Diagnosis of Parametric Faults Based on Identification and Statistical Methods

This paper proposes a fault diagnosis approach for detection and diagnosis of parametric faults in linear dynamical systems. It is assumed that the system is modeled by an auto-regressive (ARX) model, and a recursive least-squares (RLS) algorithm with variable forgetting factor is used for online parameter identification. The fault detection task is done using a statistical principal component analysis (PCA) method applied to ARX parameters. The fault isolation is done using an influence matrix method. The proposed methodology is applied to a simulation model of a DC motor under closed-loop control, and the performance is discussed.