Machine Learning For Spacecraft Time Series Anomaly Detection

Detecting anomalies in spacecraft telemetry is important due to the extreme operational environment of these systems. Detecting anomalies can serve as early warnings for potential system failures. Typically, a spacecraft system contains hundreds (or thousands) of telemetry channels. There is a need for automated anomaly detection to enhance efficiency and accuracy. This thesis focuses on developing methods for automated anomaly detection in spacecraft multivariate time series. We present a reconstruction-based method for spacecraft multivariate time series anomaly detection in the presence of non-anomalous spikes.

In reconstruction-based anomaly detection methods a network is used for time series reconstruction and anomalies are detected based on the error sequence between the reconstructed time series and the original time series. Many approaches cannot accurately reconstruct abrupt changes, such as spikes, which may result in false or missed detections. We use a Long Short-Term Memory (LSTM) autoencoder network for multivariate time series reconstruction and a post processing of the error sequence to detect at which time anomalies occur. We examine the performance of the proposed method by using a synthetic dataset characterized by having non-anomalous spikes, large periods and correlation among channels.