주성분-분산분석의 주축 수정을 통한 다중 성능 시스템의 입력변수 스크리닝 기법

Abstract

For efficient design, it is necessary to screen design variables that significantly affect performance, and for this purpose, analysis of variance (ANOVA) method that tests the difference among group means on individual response axis is commonly used. Multivariate ANOVA (MANOVA) method which is extension of ANOVA is developed to enhance test power especially in system with correlated multiple responses. However, there are limitations such as the fact that MANOVA contains several tests, and that each test’s test power changes depending on status of data, so additional analysis is required. 4.Thus, in this research, PC-ANOVA combining principal component analysis and ANOVA is introduced into screening method, and modified principal axis for performing ANOVA is proposed to improve test power of existing. Also, characteristic and usefulness of proposed method will be verified by applying it to simple example and engineering problem.