Health Indicator Construction of High-Speed Rotating Bearings in Aerospace CMG Based on Physics-Inspired Machine-Learning Approach
Diyin Tang,Mengtong Gong,Limei Tian,Jinsong Yu,Jiyang Zhang,Qiang Zhang
DOI: https://doi.org/10.1109/tim.2022.3208253
IF: 5.6
2022-10-08
IEEE Transactions on Instrumentation and Measurement
Abstract:Control moment gyro (CMG) is a critical component for attitude control of large spacecraft. To avoid its accidental failure caused by the performance degradation of high-speed rotating bearings, a metric, which is referred to as the health indicator (HI), should be constructed and monitored. Although the HI construction problem has been widely studied to evaluate the degradation process of machinery, the poor quality of telemetry data makes it impossible to apply existing solutions to the bearings in CMG. Therefore, a novel HI construction method based on energy theory and physics-inspired machine learning is proposed. Generally, the energy balance relationship of the system would change with the degradation of its components. Based on this idea, a physics-inspired convolutional neural network (CNN) model is established to formulate the relationship between monitoring signals and total energy consumed by the driving motor. Then a transfer learning strategy is developed to capture the changes in model parameters and perform HI construction. Compared with existing approaches, the proposed method does not rely on vibration signals or high-frequency data and can extract degradation information obscured by the controller and other influences. A real dataset collected from an aerospace CMG is utilized to verify the effectiveness of the proposed method. Results show that the constructed HIs have good performance in monotonicity, robustness, and time correlation. The application of constructed HIs for CMG anomaly detection is also discussed.
engineering, electrical & electronic,instruments & instrumentation