摘要
多元自相关过程不满足现行多元质量控制理论的前提假设。该文探讨了两个随机变量相互独立,其中一个随机变量的观测值相互独立、另一随机变量服从一阶自回归模型的二元自相关过程。在参数已知的条件下,提出了二元自相关过程的残差T2控制图。通过M on te C arlo模拟,得到了一族该二元自相关过程在不同偏移量下的平均链长。分析结果表明残差T2控制图的适用范围由自相关的强弱和偏移量的大小决定,可以有效监控大部分该类二元自相关过程。
Multivariate autocorrelated processes violate the fundamental assumption of independence for traditional statistical process control. This paper investigates bivariate autocorrelated processes in which the observations of one characteristic are autocorrelated following a first-order autoregressive model while the observations of the other characteristic are independent. The system parameters were monitored using an analysis for bivariate autocorrelated processes, termed the residual-based T^2 control chart. The average run lengths predicted by Monte Carlo simulations were given for a cluster of such processes with various mean shifts. Analysis of the result shows that the autocorrelation and mean shifts determine whether the residual-based T^2 control chart can be applied and that this chart can efficiently monitor most such processes.
出处
《清华大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2006年第3期403-406,共4页
Journal of Tsinghua University(Science and Technology)
基金
国家自然科学基金资助项目(70472010
70321001)
关键词
多元统计过程控制
自相关过程
T^2控制图
multivariate statistical process control
autocorrelatedprocess
Hotelling＇s T^2 chart
