A Novel Dynamic Process Monitoring Method Based on the Autocorrelation and the Cross-Correlation Analysis of Process Variables

This is the authors' abstract. We don't add key points for this paper.

Detecting anomalies is crucial to ensuring the safety and stability of the process industry. However, industrial processes often exhibit dynamic characteristics with process monitoring variables distributed across various spatial and temporal points. The intricate relationship between these variables poses challenges to process monitoring. Hence, this paper proposes a dynamic process monitoring method utilizing bidirectional two-dimensional principal component analysis to tackle this challenge. It constructs the monitoring sample with a two-dimensional structure, where the autocorrelation of process variables distributed across different time points can be monitored by conducting two-dimensional principal component analysis in the column direction, while the cross-correlation of the process variables distributed across different spatial points can be monitored by performing the two-dimensional principal component analysis in the row direction. The effectiveness of the proposed methods has been demonstrated in the Tennessee Eastman process and the zinc roast process. Results indicate a significant improvement in fault monitoring rates through the joint analysis of autocorrelation and the cross-correlation of monitoring variables.