Automated Interpretation of PCA-Based Process Monitoring and Fault Diagnosis Using Signed Digraphs

Abstract

Listen

Translate article

Principal component analysis (PCA) has emerged as a useful tool for process monitoring and fault diagnosis. The general approach requires the user to identify the root cause by interpreting the measured variable contributions to the residual and/or principal components. This could be tedious and often impossible for a large process. It also hinders the automation of high level supervisory tasks like choice of corrective actions and their associated costs. In this paper, the interpretation of the PCA-based contributions is automated using signed directed graphs (SDGs). The implementation of the PCA-SDG based fault diagnosis algorithm is done using Gensym's expert system shell G2. Its application is illustrated on the Amoco Model IV Fluidized Catalytic Cracking Unit (FCCU).