システムの健康状態を表現する変数を用いた異常検知手法の検討

抄録

In spacecraft systems, it is important to quickly detect and deal with anomalies. However, the increasing complexity of spacecraft systems makes it difficult to avoid failures using previous anomaly detection methods. In recent years, anomaly detection methods based on machine learning have been studied. But it is difficult for a spacecraft to detect anomalies autonomously in orbit. In this study, we propose an anomaly detection method that defines and monitors a single variable representing the health status of a spacecraft system, instead of monitoring a huge amount of telemetry data. We have confirmed that the single variable can represent the health status of a spacecraft system and can be applied to the anomaly detection method.