Actuator fault detection in nonlinear uncertain systems using neural on-line approximation models

This paper describes actuator fault identification in unknown, input-affine, nonlinear systems using neural networks. Neural net tuning algorithms have been derived and identifier have been developed using the Lyapunov approach. The paper defines and analyses the fault dynamics i.e., the dynamical properties of a failure process. A rigorous detectability condition is given for actuator faults in nonlinear systems relating the actuator desired input signal and neural net-based observer sensitivity. Sufficient conditions are given in terms of the input signal and related actuator fault such that a fault can be detected. Simulation results are presented to illustrate the detectability criteria and fault detection in nonlinear systems