Profile Monitoring of Multivariate Processes for Efficient Detection of Parameter Changes

Abstract

Profile monitoring has been extensively studied when the profile of a quality characteristic is normally distributed. There are a limited number of studies for the case when a profile follows other distributions such as Weibull, lognormal or Gamma. A profile, having these last three distributions, has many practical applications. It is also of interest to determine how well profile-monitoring of estimators can detect changes in parameters of an underlying distribution. Control chart methods are utilized to monitor such parameter estimates. The performance of a few monitoring statistics is investigated in this dissertation. A form of the Hotelling’s T2 statistic and another that utilizes the concept of an exponentially weighted moving average are investigated. As a performance measure, the mean and standard deviation of the time to first detection of shift in process parameters, are explored. We also investigated a method that utilizes estimated percentiles of the distribution for profile monitoring.