Crashworthiness-based lightweight design problem via new robust design method considering two sources of uncertainties

Metamodel-based robust design methods are commonly used to mitigate the influence of parametric uncertainty associated in sheet gauges and material properties in crashworthiness-based vehicle lightweight design. Since the crash performances are highly nonlinear and high-dimensional responses, the prediction error of metamodels inevitably introduces the so-called metamodeling uncertainty in robust design that may mislead to a wrong solution. In this article, a new robust design method considering both parametric uncertainty and metamodeling uncertainty is proposed in the autobody lightweight design problem. Six crash responses in side impact and roof crush are defined as the constraint responses. The results demonstrate that the proposed robust design method is superior to the conservative-surrogatebased robust design method. The final confirmed robust solution achieves 14.39% weight reduction. The method provides an efficient way to reduce the risk of constraint violation and avoids an over-conservative design due to metamodel uncertainty in crashworthiness-based lightweight design problems.