基于神经网络的模型参考飞控系统容错控制

Neural Networks Based Model Reference Fault-tolerant Control for Flight Control Systems

针对飞控动态系统故障状态下安全飞行问题,提出了一种基于神经网络动态补偿的模型参考鲁棒容错控制方法。在经典RBF网络控制基础之上,设计了一种改进的神经网络结构,通过添加直接输入输出线性环节,应用最近邻聚类和新的自适应C-均值聚类法训练改进后的神经网络,以及在线调整神经网络的权值和阈值,提高了网络的收敛速度和泛化能力,达到了飞控系统在线实时快速容错控制和抗干扰的目的。同时,证明了该闭环鲁棒容错控制算法的稳定性。在波音747-100/200模型上仿真实验表明了该方法的有效性和可行性。

With respect to safety problem of flight dynamic control systems under faulty case,a technique of model reference robust fault-tolerant control using neural network compensation is proposed.In order to improve convergence rate of neural network as well as the performance of fault-tolerant control with disturbances,an improved neural network structure based on traditional RBF neural network is proposed,in which linear connections between input and output layers are introduced.The nearest neighbor-clustering algorithm and the adaptive C-means clustering algorithm are used to train the network,and the weights of neural network are adjusted on-line.The stability of the closed-loop system is rigorously proved.Simulation results on Boeing 747-100/200 model show that the presented scheme is effective.