Approach for machine learning based design of experiments for occupant simulation

The complexity of crash scenarios in the context vehicle safety is steadily increasing. This especially case on way to mixed traffic challenges with non-automated and automated driving vehicles. number simulations required design a robust restraint system thus also vast range possible here causing huge parameter space. Simultaneously biofidelic simulation models are resulting very high computational costs therefore should be limited feasible operational range. In this study, machine-learning based experiments algorithm developed, which specifically addresses issues when designing samples taking diversity occupant accident scenario into account. contrast an optimization task, where aim meet target function, our job has been find critical load combinations make sure that these addressed not missed. A combination space-filling approach metamodel established order improve for those cases. It focuses areas difficult predict by metamodel. developed method was applied iteratively generate matrix total 208 generic interior model detailed FE human body model. Kinematic strain-based injury metrics were used as output. These train metamodels after each iteration derive next iteration. paper we present allows training prediction criteria, considering both varying cases parameters individual anthropometries seating postures. Based that, systems or can optimized achieve best overall performance variety specific focus scenarios.