Abstract:

To address the issue of poor robustness to observational noise in thruster fault diagnosis methods for On-Orbit Service （OOC） satellite， a diagnosis method based on the relative position information of OOC is proposed. Firstly， a 2-fault-type model is constructed based on the mechanism of the propulsion system. A data fusion method based on the posterior covariance matrix is used to process relative posedata， preventing excessive measurement errors in relative pose from leading to misdiagnosis. Subsequently， an adaptive EKF is used to solve the problem of inaccurate prediction step when thruster failure occurred. The influence of noise on the fault diagnosis algorithm is analyzed， and a fault diagnosis and isolation strategy based on Generalized Likelihood Ratio （GLR） is proposed. Numerical simulation results show that the filtering error of the proposed adaptive EKF was reduced by about 80% compared with that of traditional EKF in the case of faults， and the speed of fault estimation is increased by 34%. When the thrust change under fault is subtle， the fastest diagnosis speed for stuck fault is 10 steps， and the minimum diagnosis error is about 0.5%. For efficiency reduction faults， the fastest diagnosis speed were 38 steps， and the minimum diagnosis error was 3%. At last， An on-orbit mission scenario was constructed， and the LQR control law was used to control the OOC to approach the target satellite in a near-field. The efficiency reduction fault of the OOC’s thruster is simulated in the process， which takes 4 s for the algorithm to detect and isolate fault and 50 s to obtain efficiency estimation， which verified the validity of the proposed method.

Key words: adaptive Kalman filter, thruster, observer, fault detection, on-orbit service