Worst-case robust optimization based on an adaptive incremental Kriging metamodel

Worst-case robust optimization problem is concerned with finding a candidate solution that is insensitive to uncertainty. This problem involves nested-loop structure based on the worst-case analysis. It is an expensive optimization problem which has large computational complexity. This paper presents a robust optimization method of reduced computational complexity. Firstly, the state transition algorithm (STA) is utilized to explore and exploit the candidate solutions of the search space. Then, an adaptive incremental Kriging (IKriging) metamodel is proposed to replace the evaluation functions for evaluating the robustness of candidate solutions. Finally, a preferential selection strategy is presented to select the optimal solution in terms of objective function value, constraint and robustness violation. Four engineering examples are studied to analyze the performance of the proposed robust optimization method. Experimental results illustrate that the proposed method can find a better robust solution with a high computational efficiency. © 2024 Elsevier Ltd