Nonlinear multiphase batch process monitoring and quality prediction using multi-way concurrent locally weighted projection regression

Abstract

The batch process has the characteristics of nonlinear and multiphase due to variation operation conditions. Nonlinear and multiphase modeling of the batch process is very important to understand the running state of a process and improve the monitoring effect. In this work, a multiphase multi-way concurrent locally weighted projection regression algorithm is proposed. Firstly, the entire process is partitioned into phases according to local quality-related characteristics and time sequence. Secondly, the nonlinear process is modeled with locally linear models in each partitioned phase, the global approximation results are obtained by weighting all the local models. Thirdly, the complete monitoring indices of quality-related and process related are built, and the quality variables are predicted while exploiting the regression structure for quality and process monitoring. Finally, a nonlinear numerical process and the penicillin fermentation process are used to verify the effectiveness of the proposed algorithm.