Sensor Gain Fault Diagnosis for a Class of Nonlinear Systems

Identifiability of sensor gain faults is studied and a novel approach to fault detection and diagnosis of a class of nonlinear systems is presented. Sensor gain faults are divided into two classes: the conditionally identifiable faults and the conditionally detectable faults. An algorithm is proposed to obtain the number and location of the conditionally identifiable faults. Asymptotic estimation of the conditionally identifiable faults is achieved via the use of an unknown-input observer and an adaptive rule when there are no conditionally detectable faults in the system and a linear matrix inequality is satisfied, and a detection observer is presented to monitor the conditionally detectable faults. The emphasis of this paper is put on unstable systems. Finally, simulation results on a single-link flexible joint robot are presented for illustration.