Hierarchical plant-wide monitoring and triangular representation based diagnosis

A hierarchical plant-wide monitoring methodology is proposed based on hierarchical decomposition and principal component analysis to handle the complexity and interactions among process units. To obtain more reliable diagnostic results, pattern matching techniques are utilized along with the triangular representation of process trends in the reduced space that is defined by principal components. The proposed methodology is illustrated from the application to the hierarchical plant-wide monitoring of Tennessee Eastman benchmark process. The application has shown that the proposed methodology offers a fast and reliable monitoring and diagnosis for a large scale chemical plant. Copyright (C) 1998 IFAC.