Actuator Fault Detection of Satellite Based on Neural Network Observer

Satellite attitude control system is an important subsystem that ensures the normal operation of the satellite. Aiming at the problem that the model-based fault detection method relies on modeling accuracy, an actuator fault detection method based on neural network observer is proposed. The neural network is trained with the error back propagation algorithm including a correction term to ensure the accuracy and weight bounded. The neural network is used to identify and estimate actuator failure. The stability of the observer is proved by Lyapunov method. Simulation results show that this method can effectively detect and estimate actuator failure.