Anomaly Detection of Spacecraft Reconstructed Signals Based on Attention Mechanism

Compared with the traditional anomaly detection methods, machine learning algorithms do not rely on manual and have the ability to extract advanced features of data. However, anomaly detection of spacecraft telemetry data by supervised machine learning is a challenging problem due to the lack of priori knowledge. This paper presents a signal anomaly detection algorithm based on attention mechanism. First, the long-distance characteristics of spacecraft telemetry data are captured by attention mechanism. Then, the stacked autoencoder compresses the data dimension and reconstructs the input signal to obtain the error reconstruction sequences. Furthermore, the anomaly indexes of the error reconstruction sequences are marked by the window threshold method to realize the anomaly detection of the spacecraft telemetry signal. Finally, the effectiveness of the algorithm is verified by a group of examples based on multi-channel spacecraft telemetry signals.