A sliding mode observer approach for fault detection and diagnosis in uncertain nonlinear systems

Presents a sliding mode observer approach to fault detection and diagnosis for nonlinear systems with uncertainty having unknown bounds. The robustness properties of the observer ensure that no false alarms are registered due to uncertainties and disturbances in the system. The observer uses nonlinear gains that are smoothed versions of classical sliding mode gains and they are continuously updated to guarantee a globally stable observation error. A neural network is designed to capture the nonlinear characteristics of faults. Finally, simulation results have shown the feasibility and effectiveness of the method.