CMAC神经网络在应力盘光学加工中的应用研究  

Study on applications of CMAC neural network in optical processing with stressed-lap

作　　者：范斌[1] 杨力[1] 袁家虎[1] 曾志革 李晓今[1] 

机构地区：[1]中国科学院光电技术研究所,四川成都610209

出　　处：《光电工程》2002年第6期9-11,16,共4页Opto-Electronic Engineering

摘　　要：计算机数控应力盘光学加工技术有其突出的优势,但应力盘形变却难以有效控制,是实现计算机数控应力盘光学加工中的难题。对此,本文分析了应力盘形变控制系统的特点,引入了CMAC神经网络实现应力盘形变控制的模型。提出了将CMAC神经网络应用于应力盘逆变形智能控制的创意和实现方法,以应力盘面形参数和对应的驱动器电压参数作为样本训练CMAC神经网络,将训练成功的CMAC神经网络作为控制器控制应力盘变形,取得了误差小于5%的计算机仿真结果。Computer controlled stressed-lap optical processing has its outstanding advantages in fabrication of large aspherical surface. But the difficult problem is how to effectively control stressed-lap with computer. For this reason, the properties of the control system for the stressed-lap are analyzed and a model for controlling the stressed-lap with CMAC neural network is introduced. The idea and performing method for applying CMAC neural network to the intelligent control for inverse deformation of the stressed-lap are proposed. With the stressed-lap surface shape parameters and the corresponding actuator voltage parameters as sample to train CMAC neural network, and using the successfully trained CMAC neural network as controller to control the distortion of the stressed-lap, the computer simulation result with error less than 5% is achieved.

关 键 词：CMAC 神经网络 非球面加工 应力盘 计算机管理控制 仿真 镜面 

分 类 号：TQ171.6[化学工程—玻璃工业]