Spacecraft Fault Warning Method Based on Adaptive Trend Filtering

The on-orbit spacecraft monitoring system currently has a weak trend fault warning capacity that is typically relied on manually summarizing expert knowledge to identify abnormalities in advance. Furthermore, the expertise of the person in charge largely determines the breadth, depth, and accuracy of the knowledge. While the development of manually enhanced warning knowledge may also be hindered by the specific vanations of individual machines, various working modes, and environmental influencing elements. This work suggests an universal spacecraft fault warning method based on adaptive trend filtering in order to improve the automation of trend fault warning in the on-orbit spacecraft monitoring system while assuring the warning accuracy. Using the Ll trend filtering algonthm, a spacecraft fault warning model is created by fusing on-orbit telemetry data with knowledge of the mechanics of the spacecraft, expanding the fault samples to create a sample set for the improvement of the warning knowledge. On this basis, the simulated annealing algorithm is used to solve the key parameters in the spacecraft fault warning model, so as to quickly establish model for different spacecraft and provide strong support for the spacecraft on-orbit fault warning work.