Process Monitoring for Multimodal Processes With Mode-Reachability Constraints

For increased efficiency and profitability, many processes have multiple modes of operation. Switching between different operating modes is performed according to the standard operating procedures. These procedures are set by considering safety and operating limitations of various subsystems and equipment, and thus put restrictions on the switching of the process modes. In this paper, a hidden Markov model based monitoring method is proposed that can not only handle the multimodality of process data but can also capture the mode switching restrictions. A two-step Viterbi algorithm is proposed for effective mode detection in the event of faults, and a reconstruction-based fault isolation algorithm is developed to build the contribution plots. Application examples demonstrate the effectiveness of the proposed monitoring method.