework combined with ML models to predict crashworthiness metrics of carbon fiber enclosures, a novel approach in this application.

Findings

ML models achieved R2 > 0.97 in crashworthiness prediction

High throughput FEA simulations enabled virtual testing of multiple designs

Framework aids in parameter selection for carbon fiber component design

Abstract

Carbon fiber composite can be a potential candidate for replacing metal-based battery enclosures of current electric vehicles (E.V.s) owing to its better strength-to-weight ratio and corrosion resistance. However, the strength of carbon fiber-based structures depends on several parameters that should be carefully chosen. In this work, we implemented high throughput finite element analysis (FEA) based thermoforming simulation to virtually manufacture the battery enclosure using different design and processing parameters. Subsequently, we performed virtual crash simulations to mimic a side pole crash to evaluate the crashworthiness of the battery enclosures. This high throughput crash simulation dataset was utilized to build predictive models to understand the crashworthiness of an unknown set. Our machine learning (ML) models showed excellent performance (R2 > 0.97) in predicting the crashworthiness metrics, i.e., crush load efficiency, absorbed energy, intrusion, and maximum deceleration during a crash. We believe that this FEA-ML work framework will be helpful in down select process parameters for carbon fiber-based component design and can be transferrable to other manufacturing technologies.