The large-scale chemical production process contains many units,and the variables within the units have a very strong correlation.In view of this characteristic,a fault detection method based on sub-block canonical variate analysis(SB-CVA) was proposed.Firstly,the process modeling data were divided into several sub-blocks according to the principle of high correlation of intra-block variables and low correlation of inter-block variables.Then CVA models were established within each block,T² and SPE statistics were calculated as fault detection indicators.The proposed method could detect the fault of sub-blocks in time and confirm the fault unit effectively.Simulation was carried out through TE process and compared with PLS,MB-PLS and CVA to verify the effectiveness of the proposed method.

Fault Detection of Multi-Modal Process Based on Multi-Model SVM [J]. Journal of Shenyang University of Chemical Technology, 2023, 37(6): 533-541.