Neural network-based fault diagnosis scheme for satellite attitude control system

In order to improve reliability of on-orbit satellite, this paper proposes a novel fault diagnosis scheme based on BP neural network method for satellite attitude control systems (ACSs) subject to external disturbances of the system. First, a model of the satellite ACS is established. Second, the BP neural network is introduced in detail, which includes its structure and learning algorithm. Then, a fault diagnosis scheme based on the BP neural network observer is designed and further a recursive multistep prediction method is presented to train the network. Furthermore, the prediction output of the observer is used to compare with the real output of sensors to generate residual that is further used to diagnose faults of the system. Finally, numerical simulations of a satellite ACS are performed to illustrate the effectiveness and potential applications of the proposed scheme.