中国机械工程

• 机械基础工程 • 上一篇    下一篇

基于近似熵和集成经验模态分解的转子多故障诊断方法研究

韩中合;徐搏超;朱霄珣;焦宏超   

  1. 华北电力大学,保定,071003
  • 出版日期:2016-08-25 发布日期:2016-08-17
  • 基金资助:
    中央高校基本科研业务费专项资金资助项目(2015MS102);国家自然科学基金资助项目(50676031)

Research on Multi-fault Diagnosis of Rotor Based on Approximate Entropy and EEMD

Han Zhonghe;Xu Bochao;Zhu Xiaoxun;Jiao Hongchao   

  1. North China Electric Power University,Baoding,Hebei,071003
  • Online:2016-08-25 Published:2016-08-17
  • Supported by:

摘要: 为了提高汽轮机转子多故障分类的准确率,提出一种集成经验模态分解(EEMD)、近似熵和支持向量机相结合的多状态分类方法。首先进行EEMD得到各频段的单分量信号;再求出熵值作为故障信号的特征向量输入到基于二叉树的支持向量机中实现多状态分类。对比近似熵、模糊熵和能量法这三种方法,实验结果验证了利用EEMD和熵理论相结合的方法量化故障信号非线性特征的正确性。同时也表明在欧氏空间中,近似熵值组成的特征向量彼此间的距离最远,分类效果也最好。

关键词: 集成经验模态分解, 近似熵, 支持向量机, 多故障诊断

Abstract: For the purpose of accurate identification of the turbine rotor multi-fault states, a diagnosis method was put forward based on the EEMD, approximate entropy and SVM. Firstly, the fault signals were decomposed to a number of intrinsic mode functions (IMFs) by EEMD method, then calculating entropy of IMFs as the feature vector to construct samples for binary tree SVM for multi-state classification. This paper compared with approximate entropy, fuzzy entropy and energy method. The experimental results verify the correctness of quantifying the nonlinear characteristics of fault signals with EEMD and entropy theory. They also indicate that the feature vectors based on approximate entropy are farthest from each other in European space, and the classification accuracy is the highest.

Key words: ensemble empirical mode decomposition(EEMD), approximate entropy, support vector machine (SVM), multi-fault diagnosis

中图分类号: