Efficient Kriging-based robust optimization of unconstrained problems

Abstract

A novel methodology, based on Kriging and expected improvement, is proposed for applying robust optimization on unconstrained problems affected by implementation error. A modified expected improvement measure which reflects the need for robust instead of nominal optimization is used to provide new sampling point locations. A new sample is added at each iteration by finding the location at which the modified expected improvement measure is maximum. By means of this process, the algorithm iteratively progresses towards the robust optimum. It is demonstrated that the algorithm performs significantly better than current techniques for robust optimization using response surface modeling.