中文摘要：

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’s T1^2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．

英文摘要：

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.