基于神经网络的模型参考自修复飞行控制理论研究及仿真软件开发

【作者】 王鹏；

【导师】 艾剑良；

【作者基本信息】 西北工业大学 ， 飞行器设计， 2005， 硕士

【摘要】 飞行控制系统的重构技术可以有效地提高飞行器的生存能力和任务效能,已成为现代飞行器综合设计领域的热门研究专题之一。首先介绍了飞控系统重构技术的发展历史及现状,对基于神经网络的模型参考自修复飞行控制方法进行了概述和分析。在此基础上提出了一种改进的基于模糊神经网络的模型参考自修复飞行控制方法,并对所使用的BP网络学习算法进行了分析改进,进行了仿真软件的开发。对比非故障和故障状态下的飞行仿真结果表明,改进后的自修复飞行控制方法可以有效地抑制神经网络的“过学习”现象,减小了对神经网络辨识器精度的依赖程度,在故障条件下的补偿作用明显,达到了自修复飞行控制的目的,有着良好的发展潜力。更多还原

【Abstract】 The technique of reconfigurable flight control system can obviously enhance the survival ability and mission effectiveness of the aerocraft, and it has become one of the attractive research subjects in modern aerocraft integration design area. Firstly, this paper reviews the history and current state about the development of the reconfigurable flight control technique, then presents and analyses the model reference self-repairing flight control technique based on fuzzy neural network. Based on it, the paper presents an improved model reference self-repairing flight control method based on fuzzy neural network, moreover analyzes and improves the BP network learning arithmetic, and develops the simulated software. The comparison of the flight simulation in the failure and unfailure condition indicates that the improved self-repairing flight control method can effectively reduce the "over-learning" phenomenon of neural network and the dependence on the precision of neural network identifying system. And the compensation effect in the failure condition is notable, so it achieves the aim of self-repairing flight control, and holds a pleasant outlook.更多还原