Process Monitoring Based on the Multiblock Rolling Pin Vine Copula

Abstract

This paper presents a process-monitoring method based on multiblock rolling pin vine copula. The multiblock strategy not only reduces the impact of the local errors of the global variable during the monitoring process but also reduces the number of parameters that must be estimated and the number of copula trees. Introduction of rolling pin solves the problem that vine copula is difficult to accurately describe nonmonotonic data. The combination of multiblock and rolling pin reduces the risk of selecting a single reference variable to transform data. The root node has the greatest correlation with other variables. The root node in each sub-block is set as a reference variable, which makes the method have better robust characteristics. The good performance of this method during process monitoring is verified by a numerical example and a chemical-engineering example.