Time-Frequency Vision Transformer Based on Fault Diagnosis Method for Spacecraft Attitude Control System

The data-driven fault diagnosis methods of spacecraft attitude control systems are rapidly advancing, becoming a key technology for ensuring the stable operation of spacecraft. However, existing methods have certain limitations in capturing and representing temporal features. To address these issues, this paper proposes a Time-Frequency Vision Transformer (TFVT) model for recognizing fault categories in spacecraft attitude control systems. Firstly, we apply the wavelet transform to the raw signals from the spacecraft semi-physical simulation platform, generating Time-Frequency Representations (TFRs). We introduce a new data segmentation method to effectively capture the temporal features. Secondly, experiments are conducted using the dataset to analyze the model's performance and fault diagnosis results. The results indicate that the method demonstrates significant effectiveness in fault diagnosis, achieving a high level of accuracy. Finally, through comparative analysis with other advanced methods, we demonstrate the effectiveness and superiority of the TFVT method in fault diagnosis.