Evaluating the performances of statistical and neural network based control charts

Control chart is used widely in many fields and traditional control chart is no longer adequate in detecting a sudden change in a particular process. So, run rules which are built in into Shewhart X¯ control chart while Exponential Weighted Moving Average control chart (EWMA), Cumulative Sum control chart (CUSUM) and neural network based control chart are introduced to overcome the limitation regarding to the sensitivity of traditional control chart. In this study, the average run length (ARL) and median run length (MRL) in the shifts in the process mean of control charts mentioned will be computed. We will show that interpretations based only on the ARL can be misleading. Thus, MRL is also used to evaluate the performances of the control charts. From this study, neural network based control chart is found to possess a better performance than run rules of Shewhart X¯ control chart, EWMA and CUSUM control chart.