A Novel MPIPCR Diagnosis Algorithm with Quality-Related Faults for TEP

Multivariate statistics, such as principal component analysis (PCA), principal component regression (PCR) and partial least squares (PLS), etc., has been putted a broader exposure by the researchers. Typically, the improved PCR has strengthened the detection ability for quality-related faults. However, its detection ability for both quality-unrelated faults and regression faults still needs to be promoted. Considering the advantages of Mahalanobis distance and Pearson coefficient, they all can compare the relevance of two samples. Therefore, both Mahalanobis distance and Pearson coefficient are all employed to do a comparison for the procedure variables and quality variables, respectively. Then, the procedure variables of the highest quality-related are selected for modeling before IPCR, which defined MPIPCR. MPIPCR not only keeps the virtue, the fault detection ability for the quality-related faults, of IPCR but also adapts to the quality-unrelated faults and regression faults. The Tennessee Eastman Process (TEP) is applied to the comparison of PLS and MPIPCR in the simulation to verify the validity of the results.