Machine Learning for Reaction Wheel Fault Detection using Simulated Telemetry Data
João Vaz Carneiro, Hanspeter Schaub, Morteza Lahijanian, Kendra Lang, Konstantin N. Borozdin
University of Colorado Boulder Verus Research (United States)
阅读操作
确认中在文库中上传 PDF 后可生成中文音频讲解。
摘要与影响
View Video Presentation: https://doi.org/10.2514/6.2022-2507.vid Autonomous fault detection in satellites is an enticing proposition that will be essential to prevent catastrophic failures and guarantee the success of future missions. The current approach relies on highly qualified human technicians to search for anomalous behavior in large amounts of telemetry data. This method raises some challenges; as the complexity of missions increases, the ability for those technicians to detect faults declines. Furthermore, the growing number of missions with thousands of satellites in complex constellations makes traditional fault detection techniques unfeasible. Machine learning algorithms can be used to solve these issues. They require less human interaction and can be used in large operations with multiple active satellites. This paper proposes supervised and unsupervised learning algorithms to detect whether a fault has occurred and when it happened. The focus is to study faults related to reaction wheels and their subsystems, such as wheel speed encoders. A high-fidelity simulation is developed using the Basilisk software tool and the produced synthetic telemetry stream is fed into the machine learning algorithms for fault detection. The approaches used proved successful in detecting faults and underlined how some data parameters are more useful for fault detection than others.
逐年被引趋势
关键指标
同类平均 = 1
同领域 · 同年份 · 同类型
Google Scholar 与 OpenAlex 的被引统计范围不同,数值存在差异属正常。
AI 辅助阅读
依据:摘要
可就本文提问;依据不足时会说明。
学术脉络
学科主题
工程Fault Detection and Control Systems
Engineering and Test Systems · Machine Fault Diagnosis Techniques
参考文献 12
此处列出前 3 条
引用本文 9
按被引量排序,此处列出前 3 条