Fault Detection of Multimode Processes Using Concurrent Projection to Latent Structures

Abstract

Process monitoring based on Concurrent Projection to Latent Structures (CPLS) has been proposed recently as an efficient process monitoring tool to detect the input-relevant and output-relevant faults. In this paper, a novel multimode process monitoring approach is proposed to introduce external analysis into the CPLS monitoring framework. The process variables are divided into external variables, main variables and the output variables. The mode change is detected by monitoring the variation of the external variables through a mode-detection CPLS. The prediction of the main and quality variables by external variables is done by a PLS. After the influence of the external variables is removed from the main and quality variables, another process-monitoring CPLS is built to detect the main-variable-relevant faults and quality-variable-relevant faults. The proposed approach is illustrated with a simulation process.