
Research Article
Importance Measurement of Parameters for Satellite Attitude Control System Fault Diagnosis Based on DBN
@INPROCEEDINGS{10.1007/978-3-030-69072-4_6, author={Mingjia Lei and Yuqing Li and Guan Wu and Junhua Feng}, title={Importance Measurement of Parameters for Satellite Attitude Control System Fault Diagnosis Based on DBN}, proceedings={Wireless and Satellite Systems. 11th EAI International Conference, WiSATS 2020, Nanjing, China, September 17-18, 2020, Proceedings, Part II}, proceedings_a={WISATS PART 2}, year={2021}, month={2}, keywords={Importance measurement Deep belief network (DBN) Satellite fault diagnosis Attitude control system}, doi={10.1007/978-3-030-69072-4_6} }
- Mingjia Lei
Yuqing Li
Guan Wu
Junhua Feng
Year: 2021
Importance Measurement of Parameters for Satellite Attitude Control System Fault Diagnosis Based on DBN
WISATS PART 2
Springer
DOI: 10.1007/978-3-030-69072-4_6
Abstract
Efficient and accurate fault diagnosis of satellite attitude control system has an important role and significance to ensure the reliability of satellites in orbit. Recent researches on satellite fault diagnosis focuses on diagnosis methods, but less on the importance of telemetry parameters. Since the satellite itself is a highly complex nonlinear system, there are many types of telemetry parameters that can be used for fault diagnosis. The importance of different parameters has a greater impact on fault diagnosis. Aiming at the above problems, a new DBN-based parameter importance measurement method (PIM-DBN) is proposed by constraining the DBN network structure and fixing some part of weights during the update process. The proposed method can automatically solve the importance weights of the telemetry parameters by training network with the CD-K divergence algorithm. This method was applied to the fault diagnosis of the momentum wheel of a satellite with 10 telemetry parameters. In order to verify the effectiveness of PIM-DBN, three diagnosis method (SVM, ANN and DBN) were used to classify the data set. The accuracy of there methods above achieved 87.54%, 68.83% and 93.45% respectively with using the weighted data. These results show that the proposed importance measurement method is effective for the data-driven fault diagnosis field and health assessment field.