This thesis is concerned with the design of control charts for autocorrelated data which are common in continuous flow processes, and which lead to increase in the number of false alarms when conventional control charts are applied. The usual approach to monitoring autocorrelated data is to apply residual control charts in which forecast errors from a time series model are plotted on the charts. In some situations where the positive autocorrelation is high, however, the residual control charts are insensitive to the shifts in the mean.
For typical time series models, data patterns of residual control charts are identified by computing the expected values of residuals after the mean shift occurs. Based on the change patterns, runs rules are proposed to improve the performances of the residual control charts. Simulation studies conducted to evaluate the performances of the proposed control charts indicate that there exists an effective runs rule for each time series model.