改进小脑模型网络对轧辊偏心谐波的分频辨识  

A Identification Method of Separated Frequency for Roller Eccentricity Harmonic by Improved Network of Cerebellar Model Articulation Controller

作　　者：侯媛彬[1] 杜京义[1] 高赟[1] 

机构地区：[1]西安科技大学,西安710054

出　　处：《中国机械工程》2007年第8期938-941,共4页China Mechanical Engineering

基　　金：陕西省科技基金资助重点项目(DK04JC12)

摘　　要：针对连续轧钢机轧辊偏心谐波信号具有严重的非线性特性,提出一种采用改进的小脑模型控制器(ICMAC)神经网络对轧辊偏心进行分频辨识的方法。该方法基于改进的Prony参量法对轧辊偏心信号进行估计,利用ICMAC对非线性的逼近能力,对不同频率不同幅值的轧辊偏心谐波进行分频辨识,然后提取连轧机轧辊偏心信号非线性谐波,从而得到混合了各次谐波的轧辊偏心信号的最简模型。仿真结果表明,该方法与常规的BP网络辨识建模方法相比,不仅辨识结果的置信度高,而且能明显识别出对控制精度影响较大的谐波,可为进一步消除连轧机轧辊偏心信号提供依据。Aiming at nonlinear trait of the roller eccentricity harmonic of the continuous rolling steel machine, an identification method of separated frequency region for the harmonic was presented using improved network of cerebellar model articulation controller （ICMAC）. It was based on improved Prony parameter estimation in roller eccentricity, using learning nonlinear trait of ICMAC. The different harmonics of the separated frequency and the breadth value were identified; whole harmonic waves of the roller eccentricity were extracted; then the simplest model in roller eccentricity was established. The method was proved by simulation, and compared with normal BP network under same conditions, the identification reliability of this method is higher, and can obviously distinguish the main frequency from the harmonics, where the main frequency affects bigger to the control accuracy,further the method of eliminating harmonic is provided in roller eccentricity.

关 键 词：CMAC网络 轧辊偏心谐波 分频辨识 控制精度 

分 类 号：TG33[金属学及工艺—金属压力加工] TP18[自动化与计算机技术—控制理论与控制工程]