Optimisation and robustness for crashworthiness of side impact

TLDR

The study proposes a reliability‑based design optimisation framework that uses a nonlinear response surface and Monte Carlo simulation to robustly improve vehicle side‑impact crash safety. The authors build an efficient surrogate model via stepwise regression and Latin hypercube sampling, then apply sequential quadratic programming with mixed variables for optimisation, and validate the results with CAE simulation. Applying the method to side‑impact crash safety design demonstrates that vehicle weight can be substantially reduced while achieving higher safety performance and confidence.

Abstract

This paper presents a nonlinear response surface-based safety optimisation and robustness process. The stepwise regression and optimal Latin hyper cube sampling methods are employed to construct the "efficient-to-compute" surrogate model. A sequential quadratic programming method with mixed type of variables is employed for the design optimisation. A reliability based design optimisation model for robust system parameter design of vehicle safety is proposed and a Monte Carlo based stochastic simulation is used to perform the robustness assessment and the reliability-driven robust design. The methodology has been applied to the vehicle crash safety design of side impact. It shows that the vehicle weight can be significantly reduced with an improved safety performance and with a higher level of confidence. CAE simulation is used to validate the optimal results.