Study of the Long-Term Performance Prediction Methods Using the Spacecraft Telemetry Data
Fang Hongzheng,Ma Haodong,Luo Kai,Han Liming,Xiong Yi
DOI: https://doi.org/10.1109/phm.2012.6228775
2013-01-01
Abstract:The study of the changing trends of the performance of the spacecraft is helpful to the estimation of the remaining usage life (RUL), which is the important basis of the realization of the spacecraft prognostics and health management. The types of the spacecraft telemetry data include the slow change, periodic change, abrupt change and the combination of the above three changes. This paper firstly analyzes the spacecraft telemetry data change rule. Secondly, the telemetry data is decomposed into 3 types, which is the trend, seasonal and random component, through the method of X-11. After that, the different prediction methods, such as the AR linear regression method, the BP neural network method, the nonparametric regression method, are taken respectively to predict the long-term performance trend of the above 3 types of the telemetry data. Finally, the predicted data using the above methods are fused to form the final prediction result, and the decay factor between the predicted data and the original data is computed. Furthermore, the experiment result shows the proposed prediction method can be effectively applied to the prediction of the performance trend of the spacecraft telemetry data, and has strong practical significance in the field of the spacecraft engineering project.