A Novel Dual Attention Convolutional Neural Network Based on Multisensory Frequency Features for Unmanned Aerial Vehicle Rotor Fault Diagnosis

By virtue of their convenience, reasonable cost and high efficiency, Unmanned Aerial Vehicles (UAVs) have been widely applied in every aspect of life. However, complicated operating conditions are prone to causing mechanical failure in UAVs, especially the rotor fault. Therefore, a novel dual attention convolutional neural network based on multisensory frequency features is proposed for UAV rotor fault diagnosis in this study. Firstly, according to the collected multisensory acceleration vibration signals of UAV rotors, time and frequency features in different health states (normal, rotor broken and crack fault) are compared and analyzed in detail. Secondly, a novel dual attention mechanism is proposed to not only focus on the effect of different sensors but also different frequency features of UAV. Moreover, it could adaptively assign larger weight to more important features to improve the fault diagnosis accuracy. Finally, a one-dimension convolutional neural network is adopted to extract the feature of signals and implement rotor fault diagnosis of UAV. The results derived from experimental signals demonstrate the superiority of the proposed method by comparison study. Additionally, it is found that the fault diagnosis accuracy of frequency features as input is much higher than that of time features and single frequency features as input.