Abstract
This paper proposes a cumulative function approximation method based on Voronoi diagram for engineering optimization. It consists of a function approximation by blending local information at sample points based on the spatial structure of Voronoi diagram, an estimation method of function gradients and an integrated optimization scheme over approximated functions. The algorithm can gradually improve the fidelity of optimal solution by adaptively arranging new samples and cumulatively updating approximations based on tentative optimization results. After describing the approximation method and optimization scheme, this paper demonstrates optimization computation of a two-dimensional optimization problem and discusses its promise.
