统计过程控制的三个基本问题

Three Basic Problems in Statistic Process Control

本文围绕统计过程的监测、诊断和变点时间估计三个基本问题,首先概括地介绍了过程监测的最优控制图—累积和控制图;然后,详细介绍了如何利用多重累积和控制图对过程的异变进行统计诊断;最后简述了三种常用的变点估计方法:最大似然估计,Schwarz信息判别估计和Bayesian估计。此外,文中还列出6个亟待解决的问题。

Based on the three basic concepts： detection, diagnosis and change-point estimation in the statistic process control, we first introduce the optimal control chart, the CUSUM control chart, in detecting the distribution change of a process, then describe how to make a statistic diagnosis by using the CUSUM multi-chart scheme, and finally investigate three ways of estimating the change-point： maximum likelihood, Schwarz information criterion and Bayesian approach. Moreover, six unsolved problems of detection, diagnosis and change-point estimation are proposed in the paper.