Multivariate process control chart based on sign and signed-rank tests for mixture distributions

Abstract

The development and application of the traditional multivariate parametric control charts have been distribution-based procedures. This is because the process output is assumed to come from some specified probability distributions. However, in many quality/ process control applications, these distributional assumptions are either not satisfied or there is simply lack of enough information to verify the assumptions and this condition drops the performance of the charts. The purpose of this paper is to develop nonparametric control charts for mixture of multivariate normal and t-distributions with the view to evaluating the performance of the charts in simulation study. The results show that the charts are effective in monitoring process variability.