Copula-based monitoring schemes for non-Gaussian multivariate processes

Abstract

Multivariate statistical monitoring charts are efficient tools for assessing the quality of a process by identifying abnormalities. Most commonly used multivariate monitoring charts, such as the Hotelling T2 rule, however, assume the availability of uncorrelated Gaussian observations. Unfortunately, very often, real data do not satisfy these assumptions, and thus limit the usefulness of these techniques in practice. Furthermore, in many real applications, changes can occur in the shape of the multivariate distribution of the process while its mean or variance remains the same. Conventional process monitoring charts, such as the T2 chart, fail to detect such changes in the distribution. In th..