A data-driven methodology for the automated analysis and explanation of system behavior in crash simulations

Attributable to model size and complexity of numerical crash simulations, it is not feasible for the engineers to analyze each area or component in detail, especially when these are not the core subject of investigation. In the field of occupant safety, the main explanatory objective is given by the signals of anthropometric test devices (ATD), as they are relevant for the fulfillment of legal regulations and consumer protection guidelines. Hence, this study proposes a data-driven methodology to automatically determine deviations in ATD behavior in a set of simulations and provide possible causes for the prevalence helping the engineer to understand simulations faster and to ensure quality.