Search results

Following an alarm signal from a control chart it may be necessary to conduct an adjustment to bring the process mean towards a target value. Addresses the problem of how such an adjustment should be made, based on the use of an EWMA control chart, including the special case in which the EWMA is equivalent to the Shewhart chart. The distribution of the EWMA statistic, at the occurrence of an alarm signal, is approximated by applying a Markov chain approach. This distribution is used to obtain a maximum likelihood estimator of the process mean. Adjustment strategies are defined based on the EWMA statistic, the ML‐estimator and a modification of the ML‐estimator. Compares the performance of these adjustment strategies in a simulation study, although the Shewhart case allowed for a simplified treatment requiring no simulation. Results indicate that adjustment based on the ML‐estimator provides substantial improvement over the EWMA statistic when the smoothing constant of the control chart is moderate or high. When the smoothing constant is given a very low value the EWMA statistic provides the preferable adjustment strategy. Differences between control chart designs were less substantial than the differences between adjustment strategies for a given control chart.