Inter-batch-evolution-traced process monitoring based on inter-batch mode division for multiphase batch processes

Abstract

The multiplicity of operation phases is an inherent nature of many batch processes and each phase exhibits significantly different behaviors. Besides process variation along the time direction within each batch, process variation along the batch direction, called inter-batch evolution here, widely exists in batch processes. Inter-batch evolution has caused different operation modes along the batch direction, which may not be well captured by a single model. In this work, the inter-batch evolution is tracked, based on which multiple modes are separated and modeled for online process monitoring along the batch direction. First, a batch cycle is divided into multiple phases. Then sliding windows are constructed for analysis of the inter-batch evolution within each phase. Based on the comparison between reference windows and sliding windows, different process modes are separated in order along the batch direction. Four monitoring subspaces which cover different types of relative variations are used in monitoring. During online monitoring, these different variations are then supervised respectively to trace the inter-batch evolution. The application to a typical multiphase batch process with inter-batch evolution, injection molding start-up process, illustrates the feasibility and performance of the proposed algorithm.