Research On Life Forecasting Methods Of The Momentum Wheel Based On Support Vector Machine

With the continuous development of science and technology, the life of products is required more and more strictly.Being the key component of the posture control system in three-axis stabilization satellites, the Momentum Wheel’s life and reliability is an important index which decides the entire satellite’s life. Because the Momentum Wheel has the feature of higher reliability, longer life and small sample, most of the life prediction methods for the Momentum Wheel are the life prediction methods based on performance degradation data. However, the traditional life prediction methods based on performance degradation data are generally model-based approach, which requires pre-specified mathematical model of the degradation path. Because of the difference between each models and the complexity of Momentum Wheel’s performance degradation data, the traditional life prediction methods based on p erformance degradation data often can not achieve ideal forecast accuracy.For these seasons, we select the satellite momentum wheel as the object of the research with the bearing temperature data of the Momentum Wheel as the performance degradation data, and the research gives a new data-driven prediction algorithm based on support vector machines without pre-specified mathematical model of degradation path, but using support vector machine’s ability to learn from the strong performance degradation data to find the law, then building model on this basis to establish its life distribution model and assess its reliability level. This research provided a new idea of the Momentum Wheel’s life prediction. The works of the Thesis are as follows:(1) Using PSRT and GA to improve the SVR-AR algorithm,and giving the PSRT-SVR algorithm.(2) Combining improved SVR-AR algorithm and EMD algorithm to construct a new life prediction method based on EMD-SVR.(3) Selecting the bearing temperature data of the Momentum Wheel as the performance degradation data, and applying the new life prediction method based on EMD-SVR to forecast the life of the Momentum Wheel.Finally, the paper summarizes the work of the research, and further researches are discussed.