Proposed multivariate control chart using winsorized Mean based on the |S| chart

Abstract

The global quality of a product is often subject to the simultaneous evaluation of several of its features; under these circumstances, it becomes necessary to implement a multivariate analysis. The variance-covariance matrix is one of the most important quality control statistics in multivariate settings and it can be monitored using different methods, such as the Shewhart, CUSUM and EWMA charts. In this article, a construction method for the multivariate control chart |S| is proposed using the Winsorized Mean as the mean vector estimator. It was demonstrated that this modification was more efficient when the sampled data hat outliers. This study shows a comparison between the traditional multivariate control chart and the proposed chart, which was found to have more sensitivity.