基于神经网络的负载模拟器控制方法研究

【作者】 吴建德；

【导师】 袁朝辉；

【作者基本信息】 西北工业大学 ， 机械电子工程， 2004， 硕士

【摘要】 负载模拟器是半实物仿真实验室中模拟无人机等飞行器在飞行中舵面所受空气动力载荷的重要设备。它为飞行器提供性能试验数据，同时也为新技术的应用提供可靠的实验资料。随着国防工业的发展及新型武器系统研制的需要，对此类伺服系统的需求越来越多，并且对其性能指标要求也越来越高。论文的目的就是提出一种更有效的提高控制系统性能的方法。 衡量负载模拟器系统性能的关键指标是多余力矩的抑制。论文从介绍以往消除多余力矩的方法开始，分析了多余力矩的产生原理以及影响多余力矩的因素。针对多余力矩严重影响施力系统动态加载性能的特点，依据神经网络的非线性逼近和自学习特性，提出一种基于神经网络的复合控制方法宋提高系统性能。文中给出了具体的控制结构和算法。通过仿真可以看出，复合控制器利用神经网络进行在线辨识、控制，基本消除多余力矩，系统动态性能得到改善，仿真效果令人满意。在某型无人机负载模拟器中的应用表明，该方法极大的改善了系统动态加载性能，有很强的鲁棒性。更多还原

【Abstract】 Load simulator was an important simulation system in hardware-in-loop flight simulation laboratory, which could simulate aerodynamic loads acting on flight vehicle, such as unmanned aerial vehicle, planes. It was one of most important devices in the whole flight vehicle simulation system. It afforded datum, which were contributed to the use of new flight vehicle. With the development of national defence industry, load simulator was becoming more and more important. In this thesis, a method that could improve the system performances of the load simulator has been given.How to eliminate the surplus torque of a loading system was one of the key problems to design a load simulator. First, in the thesis some effective control methods were introduced to eliminate the surplus torque. Then the principle of surplus torque was analyzed in detail, and some factors that influenced the surplus torque were given. Based on the learning characteristic of neural network and the function approximation ability of the neural networks, a hybrid control based on neural networks was proposed. The application results of the unmanned aerial vehicle load simulator showed that the proposed controller could eliminate the surplus torque effectively and improve the dynamic loading performances of the load simulator fairly. In addition, the results showed that the proposed controller was of fine robustness to unknown external load disturbances.更多还原