基于高斯基函数CMAC神经网络的发电机故障诊断方法  

Fault diagnosis of a generator using a CMAC neural network based on Gauss basis functions

作　　者：万书亭[1] 何鹏[1] 赵松杰[2] 

机构地区：[1]华北电力大学机械工程系,保定071003 [2]邯郸职业技术学院机电系,邯郸056000

出　　处：《振动与冲击》2010年第4期84-87,134,共5页Journal of Vibration and Shock

基　　金：国家自然科学基金项目(50677017);中央高校基本科研业务费专项资金资助项目(09MG30)

摘　　要：提出一种基于高斯基函数小脑模型神经网络(CMAC)的汽轮发电机故障诊断新方法,为了达到更高的精度和更好的泛化能力,该方法以高斯函数作为CMAC神经网络的基函数,针对发电机的机电耦合特性,将发电机机电综合特征作为神经网络的训练样本输入,经MATLAB仿真得到了完全正确的诊断结果,收敛速度快,精度高,可以满足在线监控的要求。通过比较学习率和泛化常数取值不同时CMAC网络的训练结果,分析了学习率和泛化常数对该网络的影响。Based on a CMAC neural network with Gauss basis functions,a novel method was proposed for fault diagnosis of a turbo-generator.In order to achieve higher precision and better generalization ability,this method used Gauss basis functions in the CMAC neural network.Aiming at electrical and mechanical coupling characteristics of the generator,its integrated mechanical and electrical features as a neural network training sample were imputted.Through MATLAB simulation,the completely correct diagnosis results with higher convergence speed and accuracy,which met the requirements of on-line monitoring,were obtained.By comparing the CMAC network training results using the different values of the learning rate and generalization constant at the same time,the influence of the learning rate and generalization constant on the neural network was analyzed.

关 键 词：小脑模型神经网络(CMAC) 高斯基函数 发电机 故障诊断 机电综合特征 

分 类 号：TM31[电气工程—电机]