摘要
研究了BP网络输入数据的归一化方法,提出了一种联合归一化的新方法,从而加快了网络的学习训练速度,提高了分类精度。在此基础上,建立了用于机械故障诊断的三层BP神经网络模型,编写了基于BP神经网络的故障诊断软件,并在实际的齿轮箱状态监测与故障诊断研究中得到了工程应用。
In this paper,the methods of input data normalization in BP neural network are studied,and a new method of combined normalization is presented.This method can make input vector of BP neural network more suitable for own structure,so it can quicken up the learning of network,and improve the classification precision.Based on this method,a three-layer BP neural network model for machine fault diagnosis is built,and the software of BP network for fault diagnosis is designed,which is used for state monitoring and fault diagnosis for gearbox.
出处
《机械工程与自动化》
2010年第3期122-123,126,共3页
Mechanical Engineering & Automation
关键词
BP神经网络
归一化
故障诊断
BP neural network
normalization
fault diagnosis
