A fast monitoring method for multiple operating batch processes with incomplete modeling data types

• A novel up-down multi-model dynamic PCA is proposed with incomplete modeling data.
• Multi-model structure is constructed along with the different cluster number.
• Selection method of the best monitoring model from multiple models is described.
• The proposed approach shows better performance than MPCA.

For complex industrial processes with multiple operating conditions, a novel fast monitoring method, called up-down multi-model dynamic principal component analysis, is proposed with incomplete modeling data types in this paper. The method firstly classifies the process into several stages according to the operations. In each stage the data with similar operations are clustered together. Multiple model structures are constructed along with the variation of the cluster number. When on-line monitoring, the up-down monitoring method is proposed to monitor a new batch type. The effectiveness of the proposed monitoring method is demonstrated through the 120 t ladle furnace (LF) steelmaking process.