PCA‐based fault diagnosis in the presence of control and dynamics

In PCA‐model‐based fault diagnosis, using the isolation enhancement approaches of analytical redundancy, mis‐isolation of some sensor and actuator faults may arise if the training data is collected under constant control. A detailed analysis is provided of how control actions affect PCA‐model‐based diagnosis. Ratio and feedback control in linear static and discrete dynamic systems, with full and partial PCA models is investigated. It is shown that all that it takes to eliminate the adverse effects is to vary the control set point (in feedback control) or the ratio coefficient (in ratio control) in the course of collecting the training data.