An approach to nonlinear fault diagnosis based on neural network adaptive observer

A fault detection method based on neural network online approximation structure for uncertain nonlinear systems is presented. A neural network observer is used to learn the nonlinear fault functions to monitor the abnormal behavior of the dynamic system. When system faults occur, the online learning structure can approximate all possible unknown faults, then the faults are identified and accommodated. The uniformly ultimately bounded stability of the closed-loop error system is guaranteed by Lyapunov stability theory and the weights are tuned without need of persistency of excitation.