Two robust multivariate exponentially weighted moving average charts to facilitate distinctive product quality features assessment

Abstract

This paper offers new multivariate statistical process monitoring schemes to study the process shift supported by a distinctive product quality features assessment. The proposed procedures help determine whether one or more features have undergone a shift or whether the shift is in the dependence structure but not in the study variables. We use the marginal distributions and pseudo copula observations to effectively apply feature-wise rank-based Lepage and Cucconi tests. The rank-based statistics induce nonparametric properties to the proposed chart. Therefore, its in-control performance is highly robust and nearly distribution-free. It is shown that the new chart gives better results for detecting scale shifts in one or more quality variables than some representative existing nonparametric charts. Some Monte-Carlo simulation studies have been performed to establish the effectiveness of the new charting scheme. A real application involving the production quality of cork stoppers is considered to illustrate the use of the proposed schemes in manufacturing and production. Some encouraging product feature assessment properties are observed.