Data-driven machine learning techniques for crashworthiness analysis of thin-walled structures: A review

This paper reviews recent advances in the application of machine learning techniques to the crashworthiness analysis of thin-walled structures. Thin-walled structures are widely used in aerospace, automotive, and defense industries due to their lightweight characteristics, weight ratio, and superior energy absorption capacity, thereby establishing crashworthiness evaluation as a fundamental requirement in the design process. Conventional numerical simulations and experimental approaches are constrained by high computational cost, model complexity, and limited scalability. To address these issues, machine learning methods have emerged as powerful tools for predicting crash responses, energy absorption, deformation modes, and failure mechanisms. This review discusses the capabilities of machine learning in reducing computational effort, improving predictive accuracy, and enabling rapid design optimization of crash box structures. Furthermore, the paper highlights current limitations, including data dependency, generalization issues, and the need for robust validation under diverse loading scenarios. Future research opportunities are outlined to integrate machine learning models with physics-based simulations and digital twin concepts, ultimately advancing the reliability and applicability of crashworthiness analysis for thin-walled composite structures.