이중순환 다중응답접근법 기반 다변수 크리깅 메타모델

Abstract

Metamodel have been developed with a variety of design optimization techniques to achieve better efficiency and easy interconnections. For the last decade, the kriging metamodel is widely used in the computational experiments because the correlation parameter is logically and automatically estimated by maximum likelihood estimation (MLE). However, if the number of design variables for construction of the kriging metamodel is increased, the number of presampled points and correlation parameters are also increased. Due to this increase, the optimization for MLE becomes time-consuming or construction of the kriging metamodel is impossible depending upon the computing power. In this research, we propose the multi-variable kriging metamodel based on double-loop multiple responses approach (MRA) that can split design variables and achieve better efficiency and accuracy compared to the existing technique considering multi-variable examples.