A multi-objective robust optimization approach based on Gaussian process model

The design and optimization of engineering products is usually multi-objective, constrained and has uncertainties in the inputs. It is of great importance for taking these uncertainties into consideration during the design process because these uncertainties can significantly degrade the performance of optimal solutions and even change the feasibility of obtained solutions. Most existing Multi-objective robust optimization (MORO) approaches rely on outer-inner nested optimization structures, where a large number of function evaluations is required. In this work, a MORO approach based on Gaussian process (GP) model is proposed to ease the computational burden of MORO under interval uncertainty. To consider the interpolation uncertainty introduced by GP model, an objective switching criterion is developed, which is according to whether the robust status of the individual can be changed because of the interpolation uncertainties from GP model or not. Six numerical and engineering cases with different degrees of difficulty are used to demonstrate the applicability and efficiency of the proposed approach. The objective and feasibility robustness of the obtained optimal solutions are verified via the design of experiment.