摘要
研究飞行器空间动态特性优化控制问题,高动态临近空间飞行器的姿态运动模型是一个多变量非线性时变不确定系统,由于小扰动线性化理论和系数冻结基本假设的传统三通道独立综合(设计)方法已很难适用。针对多变量非线性时变不确定系统对象的非线性设计方法来解决控制问题。提出了一种神经网络自适应反馈线性化方法,用FCMAC神经网络估计实际模型和标称模型间精确线化反馈阵的误差,修正实际系统的反馈线性化模型,有效克服了反馈线性化需要被控对象精确建模的局限性,实现了高马赫、大空域、快角度机动条件下的姿态解耦控制,并进行仿真。仿真结果表明:解耦控制系统在气动拉偏30%下超调小于7%,通道间交叉耦合小于1°,能够满足对象的控制要求。
The motion model of high - dynamic near space vehicles is a multivariable nonlinear time - variant un- certain system. The traditional three - channel independent design method is no longer useful because it is based on little disturbance linearization theory and fundamental hypothesis of coefficient freezing method. Instead of that, a nonlinear design method is necessary which is fit for multivariable nonlinear time - variant uncertain plant. Based on differential geometry method and modem control theory, this paper suggested an approach of neural networks adaptive feedback linearization. The errors of precision linearization feedback matrix between actual model and nominal model were estimated by FCMAC neural network and the feedback linearization model was modified. It effectively solved the limitation of feedback linearization that needs accurate mathematic model. The difficult problems of attitude decou- piing control on high -mach, large airspace and large angle mobile condition were Solved. The results of simulation on the condition of 30% aerodynamic interference show that the overshoot of decoupling control system is less than 7%, residual degree of channel cross decoupling is less than 1°, and the design can meet the control requirement for this kind of flying vehicles.
出处
《计算机仿真》
CSCD
北大核心
2010年第4期57-61,共5页
Computer Simulation
关键词
高动态
临近空间飞行器
神经网络
反馈线性化
解耦控制
High - dynamic
Near - space flying vehicles
Neural networks
Feedback linearization
Decoupling control