UAV fault detection based on GA-BP neural network

Abstract

Unmanned Aerial Vehicle has a high complexity and nonlinear control system, the sensor is the important device to provide flight data for system, so it's performance is the major determinant of Unmanned Aerial Vehicle. The sensor generally damaged suddenly and lack of many fault data samples. So specific to this reason, this paper uses wavelet packet to extract fault energy feature, and designs a fault detection system, which is the BP neural network optimized by genetic algorithm. The algorithm not only overcomes the shortcoming of BP neural network learning slow convergence and easily falling into the local minimum, but also improves the system performance and fault recognition. The simulation studies shows that the system, which has a good application prospect, can efficiently detect Unmanned Aerial Vehicle sensor fault.