复合正交神经网络与CMAC在PID并行控制中的比较研究  

Research on the Comparison between CONN and CMAC in PID Parallel Control

作　　者：叶军[1] 

机构地区：[1]绍兴文理学院机电系,绍兴312000

出　　处：《机床与液压》2005年第3期153-154,共2页Machine Tool & Hydraulics

基　　金：浙江省自然科学基金资助项目(M603070)

摘　　要：由于正交神经网络算法简单,学习收敛速度快,具有线性、非线性逼近精度高等优异特性,取得了较好的应用 效果,但在动态建模与实时控制问题上研究较少。为此,本文提出复合正交神经网络(CONN)与PID并行控制算法,并 对小脑模型(CMAC)与PID并行控制作一比较研究。仿真结果表明,当方波(阶跃)输入与正弦输入时CONN实现的前 馈控制效果比小脑模型实现的前馈控制效果更好,响应速度更快,这充分地体现了复合正交神经网络的特点,即输出误差 小、实时性好、鲁棒性强。In order to research the dynamic system model-building and real-time control of compound orthogonal neural network (CONN),the parallel control algorithm between CONN and PID was presented,and was compared with the parallel control between CMAC and PID.The simulation results prove that the former is better efficiency than the latter in fast respondent velocity when signals of square wave and sine wave are inputted in the feedforward control.It is showed that CONN has the characteristics of smaller error,better real-time control,and stronger robust.

关 键 词：复合正交神经网络(CONN) CMAC PID 并行控制 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]