基于多点报警的多元统计过程控制

Multivariate Statistical Process Control Based on Multiple Dots Alarm Rules

本文主要研究了一种将控制图的报警准则由一点改为多点时多元控制图的构造,对于不同的多点报警准则的控制图可以通过计算平均运行长度的方法得出了不同的控制限参数。然后本文对各种报警准则下的多元控制图与Hotelling T2控制图,MCUSUM(多元累计和控制图)和MEWMA(多元指数加权移动平均控制图)进行了比较,说明这种改变报警规则的多元控制图在监测过程中出现的微小波动时具有一定的优势。

This paper mainly studied to building multivarate control charts under different muhiple dots alarm rules. For different multiple dots alarm rules, control limit parameters can be given by a kind of method of calculating average run length. Then the performances of those kinds of multivariate control schemes under different alarm rules were compared with Hntelling T^2 chart, MC USU M and MEWMA. We can find from this compare that those charts under different alarm rules have advantage in detecting small changes in the mean vector of a multivariate process. At last a example is used to illustrated how this method can used in practice.