Abstract：With the increase of the number of UAVs, the possibility of fault occurrence and the difficulty of fault diagnosis increase. In this paper, taking the UAV flight control system actuator as the research object, considering the nonlinearity and external disturbance of the flight process, a neural network observer fault diagnosis method for the actuator is proposed, in which the weights and center values of the neural network can be updated online to avoid the difficulty of parameter selection. At the same time, the nonlinear term of the system is effectively dealt with. The relative output error is introduced into the UAV formation by using the long-wingman formation strategy, and the relative output error is described by undirected topological structure diagram, so that the designed neural network observer can reflect the communication connection between UAVs. The Kronecker product is used to represent the global vector of UAV formation system, and the stability conditions of the designed neural network observer are derived from the global point of view by using Lyapunov stability theory. The simulation results show that the designed neural network observer can diagnose the constant and time-varying faults occurring at the same time, different times, same channel and different channels.
聂瑞, 王红茹. 基于神经网络观测器的无人机编队执行器故障诊断[J]. 空天防御, 2022, 5(2): 32-41.
NIE Rui, WANG Hongru. Fault Diagnosis of UAV Formation Actuator Based on Neural Network Observer. Air & Space Defense, 2022, 5(2): 32-41.
