摘要
针对影响三轴电液仿真转台动、静态性能最大的同步驱动、摩擦和大惯量负载干扰三个问题,采用了模拟人脑基于经验控制的FNN(模糊神经网络)控制器和基于学习校正的PNN(预测神经网络)控制器分别对应转台内环(角速度环)和外环(角度环)反馈系统。FNN同步控制器分为等同和主从同步控制模式,两种模式相互切换,提高了系统同步性能;PNN摩擦干扰控制器采用了基于双网络模型的NARMA(非线性自回归滑动平均)预测模型,具有较强的非线性系统辨识能力,提高了系统抗干扰能力。软件仿真结果表明,当转台外框负载发生变化或外框两马达转速相差较大时,使用PNN-FNN模型的智能控制系统仍具有较高的位置跟踪精度和动态性能。
Friction, disturbance and non-synchronization are main problems of three gimbals Rotary-Table. This paper adopt Fuzzy neural net（FNN） to deal with double hydraulic motors velocity feedback and Predictive neural net（PNN） for outer gimbal angle feedback, FNN controller simulate cerebra neural net（CNN） based on experiential control and PNN controller depend on self-study simulate CNN. FNN controller include master-slave and equation, two models switch for different state, improve system synchronization. PNN controller select nonlinear auto regressive moving average（NARMA） for system identification. Emulational results of PNN-FNN controllers show that the proposed approach can achieve high displacement tracking accuracy and dynamic performance when the loads of the simulator are changeable or two hydraulic motors＇ rotate speeds are different greatly.
出处
《宇航学报》
EI
CAS
CSCD
北大核心
2007年第1期178-182,共5页
Journal of Astronautics
