Anomaly Detection in Spacecraft Telemetry using Similarity Metrics and Isolation Forest

Anomaly detection plays key role in spacecraft operations. One way of classifying anomalies is to find deviation in parameter trend by comparing it with previous data. This study aims to find such anomalies by using distance metrics like Euclidean, Cosine and using isolation forest to find out outliers in the data. Application of this method has been verified using temperature sensor data of spacecraft. To evaluate the model, outputs generated were verified with known anomalous observations and manual analysis was done for other observations. The results obtained are promising with high detection and low false alarm rates.