跳到主要导航 跳到搜索 跳到主要内容

Analytical Investigation of Anomaly Detection Methods based on Time-Domain Features and Autoencoders in Satellite Power Subsystem

  • Weihua Jin
  • , Bo Sun
  • , Zhidong Li
  • , Lei Zhang
  • , Shijie Zhang

科研成果: 书/报告/会议事项章节会议稿件同行评审

摘要

This study investigates the potential of autoencoders in detecting anomalies in satellite power subsystem telemetry data. An autoencoder-based ensemble method is proposed while providing a comprehensive comparison on different autoencoder types and time-domain features. Considering the unique learning mechanism of autoencoders, two specific methods have been designed to evaluate the autoencoder performance. The research results show that the combination of kurtosis feature and convolutional autoencoder has stronger detection ability for satellite power subsystem.

源语言英语
主期刊名Proceedings - 2021 8th International Conference on Dependable Systems and Their Applications, DSA 2021
出版商Institute of Electrical and Electronics Engineers Inc.
451-460
页数10
ISBN(电子版)9781665443913
DOI
出版状态已出版 - 2021
已对外发布
活动8th International Conference on Dependable Systems and Their Applications, DSA 2021 - Yinchuan, 中国
期限: 11 9月 202112 9月 2021

出版系列

姓名Proceedings - 2021 8th International Conference on Dependable Systems and Their Applications, DSA 2021

会议

会议8th International Conference on Dependable Systems and Their Applications, DSA 2021
国家/地区中国
Yinchuan
时期11/09/2112/09/21

指纹

探究 'Analytical Investigation of Anomaly Detection Methods based on Time-Domain Features and Autoencoders in Satellite Power Subsystem' 的科研主题。它们共同构成独一无二的指纹。

引用此