Enhancing numerical simulation analysis with the use of explainable artificial intelligence and large language models: a case study on vehicle side crash optimization

Substantial engineering efforts are dedicated to reducing injury risks in crash scenarios during the development of new vehicles. This is achieved by performing crash simulations to optimize the nonlinear behavior of systems. However, the complexity makes their behavior difficult and time-consuming for engineers to understand. To reduce the analysis time, this study introduces a modular framework combining Explainable Artificial Intelligence and Large Language Models (LLM). Shapley Additive Explanation values allow for simulation-wise feature importance attribution and generate a data-driven understanding. An LLM assists by making result data interactively accessible and supports technical report generation. Validated through a real-world vehicle side crash optimization use case, the framework demonstrates enhanced and accessible insights into system behavior within virtual engineering.