Statistical Modeling of Dynamic Multivariate Process Using Canonical Variate Analysis

For quality control or statistical process monitoring in an industrial process, the valid models are necessary. While the mechanical modeling is a time-consuming and difficult task, the statistical modeling is adopted in many industrial continuous. But most of these processes always have a large number of process variables and are usually operated under closed-loop control, yielding process measurements that are auto correlated, cross correlated, and colli near. In this paper, a statistical method named canonical variate analysis (CVA) is introduced to a chemical separation process. The proposed modeling method based on CVA focuses on canonical correlations using not only the past process outputs but also the future process outputs. The canonical variates are linear combinations of the past process measurements which explain the variability of the future measurements the most. Experimental results illustrate a good performance.