Skip to Main Content
A neural adaptive observer (NAO) based fault detection and identification (FDI) strategy for a class of nonlinear systems is presented in this paper. The observer input is designed in a structure similar to feedback neural networks. The parameters in the NAO input are updated by using the extended Kalman filter (EKF) algorithm. The convergence of the learning process is analyzed in terms of a quadratic Lyapunov function. Moreover, stability of the observer input and the NAO-based system are investigated respectively. Finally, the proposed FDI strategy is applied to a micro-satellite attitude control system. Several simulation results demonstrate that the NAO based FDI method can detect and specify both abrupt and incipient faults with satisfactory performance.