A Nonparametric Multivariate Statistical Process Control Chart Based on Change Point Model
Ya Fei Xu
Beijing AI Lab, VIVO Communication Technology Co. Ltd., China
Chair of Econometrics and Statistics, Technische Universitat Dresden, Germany
Abstract. This article presents a nonparametric control chart based on the change point model, for multivariate statistical process control (MSPC). The main constituent of the chart is the energy test that focuses on the discrepancy between empirical characteristic functions of two random vectors. This new multivariate control chart highlights in three aspects. Firstly, it is nonparametric, requiring no pre-knowledge of the data generating processes. Secondly, this control chart monitors the whole distribution, not just specific characteristics like mean or covariance. Thirdly, it is designed for online detection (Phase II), which is central for real time surveillance of stream data. Simulation study discusses in-control and out-of-control measures in the context of mean shift and covariance shift. The obtained results are compared with benchmarks and strongly advocate the proposed control chart.
Keywords: Change point model • Energy test • Multivariate statistical process monitoring • Phase II statistical process control.
To buy this chapter or the book, use the form below