Evaluation of the effect of parameter estimation on the performance of exponentially weighted moving average control chart

The first step in process monitoring with control charts is to estimate process parameters. The effect of this phase, called Phase I, on online process monitoring (Phase II) performance is known to be significant. In this study, presence of contaminated data in the control chart design is discussed. By simulating contaminated data to represent out of control conditions of the process in Phase I and using Shewhart I Control Charts in this phase, process parameters are estimated. Performance of the Exponentially Weighted Moving Average (EWMA) Control Charts designed by using these parameters for Phase II are calculated with Integral Equation approach. Combinations of different parameter estimators (Sample Standard Deviation, Mean Moving Range, Median Moving Range, Root Mean Square Successive Differences, Median Absolute Deviation, Interquartile Range) as well as Phase I control chart parameters (sample sizes of 50, 100, 200, 500, 1000, 10000 and control limit width L in the interval of [2, 4] with 0.1 increments), Phase II performances are evaluated by Average of Average Run Length (AARL) and Standart Deviation of Average Run Length (SDARL) metrics. Design suggestions of the Phase I and Phase II control charts are provided for practioners.