Purpose: To propose a multivariate cumulative score (Cuscore) control chart for monitoring the mean vector shift in autocorrelated multivariate process. Summary: It is discussed about the features of univariate and multivariate statistical process control charts that are useful in monitoring the process for its state of control. When the observations available under study are independent, there are as many traditional control charts available for use. However, as the intervals between observations become shorter, the autocorrelation may occur and lead to high false alarm rate in traditional charts. Though there exist many multivariate control charts of different situations, in this article multivariate Cuscore (MCuscore) control chart is proposed for monitoring the mean vector of an autocorrelated multivariate process. This approach is, in fact, based on the likelihood ration test and fault signature analysis techniques. Further, the MCuscore charts for specific signals due to transient mean shift, sudden steady mean shift and ramp mean shift are also designed. It is observed that the MCuscore chart outperforms the traditional multivariate chart in detecting shifts in autocorrelated processes. (39 refs.)
展开▼
