Spacecraft sensor data reliability improvement based on spatio-temporal information fusion model

Abstract

Spacecraft sensor data is crucial for evaluating the operating status and environment of spacecraft. However, due to factors such as component aging, space environment interference, and unstable satellite-to-ground communication, obtaining high-reliability sensor data is challenging. Additionally, the high dimensionality, complex correlations, strong temporal dependencies, and noise in sensor data further complicate efforts to improve data reliability. To address these challenges, this paper proposes a sensor reliability improvement method based on a multi-layer spatio-temporal information fusion model (STIFM). First, a moving average filter is applied to the raw data to reduce the impact of noise on modeling. Next, the Transformer model is used to establish data estimation models for different sensors in spatial scale and the same sensor in temporal scale. The outputs from these spatiotemporal models are then fused using particle filtering, and the uncertainty of the results is quantitatively assessed. Based on this, data anomaly detection and recovery are performed using the confidence interval of the STIFM output. Finally, the proposed method is validated using satellite flywheel on-orbit data. Experimental results show that the proposed method achieves at least 93.55 % accuracy in abnormal scenarios and significantly extends the mean time to failures (MTTF), outperforming existing methods. This indicates that the method proposed in this paper can effectively enhance the reliability of spacecraft sensor data.