A new statistical method for recognition of control chart patterns

This paper presents a new statistical method for control chart pattern (CCP) recognition, based on Bayesian inference and Maximum Likelihood Estimation. In this method, by assuming the existence of each pattern, the Maximum Likelihood Estimator of pattern parameter(s) is obtained and then a measure called Belief is determined. Beliefs denote the probability of existence of each pattern in the process. Using Bayes' Rule, we update beliefs recursively through the observations window points and the pattern with the greatest belief is recognized. Simulation results show the accuracy of the new method to detect the abnormal patterns as well as satisfactory results in the estimation of pattern parameters.