Fault diagnostic using improved CDKF and neural network for attitude sensor of satellite

The main aim of this paper is to develop a superior fault detection and isolation scheme (FDI) for the attitude sensor of a satellite. Towards this end, we present a generic data-driven prognostics framework utilized improved central divided-difference Kalman filter (CDKF) algorithm proposed in our previous work for fault detection and combine RBF neural network with fuzzy logic for fault identification process. The proposed method is applied to the micro satellite which employs three-axis magnetometer (TAM) and fiber optic gyroscope (FOG) as attitude sensors. Two types of the typical sensor fault scenarios are considered in the paper. The simulation studies demonstrate that our method shows better performance and capabilities, and thus is a good candidate for on-board fault diagnosis scheme for attitude sensors of satellite.