自相关过程的多变量统计过程监控  

Multivariate Statistical Process Control for Autocorrelated Process

作　　者：田海晶[1] 潘尔顺[1] 

机构地区：[1]上海交通大学机械与动力工程学院,上海200240

出　　处：《上海交通大学学报》2008年第3期496-499,共4页Journal of Shanghai Jiaotong University

基　　金：国家自然科学基金资助项目(50405016)

摘　　要：探讨了一种基于时间序列自相关移动平均(ARIMA)模型的自相关过程的多变量统计过程监控方法.以神经网络为基础建立预测模型,计算残差值,采用Hotelling’sT21控制图进行监控,结合MYT分解法进行故障诊断.并模拟受控和失控状态的具有自相关的多变量过程,运用常规控制图和残差控制图对案例进行了分析比较,说明了本文方法的有效性.A methodology on statistical process control for autocorrelated multivariate process based on ARIMA （Autoregressive-Integrated Moving Average） time-serial model was proposed. A neural network model is applied to calculate the residual of output-variation and Hotelling＇s T1^2 control chart of residuals is established to eliminate the autocorrelation effect; when the multivariate process is out-of-control, the MYT method is used to analyze and identify the cause. Autocorrelated multivariate data from in-control and out-of control processes are simulated and the case study of an individual control chart and a residual chart for the processes shows that the design is reasonable.

关 键 词：自相关性 多变量过程 统计过程控制 神经网络 

分 类 号：O213.1[理学—概率论与数理统计]