基于径向基函数网络的光电编码器误差补偿法.docVIP

第16卷　第4期 2008年4月　　　　　　　　　　　光学精密工程 　Optics and Precision Engineering 　　　　　　　　 Vol.16　No.4 　　Apr.2008 　　收稿日期:2007209213;修订日期:2007212210. 　　基金项目:国家自然科学基金资助项目(No文章编号 (20080420598207 基于径向基函数网络的光电编码器误差补偿法 洪　喜1,2,续志军1,杨　宁1 (1.中国科学院长春光学精密机械与物理研究所,吉林长春130033; 2.中国科学院北京研究生院,北京100039 摘要:提出了一种新的编码器误差修正技术。建立了径向基函数网络模型,以高精度检测仪器的检测值为学习目标,以生成最小映射误差为原则调节网络权因子、径向基函数中心和宽度,使建立的网络具有良好的学习能力和泛化能力。利用DSP 的在线烧写技术保存网络建模参数,从而无需拆改编码器系统即可实现误差的补偿与修正,并且可根据使用需要进行程序更新。仿真和实验表明,采用此种方法可将编码器的精度提高至原来的1~3倍,有效地改善了编码器的系统精度,并且很好地解决了非线性误差对系统的影响。 关　键　词:光电编码器;神经网络;径向基函数;误差补偿;数字信号处理器中图分类号:TN762;TP183　　文献标识码:A E rror compensation of optical encoder based on RB F net w ork HON G Xi 1,2,XU ZHi 2jun 1,YAN G Ning 1 (1.Changchun I nstit ute of O ptics ,Fi ne M echanics and Physics , Chi nese A ca dem y of S ciences ,Changchun 130033,Chi na; 2.Grad uate S chool of t he Chi nese A cadem y of S ciences ,B ei j i ng 100039,Chi na Abstract :A new met hod to correct and compensate t he dynamic error of a optical encoder was presen 2ted by using t he neural network and t he digital signal p rocess technology.A modeling met hod based on t he Radial Basis Function (RB F was set up ,in which t he outp ut was t he test value of high preci 2sion inst rument and t he inp ut was t he angle values of sample point s.According to t he inhibiting con 2dition between t he test value and t he outp ut of network ,t he power factor formula ,t he center and widt h of t he RBF were adjusted to make t he model have a good learning ability and a generalization a 2bility.The relationship between sampled angles and errors was determined by t raining t he neural net 2work.After t he sampled angles were measured ,t he unknown error of t he encoder could be calculated via a trained neural network even if t he error was nonlinear.The tested result s show t hat t he precision of system error by t his met hod is 1~3times t hat by t radiotional met hod.And not only t he error of each network point but also t he interpolated network