Probabilistic analytical target cascading combined with Kriging metamodel for multidisciplinary robust design optimization

Abstract

Robust design under a multidisciplinary environment is a challenging problem in modern quality engineering because of the organizational and computational burden. In this paper, a probabilistic analytical target cascading (PATC) methodology combined with Kriging metamodel (PATC-KRG) is proposed for robust design of engineering systems considering both objective robustness and feasibility robustness. PATC is an effective methodology for hierarchical multilevel optimization under uncertainty. However, when mathematical models are complex and highly nonlinear, especially when simulations are used, multidisciplinary robust design optimization (MRDO) will be highly computationally expensive. To obtain better computational efficiency, in the PATC-KRG method, Kriging metamodels are used to replace the complicated models, and an exponential penalty function (EPF) formulation is employed for target-response coordination in PATC. A geometric programming problem is used to demonstrate the feasibility and computational efficiency of the PATC-KRG method. Results indicate that the PATC-KRG method can achieve a higher efficiency in support of MRDO.