适应运载火箭推力下降故障的神经网络容错控制方法

A Neural Network Fault-Tolerant Control Method for Launch Vehicles with Thrust Decline

针对运载火箭单台发动机推力下降故障,提出了一种基于径向基神经网络(Radial Basis Function Neural Network,RBFNN)的容错姿态控制方法。该方法无需故障诊断系统,根据运载火箭姿态动力学控制通用模型,使用RBFNN在线辨识并补偿模型的故障变化和不确定干扰,得出容错控制律。仿真结果表明,在单台发动机发生推力下降故障时,本文方法与传统PD方法、自适应增广控制方法(Adaptive Augment Control,AAC)相比,可有效保证姿态稳定和控制精度。

To tolerate thrust decline of launch vehicles’single engine,a fault-tolerant attitude control method based on radial basis function neural network(RBFNN)is proposed.The method does not require a fault diagnosis system.Based on the general model of launch vehicle attitude dynamics,RBFNN is used to online identify and compensate the fault change and uncertain disturbances in the model to derive the fault-tolerant control law.The simulation results show that this approach can ensure the stability as well as accuracy by comparing with the traditional PD method and adaptive augment control(AAC)method.