크리깅 메타모델의 전역 모델에 따른 예측 성능

Many meta-models have been developed to approximate relationship between response and design variables and replace expensive computer simulations. Kriging meta-model especially approximates complex nonlinear systems with many peaks and valley. Furthermore, it offers the statistical estimations and optimizes parameters by using Maximum Likelyhood Estimation. so it has no user-defined parameters based on design's know-how. Although Kriging is an efficient meta-model. we should select appropriate forms of the Kriging models. Furthermore, it is lack of guidance in selecting the appropriate form of the Kriging models. The goal of this research is to study an influence of global model of estimation capability. In this work, a valiance of estimation capability of Kriging global models is investigated for several lest problems. Then suing error measurements we evaluated an accuracy of estimation capability and proposed the guidance in selecting the appropriate form.