A modified multivariate EWMA control chart for monitoring process small shifts

In this paper, a novel data-driven approach is presented to monitor processes influenced by gradual small shifts. The primary idea is to first build multivariate exponentially weighted moving average (MEWMA) model based on the originally measured variables to keep the memory effect of the process trend. Then introduce a unified Mahalanobis distance based monitoring statistic, which makes full use of the feature of the normal distribution of the process variables, to better capture the deviation of the process variables. A case study of the Tennessee Eastman process (TEP) is used to demonstrate the superiority of the proposed method over other conventional ones in performance and workload of the gradual small shifts monitoring.