基于EEMD-GRNN网络的滚动轴承故障诊断试验研究  

Fault diagnosis tests research of rolling bearing based on EEMD- GRNN network

作　　者：李善[1] 谭继文[1] 俞昆 

机构地区：[1]青岛理工大学,山东青岛266520

出　　处：《制造技术与机床》2016年第3期55-60,共6页Manufacturing Technology & Machine Tool

基　　金：国家自然科学基金项目(51075220);山东省高等学校科技计划项目(J13LB11);高等学校博士学科点专项科研基金(20123721110001);青岛市科技计划基础研究项目(12-1-4-4-(3)-JCH)

摘　　要：提出了一种基于总体平均经验模态分解和GRNN神经网络的滚动轴承故障诊断方法。首先通过EEMD方法将非平稳、非线性的滚动轴承振动信号分解为若干个平稳的固有模态函数(IMF)之和,提取前8个IMF分量作为频域特征,同其他14个时频域特征指标组成特征集输入到GRNN神经网络中,建立起GRNN网络模型,对滚动轴承三种故障状态进行模式识别。通过分析比较BP和GRNN两种网络模型对故障的诊断结果,验证了GRNN网络的优越性和可行性。Propose a fault diagnosis methods for rolling bearing based on an ensemble empirical mode decomposition and GRNN neural network. First of all,through the EEMD method,non- stationary and nonlinear properties of rolling bearing vibration signal is decomposed into several stationary intrinsic mode function（ IMF）,the sum of first eight IMF component as a frequency domain feature extracting,with 14 other time- frequency domain characteristic index of characteristic collection input into the general regression neural network（ GRNN）,establish the GRNN network model,for the three kinds of rolling bearings fault state for pattern recognition. Through analyzing and comparing the BP and GRNN two network model for fault diagnosis,verified the superiority and feasibility of GRNN network.

关 键 词：滚动轴承 总体平均经验模态分解(EEMD) IMF分量 GRNN神经网络 故障诊断 

分 类 号：TH17[机械工程—机械制造及自动化]