Batch Process Monitoring Based on Multilinear Principal Component Analysis

Abstract

A new batch process monitoring based on Multilinear Principal Component Analysis (MLPCA) is proposed in this paper. In the existing vector-based method on batch process monitoring such as Multiway Principal Component Analysis (MPCA), a batch data is represented as a vector in high-dimensional space. But vectorizing the batch data will lead to large storage requirements and information loss. MLPCA can be used to deal with the three-way data (or tensor) directly instead of performing vectorizing procedure. Hence, MLPCA has some advantages such as low memory and storage requirements for Normal Operation Condition (NOC) model. Furthermore, the MLPCA is able to extract more meaningful information from the batch dataset. The MLPCA monitoring approach is tested with the data from a Reactor of Thermal Anneal (RTA) batch process. Simulation results show that MLPCA early finds the process fault and improves the accuracy of process monitoring compared with MPCA.