Adaptive generalized likelihood ratio control charts for detecting unknown patterned mean shifts

摘要

Purpose:To develop new adaptive control charts for monitoring processes subject to arbitrary and unknown patterned mean shifts. Summary:It is discussed that while traditional control charts are useful for detecting constant mean shifts, their performance is not good enough when the mean of the observed sequence exhibits a time-varying behavior after the fault occurrence. For this purpose, it is observed that some adaptive cumulative score (CUSCORE) schemes have been developed for detecting unknown patterned mean shifts. However, these CUSCORE schemes assume the occurrence of one-sided mean shifts only. Therefore, in order to overcome these limitations, an adaptive generalized likelihood ratio (AGLR) control chart is proposed in which a possible shift pattern in the mean is estimated using an exponentially weighted moving average combined with a wavelet smoother. Further, the AGLR control chart is developed for the observations from both normal and non-normal distributions. Simulations and real time examples are considered to study the performance of the proposed AGLR chart.