A low false alarm rates oriented design scheme of multivariate control chart

Multivariate statistical control charts (MSCC) are used broadly in manufacturing process. It is used to deal with multi-sensor measurements to detect the product variation which will induce degradation of product quality such as reliability, durability etc. High false alarm rate is one of disadvantages to the MSCC engineering applications. In this paper, a low false alarm rates oriented design scheme of multivariate control charts is proposed in order to make MSCC more practicable. Firstly, the paper analyzes causes of false alarms in multivariate statistical process monitoring, and factors leading to high false alarms rate. Then, the paper proposes a new design scheme of multivariate statistical process charts based on multiple dots alarm rules. It shows that the scheme could deliver effectively charts with reduced false alarm rate by evaluating the average run length. Finally, a comparison study of Hotelling χ² charts and multivariate cumulative sum (MCUSUM) control charts is conducted. The comparison result validates the effectivity and efficiency of the presented design scheme of multivariate control charts.