Research on Spacecraft Anomaly Detection Method Based on Transformer

Spacecraft anomaly detection is a critically important task in the modern aerospace field, essential for ensuring spacecraft safety. This paper presented a method for spacecraft anomaly detection. The method utilized one-dimensional time series of spacecraft parameters as raw data to create a dataset. Subsequently, a Transformer-based regression model was employed to analyze the dataset. The difference between predicted values and actual values was used to determine a threshold range for identifying anomalies in the samples. Experimental results indicated that this method outperformed traditional algorithms such as random forests, decision trees, and support vector machines in anomaly detection.