반도체 공정 내 비정규 데이터의 이상 감지를 위한 비모수 다변량 관리도 적용에 관한 연구

Abstract

In the univariate control charts, Two problem is commonly existed: Increasing the number of variables leads to increasing of the number of control charts and incapability to detect abnormality due to correlation among the variables. In order to improve it, an application of multivariate control chart is essential. However, Hotelling`s chart which is the most popularly used multivariate control chart is applicable only if data have the multivariate normality. Due to condition that semiconductor process data doesn't follow normal distribution, it can`t be applied Hotelling's chart Therefore, in this paper, We suggest the method of applying nonparametric multivariate control chart based on the data depth. We select the adjust control chart by analyzing simulation results for comparing several data depth based nonparametric multivariate control charts and We apply this control chart to real process data.