Self-powered real-time fault monitoring for drone blades

The vigorous development of the low - altitude economy is accompanied by a large number of small-drone being put into the market. The safety issues of drones have become increasingly prominent. In particular, the health condition of drone blades is directly related to the flight stability. Here, we introduce an innovative non-contact approach for monitoring drone blade faults utilizing acoustic signals. This methodology employs triboelectric sensor as acoustic-electric transducers, offering heightened sensitivity in acoustic signal acquisition. Through simulated fault experiments mirroring real flight scenarios, we constructed a triboelectric sensor based fault acoustic dataset. By integrating Convolutional Neural Network (CNN) with Transformer architecture, we effectively extracted intricate time-frequency features from the acoustic signals, achieving precise classification of four prevalent drone blade faults: crack, fracture, foreign object attachment, and edge deformation, with an accuracy rate of up to 95.1 %. Moreover, we developed a real-time monitoring system tailored for drone blade faults, showing its pivotal role in bolstering flight safety. This work constitutes a technological cornerstone for advancing the realm of intelligent, automated, and real-time drone blade fault monitoring.