An evaluation of forecast-based quality control schemes

Abstract

High volume production processes and many processes using automated sampling technology yield process data which are autocorrelated. One technique proposed for monitoring autocorrelated data involves the application of the Individuals control chart to forecast residuals from an appropriate time series model of the process. This study examines the following issues concerning forecast-based monitoring schemes: (i) the affect of forecast recovery from step changes on the average run length (ARL) of control charts applied to forecast residuals (ii) the proposal of the cumulative distribution function (CDF) of the run lengths as an appropriate criterion for chart comparisons and (iii) the relative performance of the Individuals control chart, the Cumulative Sum (CUSUM) control chart and the Exponentially Weighted Moving Average control chart applied to forecast residuals using both the ARL and CDF criteria