基于LM神经网络的旋转机械故障诊断  

Fault Diagnosis for Rotating Machinery Based on LM Neural Network

作　　者：王光研[1] 许宝杰[1] 

机构地区：[1]北京机械工业学院机电学院,北京100085

出　　处：《机械制造与自动化》2007年第6期41-43,共3页Machine Building & Automation

基　　金：国家自然科学基金资助项目(50375017);北京市自然科学基金资助项目(3062008);北京市属市管高校人才强教计划资助项目;机电系统测控北京市重点实验室开放课题资助项目(KF20061123201;KF20061123202)

摘　　要：BP神经网络具有优良的非线性映射能力,可以很好地描述频率特征和诊断结果之间的关系。针对BP神经网络存在局部极小值和收敛速度慢等问题,提出了一种基于Levenberg-Marquardt(LM)的改进的BP网络。经改进算法训练的网络能大大提高诊断的能力,具有广泛的应用前景和应用价值。BP Neural network is effective on dealing with non-linear mapping which could perfectly describle the non-linear relations between frequency character and diagnosis results. A kind of BP neural network based on LM Backpropagation optimization algorithm is introuduced in detail, which can comeover the disadvantages of standard BP algorithm. The improved algorithm converges very rapidly and has good precision compared BP algorithm. The improved BP neural network is suitable for diagnosis of mechanical fault,and has higher reference value in practical application.

关 键 词：神经网络 故障诊断 旋转机械 MATLAB 

分 类 号：TP206[自动化与计算机技术—检测技术与自动化装置]