Inner-Phase Analysis Based Statistical Modeling and Online Monitoring for Uneven Multiphase Batch Processes

Abstract

The multiplicity of operation phases is inherent in the nature of many batch processes, and each phase exhibits significantly different underlying behaviors. In addition, within each phase, normal processes in general follow certain underlying operation rules, called inner-phase evolution here, which however have not been addressed before. In this paper, a new statistical modeling and online monitoring method is proposed for multiphase batch processes. A two-level phase division algorithm is proposed to capture the process trend and trace inner-phase evolutions. It reveals that the inner-phase process in general goes through three statuses sequentially, i.e., transition, steady phase, and transition. Principal component analysis (PCA) and qualitative trend analysis (QTA) are combined to distinguish different inner-phase process statuses. Their different characteristics are then modeled and monitored separately, revealing more accurate process operation information. Meanwhile, the problem of uneven-duration batches is effectively handled in different inner-phase process statuses. The application to a typical multiphase batch process, injection molding, illustrates the feasibility and performance of the proposed algorithm.

Quality-related inner-phase evolution analysis and quality prediction for uneven batch processes

“…To reveal time-varying underlying process characteristics, process division is extended to two levels: conventional phase division and inner-phase division. Instead of tracking the changes of general process characteristics for process monitoring (Zhao et al (2013)), two-level phase division is performed by focusing on the quality-related process characteristics for the specific purpose of quality prediction.…”

“…A series of phases comprise a whole batch cycle and each phase has its own characteristic, which requires special attention for multiphase batch processes. Zhao and Lu et al (2004, 2008, 2013 proposed phase-based PCA/PLS methods, recognizing that phases can reflect the changes of the inherent process correlations. Considering the transition problem between neighboring phases, a soft-transition multiple PCA (STMPCA) method was proposed to detect and model transitions for online process monitoring (Zhao et al (2007)).…”

“…However, these phase-based methods ignore the process trend reflected by inner-phase variations, which may lose important information about process operation. Recently, Zhao et al (2013) investigated the inner-phase evolution of batch processes by dividing a phase into several parts and different statistical models were developed in different parts for process monitoring. In general, three sequential statuses, i.e., transition, steady part and transition, are the basic structure to describe the process variation within a phase, called inner-phase evolution.…”

Phase division and transition modeling based on the dominant phase identification for multiphase batch process quality prediction

“…Some works on transitions identification based on subspace separation are proposed (Zhao and Gao, 2012; Zhao et al, 2011). Later, Zhao et al (2013, 2014) proposed an inner-phase analysis method considering the inner-phase evolution. Although transition can be identified based on above methods, the specific number of phases and hard partitions are always needed before actual phase division.…”

Intra-Batch Evolution Based Process Monitoring for Multiphase Batch Processes

“…et al (2008Zhao C.H. et al ( , 2013a, Zhao L.P. et al (2012Zhao L.P. et al ( , 2013) for process monitoring and quality prediction focusing on multiphase characteristic of batch processes. Different statistical models were established to capture different characteristics of phases on the basis of such recognition that the underlying variable correlations are similar within the same phase.…”