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