Quadratic Neural Network-Based Fault Diagnosis for UAV

The unmanned aerial vehicle (UAV) fault diagnosis is of crucial significance for ensuring the reliability, stability and safety of UAVs in modern military and industry. In this paper, we propose a data-driven approach to introduce a Residual Quadratic Neural Network (ResQNet) fault diagnosis network. This network improves the performance of fault diagnosis by extracting the temporal features of flight data. Through a predetermined fault injection method, we obtain a dataset of UAV faults which includes the normal state, three types of actuator faults, and four types of sensor faults. The results show that the ResQNet realizes end-to-end time-domain UAV fault diagnosis and fault localization.