基于DRNN网络的航空发动机多变量解耦控制  

A multivariable decoupling control based on DRNN for aeroengine

作　　者：朱玉斌[1] 樊思齐[1] 任新宇[1] 时瑞军[1] 

机构地区：[1]西北工业大学动力与能源学院,陕西西安710072

出　　处：《推进技术》2006年第2期150-153,共4页Journal of Propulsion Technology

摘　　要：通过分析对角递归神经网络(DRNN)及带动量项的梯度学习方法(GDM),针对某型涡扇发动机的性能控制,研究了基于对角递归神经网络的多变量自学习解耦控制算法及其在航空发动机控制中的应用。阐明了该方法的结构和原理。并在设计点处进行了发动机多变量解耦控制系统设计。在偏离设计点时,大量的仿真结果表明,系统具有较好解耦和自适应能力。A new neural paradigm called diagonal recurrent neural network （DRNN） and Gradient Decent Method （GDM） were presented. A muhivariable decoupling control algorithm based on the diagonal recurrent neural network was used for aeroengine control. A generalized GDM was developed and used to train diagonal recurrent neuroidentifier. The emphasis was focused on the research of the algorithm and the properties of the controller, as well as their application to aeroengine control by computer simulation. Finally, the aeroengine control system based on DRNN was designed. Simulation shows that the system has fine performance of decoupling and adaptive capabilities.

关 键 词：航空发动机 对角递归神经网络^+ 多变量控制 解耦^+ 

分 类 号：V233.7[航空宇航科学与技术—航空宇航推进理论与工程]