An Unsupervised Anomaly Detection Approach for Spacecraft Based on Normal Behavior Clustering

This paper presents a new unsupervised anomaly detection approach for spacecraft based on normal behavior clustering. This method takes as input a set of unlabelled historical telemetry data and automatically detects anomalies within the data. After these abnormal data are removed, the method constructs system normal behavior model based on normal data clustering. Then at run-time, it monitors the status of the spacecraft and detects any anomalies appearing in the real-time telemetry data by checking deviations from the normal behavior model. The experimental results show that the method is efficient and practical for anomaly detection of spacecraft system.