“…Because a number of solutions will be sampled in the neighbor of an individual to evaluate its robustness, the computational burden for a robust optimization is extremely higher than that for a global optimization. To address this issue, a surrogate model is generally used in lieu of the heavy simulation model (Yao et al , 2006; An et al , 2018; Ho and Yang, 2012; Ong et al , 2006; Silva et al , 2018; Tenne and Armfield, 2008). The polynomial chaos expansion is used as a stochastic response surface model of the objective function for efficient robustness performance computations, and the numerical results on two case studies confirm the feasibility (Ho and Yang, 2012).…”