Online Predictive Monitoring Using Dynamic Imaging of Furnaces with the Combinational Method of Multiway Principal Component Analysis and Hidden Markov Model

Furnace processes play a very important role in modern manufacturing industries. They are complicated transient processes and almost a "black box" for operators. It is rather difficult to diagnose those using classical methods, such as statistical classifications. In this Article, novel predictive video monitoring that utilizes prediction from the hidden Markov model (HMM) and multiway principal component analysis (MPCA) is proposed. MPCA is used to extract the cross-correlation among spatial relationships in the low dimensional space, while HMM constructs the temporal behavior of the sequence of the spatial features. Also, HMM can provide state-based segments, which allow predictive models to monitor signals at different time points. With the future predictions, the progress of the current operation can be tracked under a simple probability monitoring chart that shows the occurrence of the observable upsets in the future. A real furnace system is used to verify the effectiveness of the proposed method.