Efficient Methods for Robust Shape Optimisation for Crashworthiness.

Abstract

Recently complex geometry and detailed Finite Element (FE) models have been used to capture the true behaviour of the structures for crashworthiness. Such model complexity, detailed FE model, high non-linearity of crash cases and high number of design variables for crashworthiness optimisation add to the required computational effort. Hence, engineering optimisation problems are currently highly restricted in exploring the entire design space and including the desired number of design parameters. Hence it is advantageous to reduce the computational effort to fully explore the design alternatives and also to study even more complex and computationally expensive problems. This thesis presents an efficient robust shape optimisation approach via the use of physical surrogate models, i.e. sub-models and models derived for the Equivalent Static Loads Method (ESLM). The classical simultaneous robust design optimisation (RDO) approach (where robustness analysis of each design is assessed) is modified to make use of the physical surrogate models. In the proposed RDO approach, design optimisations are made using sub-models and robustness analyses are made using either non-linear dynamic analysis or ESLM. The general idea is to approximate the robustness of designs at the start of the optimisation (using ESLM) and use accurate robustness evaluations (via non-linear dynamic analysis) towards the end of the optimisation where the optimisation has already found interesting regions of the design space. The approach is validated on crashworthiness design cases.