Research On Robust Design Optimization Methods Considering Parametric And Metamodeling Uncertainty

Robust Design Optimization(RDO) has been widely used in engineering design optimization. The core of the robust design is to optimize the mean performance and minimize the performance variance simultaneously. Computer simulation has become a powerful tool for replacing physical tests in robust design applications. In practical engineering design optimization, simulation analysis of RDO is usually complex and time-consuming. The application of metamodel techniques greatly reduces the computational cost in RDO. However, metamodel is only an approximation of the original model resulting in metamodel uncertainty. In traditional robust design, only parametric uncertainty is considered rather than metamodel uncertainty, which may induce design error. In this paper, we mainly focuses on the compound effect of both parametric uncertainty and metamodeling uncertainty in RDO.First, analytic formula and numerical integration are used to quantify the effects of parametric uncertainty and metamodeling uncertainty on the robust responses based on Kriging model which has high prediction accuracy. This section compare the two methods with a mathematical example, and the result provides theoretical basis for RDO based on sequential sampling.Second, study the unconstrained RDO problem with considering parametric uncertainty and metamodeling uncertainty. In this paper, an improved objective-oriented sequential sampling method is proposed to improve the sampling efficiency of robust design. The sequential method relies on quantifying the effect of parametric uncertainty and metamodeling uncertainty on a robust response. The proposed method can select the updated sampling points effectively and convergent to the robust optimal quickly. Experimental result illustrates the effectiveness of the proposed method.Finally, an improved multi-point objective-oriented sequential sampling strategy is proposed for constrained robust design problems. The sequential method relies on quantifying the effect of parametric uncertainty and metamodeling uncertainty on a robust objective and constrain responses. The proposed method is validated using one mathematical example followed by structural design of a welded joint. The results show that the proposed method can obtain the robust design solution more efficiently.