中国机械工程 ›› 2016, Vol. 27 ›› Issue (04): 433-437.

• 机械基础工程 • 上一篇    下一篇

基于LMD多尺度熵和概率神经网络的滚动轴承故障诊断方法

孟宗1,2;胡猛;谷伟明;赵东方   

  1. 1.燕山大学河北省测试计量技术及仪器重点实验室,秦皇岛,066004
    2.国家冷轧板带装备及工艺工程技术研究中心,秦皇岛,066004
  • 出版日期:2016-02-25 发布日期:2016-03-03
  • 基金资助:
    国家自然科学基金资助项目(51575472);河北省自然科学基金资助项目 (E2015203356);河北省高等学校科学研究计划资助重点项目(ZD2015049);河北省留学人员科技活动择优资助项目(C2015005020) 

Rolling Bearing Fault Diagnosis Method Based on LMD Multi-scale Entropy and Probabilistic Neural Network

Meng Zong1,2;Hu Meng;Gu Weiming;Zhao Dongfang   

  1. 1.Key Laboratory of Measurement Technology and Instrumentation of Hebei Province,Yanshan University,Qinhuangdao,Hebei,066004
    2.National Engineering Research Center for Equipment and Technology of Cold Rolling Strip,Qinhuangdao,Hebei,066004
  • Online:2016-02-25 Published:2016-03-03
  • Supported by:

摘要:

研究了一种基于LMD多尺度熵和概率神经网络的滚动轴承故障诊断方法。该方法将故障信号自适应地分解为若干乘积函数分量,然后将各分量的多尺度熵作为故障特征向量输入概率神经网络进行模式识别,实现了对损伤位置和损伤程度的诊断。将该方法与基于LMD时域统计量和神经网络的滚动轴承故障诊断方法进行了对比。实验结果表明,基于LMD多尺度熵和概率神经网络的方法能对滚动轴承故障进行有效的识别与诊断。

关键词: 局部均值分解, 故障特征提取, 多尺度熵, 概率神经网络, 故障诊断

Abstract:

A rolling bearing fault diagnosis method was studied based on LMD multi-scale entropy and probabilistic neural network. In this method, the fault signal was decomposed into several product functions adaptively, and then the multi-scale entropies of each component were feed into the probabilistic neural network as fault feature vectors for pattern recognition to realize the diagnosis of damage position and damage degree. Comparing with the method based on LMD time-domain statistics and probabilistic neural network, the experimental results show that the method based on LMD multi-scale entropy and neural network can identify and diagnose the rolling bearing fault accurately and efficiently.

Key words:  local mean decomposition(LMD), fault feature extraction, multi-scale entropy, probabilistic neural network, fault diagnosis

中图分类号: