An Efficient Robust Optimal Design Method for Engineering Systems with Numerical Noise

Abstract

This study presents an efficient robust optimal design method, whose name is SARO (Sequential Approximate Robust Optimization), for engineering systems with numerical noise. Conventional optimization methods, based on the design sensitivity information, cannot accomplish optimization when the system responses have numerical noise because the design sensitivity information is not available or not reliable in this case. An RSM (Response Surface Modeling) has been applied to cope with this weakness of conventional optimization methods. It still however has two drawbacks. First, it requires a large number of design points to construct a response surface model. Secondly, it cannot capture the variations of the real objective function exactly during the robust optimization process because it uses a quadratic approximate function. In order to overcome above two faults, this study adopts PQRSM (Progressive Quadratic Response Surface Modeling) as the approximate method and the trust region model management strategy is employed. Finally, in order to show the numerical performance of the proposed SARO, a mathematical problem and two engineering system design problems with numerical noise, which cannot be solved by the conventional robust optimal design methods, are solved and their results are verified by the Worst Case analysis.