Abstract
This paper discusses a new approach for constructing an approximated surface of an unknown response surface. The proposed hybrid algorithm consists of two processes: (1) global approximation with Kriging method, (2) local approximation with polynomial approximation. One of the key ideas of the proposed method is Hessian evaluation of an estimated surface, which can be estimated using the Kriging-based formulation, for accurate approximation based on a quadratic polynomial approximation. In order to investigate validity and effectiveness of the proposed method, it is applied to a layout optimization problem of a beam structure. Compared to the result of an optimality-based approach, effectiveness of the proposed method is illustrated.
