Abstract
Multiphase is an inherent nature of many batch processes, which should be taken into consideration to achieve efficient process monitoring. Although there have been several methods for phase identification, yet process information, especially process dynamic information, is not fully utilized. In this paper, an automatic phase identification method is proposed, which not only captures the differences in variable cross-correlations, but also explores the changes in process dynamics and model-explained variance. Accordingly, better phase identification, deeper process understanding and more efficient fault detection results are obtained. The practical application to an injection molding process verifies the effectiveness of the proposed method. © 2012 Elsevier B.V.