Transformer-based fault diagnosis method for fixed-wing UAV elevator

To address the issue of low aircraft elevator fault diagnosis efficiency, this paper proposes a Transformer-based deep learning fault diagnosis algorithm aimed at enhancing flight safety and fault resolution efficiency. First, data acquisition and processing is described in detail, encompassing data sources, acquisition methods, pre-processing, and feature extraction steps. Subsequently, a fault diagnosis model is designed and assessed in both the training and testing phases. Experimental validation is conducted using actual flight data from an electric drogue fixed-wing unmanned aerial vehicle (UAV). The paper analyzes the accuracy of the diagnostic model, the trajectory of loss values during iterations, and a comparative assessment against various algorithms to affirm the model's effectiveness and accuracy. Lastly, the paper summarizes the experimental findings and puts forth potential directions for future improvements.