Between-Mode Quality Analysis Based Multimode Batch Process Quality Prediction

Abstract

In batch processes, multiple process modes usually exist because operation conditions frequently change to satisfy different market requirements. Besides multiple modes, the multiplicity of operation phases within each batch is an inherent nature of many batch processes. Different between-batch modes and different within-batch phases may exhibit significantly different underlying behaviors and have different effects on the final product quality. Therefore, for better quality predictions, the multimode and multiphase problem should be addressed. In this work, based on between-mode quality analysis, prediction strategy is developed for multiphase batch processes with multiple operation modes. First, each batch cycle is divided into multiple phases. Then within each phase, for different modes comprised of different batches, different quality regression models are constructed. Also, between-mode quality analysis is conducted to judge the relationship between the new mode and the historical modes in the mode library, regarding whether a new model is necessary to update the mode library. In the between-mode analysis, a novel model for the new mode is built by a simple linear combination of historical modes, which is then compared with the model obtained by complete remodeling for the new mode. Online quality prediction for the new mode is conducted using the selected model. The proposed algorithm is applied to a typical multiphase batch process with multiple operation modes, an injection molding process, to illustrate the feasibility.