Fault Diagnosis of Tennessee Eastman Process Based on Static PCA

Principal component analysis (PCA) as a basic technique of multivariate statistical process control (MSPC) was successfully applied to monitor large-scale plants, where it is used for modeling process data. To detect abnormal situations, PCA utilize several types of monitoring statistics, where the most popular is the squared prediction error (SPE). After detection, the root cause of the fault must be identified. For that purpose, many approaches have been proposed and developed in the literature. The aim of this work is to study the ability of static PCA to monitor the predetermined Tennessee Eastman faults where the static PCA model is constructed using Tennessee Eastman process (TEP) training set.