Joint-Individual Monitoring of Parallel-Running Batch Processes Based on MCCA

Abstract

A modern production plant may consist of several parallel-running batch processes, and the monitoring of such processes is imperative. This paper proposes a multiset canonical correlation analysis (MCCA)-based joint-individual monitoring scheme for parallel-running batch processes, which considers the individual feature of each batch process and the joint features shared by all batch processes. First, fourway batch process data are unfolded into two-way time-slice data. Second, MCCA is performed at each time instant to extract joint features throughout all running batch processes. Then, for each batch process, the measurements are projected onto a joint feature subspace and its orthogonal complement subspace that contains the individual features of the batch process. Finally, monitoring statistics are constructed to examine the joint and individual features. The proposed monitoring scheme is applied on a numerical example and the simulated parallel-running batch-fed penicillin fermentation processes. Monitoring results show the efficiency of the proposed approach.

Keywords

• Parallel-running batch processes
• fault detection
• multiset canonical correlation analysis
• process monitoring