A double multivariate exponentially weighted moving average control chart for a process location monitoring

A double multivariate exponentially weighted moving average (dMEWMA) control scheme for detecting shifts in the mean vector of multivariatelly normally distributed quality characteristics was proposed. It was shown that the ARL performance depends on the mean and variance-covariance matrix only through the non-centrality parameter value. Through the Monte Carlo simulation, the design of the dMEWMA control scheme was constructed. The performance of dMEWMA control schemes for detecting various shifts was compared to the competing control schemes, namely, MEWMA and Hotelling's chi2. An example to demonstrate the construction of the dMEWMA control scheme was introduced. It was concluded that dMEWMA control scheme outperformed MEWMA and Hotelling's chi 2 control schemes for small and larger shifts. The dMEWMA control scheme was optimal for larger values of the smoothing parameter and performed much better than MEWMA for very small shifts. These properties are desired by industry.