Abstract

Engineering design optimization often gives rise to problems in which expensive objective func-tions are minimized by derivative-free methods. We propose a method for solving such problems that synthesizes ideas from the numerical optimization and computer experiment literatures. Our approach relies on kriging known function values to construct a sequence of surrogate mod-els of the objective function that are used to guide a grid search for a minimizer. Results from numerical experiments on a standard test problem are presented.

Citations