INNOVATIVE DESIGN OPTIMIZATION STRATEGY FOR THE AUTOMOTIVE INDUSTRY

摘要

In order to effectively solve modern automotive design problems including the results of nonlinear FEA and multi-body dynamics, a progressive meta-model based design optimization is presented. To reduce the number of initial sample points, two sampling methods are introduced. Then, for efficient and stable construction of meta-models, three meta-model methods are newly introduced which are numerically based on the singular value decomposition technique. To design a practical system considering manufacturing tolerances and optimizing multiple performances, a robust design optimization, 6-sigma constraints and multi-objective strategies are implemented when solving the approximate optimization problem constructed from the meta-models. Until the convergence criteria are satisfied, the initially developed meta-models are progressively improved by adding only one point that minimizes the approximate Lagrangian in the consecutive optimization iterations. Finally, one validation sample and four automotive applications are solved to show the effectiveness of the proposed approach.