中国机械工程

• 机械基础工程 • 上一篇    下一篇

基于总体平均经验模式分解近似熵和混合PSO-BP算法的轴承故障诊断方法

张淑清;黄文静;胡永涛;宿新爽;陆超;姜万录   

  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
  • 出版日期:2016-11-25 发布日期:2016-11-23
  • 基金资助:
    国家自然科学基金资助项目(51475405,61077071);河北省自然科学基金资助项目(F2015203413);河北省高等学校科学技术研究重点项目(ZD2014100) 

Bearing Fault Diagnosis Method Based on EEMD Approximate Entropy and Hybrid PSO-BP Algorithm

Zhang Shuqing;Huang Wenjing;Hu Yongtao;Su Xinshuang;Lu Chao;Jiang Wanlu   

  1. Measurement Technology and Instrumentation Key Lab of Hebei Province,Yanshan University,Qinhuangdao,Hebei,066004
  • Online:2016-11-25 Published:2016-11-23
  • Supported by:
     

摘要: 针对机械系统的非平稳、非线性特性,提出了一种基于总体平均经验模式分解(EEMD)近似熵和混合PSO-BP算法的轴承故障诊断方法。EEMD能够解决EMD的端点效应,改善处理非线性信号时的局限性;引入随机权重和压缩因子来改进粒子群算法,优化BP神经网络的权值和阈值,解决BP网络的全局收敛问题。将信号经EEMD得到的IMF分量与近似熵结合,组成特征向量,再将构造的特征向量输入到PSO-BP神经网络中进行模式识别。实验及工程应用实例证明了该方法的有效性和优越性。

关键词: 轴承, 故障诊断, 总体平均经验模式分解, 近似熵, 混合粒子群神经网络

Abstract: According to the non-stationary and nonlinear characteristics of mechanical faults, a new method for bearing fault diagnosis was put forward based on EEMD approximate entropy and hybrid PSO-BP algorithm. EEMD might resolve the end effects of the EMD, improving the limitations of EMD when dealing with nonlinear signals; the random weights and the compression factors were introduced to improve particle swarm optimization(PSO) algorithm so as to optimize BP neural network, realizing fast convergence to optimal solution effectively of neural network. Signals were first decomposed by EEMD to get the instrinsic mode function(IMF) components, and to construct feature vectors together with the approximate entropy and the constructed feature vectors were put into the PSO-BP neural network for pattern recognition. The experiments and the engineering tests demonstrate the efficiency and superiority of this method.

Key words: bearing, fault diagnosis, ensemble empirical mode decomposition (MEMD), approximate entropy, hybrid particle swarm neural network

中图分类号: