神经网络动态逆在歼击机安全着陆中的控制  

Safe landing control of fighters based on neural network dynamic inversion

作　　者：黄小波[1] 胡寿松[1] 

机构地区：[1]南京航空航天大学自动化学院,南京210016

出　　处：《电光与控制》2007年第3期5-7,共3页Electronics Optics & Control

基　　金：国家自然科学基金重点资助项目(60234010);航空科学基金资助项目(05E52031);国防基础科研资助项目(K1603060318)

摘　　要：给出了基于神经网络动态逆的自适应跟踪控制方法,用以解决飞机着陆过程中的复杂非线性和出现舵机故障的情况。应用神经网络直接对非线性系统故障模型求逆,使得所设计的逆系统能够包含故障信息,克服了传统的控制设计中将过程模型线性化,从而将不可忽视的非线性关系用线性关系代替或忽略的弊端。对由于建模误差、不确定性因素等引起的非线性系统逆误差,通过自组织模糊小脑模型关节控制器(SOFCMAC)神经网络在线进行修正。并在此基础上对3个通道分别设计了参考模型和线性控制器,以实现对伪线性系统进行跟踪控制。通过将这种方法用于某型歼击机在着陆过程中发生平尾卡死故障控制的过程仿真,验证了该方法的可行性。A plan of self-adaptive tracking control is introduced based on Neural Network Dynamic Inversion （NNDI） for dealing with the complex nonlinear issues and uncertainties of aircraft system caused by actuator stuck during landing. The designed pseudo -linear system can contain fault information by using neural network to get dynamic inversion model of nonlinear systems with failures. Thus the drawback of the traditional control design of linearizing the process model, i.e. , using linear relationship to replace the non - neglectable nonlinear relationship, is overcome. Self-Organizing Fuzzy Cerebella Model Articulation Controller （SOFCMAC） neural network is used to correct the nonlinear system inversion error due to modeling uncertainties and disturbances. We designed reference models and linear controllers for all the three channels respectively for pseudo -linear system tracking control. The application of the method in auto -landing system of fighter with elevator stuck is studied by the simulation. Results show that the method is effective and practicable.

关 键 词：自动着陆 故障 动态逆 神经网络 自组织模糊小脑模型关节控制器 

分 类 号：V324[航空宇航科学与技术—人机与环境工程]