• CN:11-2187/TH
  • ISSN:0577-6686

机械工程学报 ›› 2019, Vol. 55 ›› Issue (23): 136-146.doi: 10.3901/JME.2019.23.136

• 机械动力学 • 上一篇    下一篇

扫码分享

基于IEWT和MCKD的滚动轴承故障诊断方法

李政1,2, 张炜2, 明安波1,2, 李峥1, 褚福磊1   

  1. 1. 清华大学机械工程系 北京 100084;
    2. 火箭军工程大学导弹工程学院 西安 710025
  • 收稿日期:2018-11-23 修回日期:2019-06-05 出版日期:2019-12-05 发布日期:2020-02-18
  • 通讯作者: 褚福磊(通信作者),男,1959年出生,博士,教授,博士研究生导师。主要研究方向为旋转机械动力学与振动控制、机械状态监测与故障诊断。E-mail:chufl@mail.tsinghua.edu.cn
  • 作者简介:李政,男,1992年出生,博士研究生,主要研究方向为故障诊断与信号处理。E-mail:l-z17@mails.tsinghua.edu.cn;张炜,男,1963年出生,博士,教授,博士研究生导师。主要研究方向为设备状态监测与故障诊断。E-mail:zhangweihuaiyu@163.com;明安波,男,1986年出生,博士,讲师。主要研究方向为机械故障诊断、信号处理。E-mail:mab10@mails.tsinghua.edu.cn;李峥,男,1984年出生,硕士,工程师。主要研究方向为机械故障诊断、信号处理。E-mail:497845611@qq.com
  • 基金资助:
    国家自然科学基金资助项目(51335006、51505486)。

A Novel Fault Diagnosis Method Based on Improved Empirical Wavelet Transform and Maximum Correlated Kurtosis Deconvolution for Rolling Element Bearing

LI Zheng1,2, ZHANG Wei2, MING Anbo1,2, LI Zheng1, CHU Fulei1   

  1. 1. Department of Mechanical Engineering, Tsinghua University, Beijing 100084;
    2. School of Missile and Engineering, Rocket Force University of Engineering, Xi'an 710025
  • Received:2018-11-23 Revised:2019-06-05 Online:2019-12-05 Published:2020-02-18

摘要: 针对经验小波变换(Empirical wavelet transform,EWT)对强噪声环境中滚动轴承微弱故障诊断的不足,主要是傅里叶频谱分段不当的问题。提出一种基于最大相关峭度解卷积(Maximum correlated kurtosis deconvolution,MCKD)降噪与改进EWT相结合的滚动轴承早期故障识别方法。首先采用最大相关峭度解卷积算法以包络谱的相关峭度最大化为目标对原信号进行降噪处理、检测信号中的周期性冲击成分,然后根据信号Fourier频谱的包络极大值进行分段,通过分析各频段平方包络谱中明显的频率成分来诊断故障。新方法能有效降噪、增强信号中周期性冲击特征、降低单次偶然冲击的影响、抑制非冲击成分。通过对含外圈、内圈故障的滚动轴承进行试验分析,结果表明,相比于快速谱峭度图和小波包络分析方法,该方法提取出的特征更加明显,能有效实现滚动轴承早期微弱故障的识别。

关键词: 经验小波变换, 快速谱峭度图, 最大相关峭度解卷积, 小波包络分析, 滚动轴承

Abstract: In order to solve the problem of Empirical wavelet transform method for the rolling element bearing fault diagnosis in strong noise condition, that is mainly the inappropriate segmentation of the signal spectrum, the combination of the Improved Empirical wavelet transform and Maximum correlated kurtosis deconvolution method is proposed. Firstly, an original signal is de-noised with MCKD algorithm, and the max kurtosis of its envelope spectrum is taken as an objective to detect its periodic impact components. Then, the envelope of the signal Fourier spectrum is segmented based on the peaks, and the most meaningful component can be found from the signal components. At last, fault features can be diagnosed by analyzing obvious frequency components in squared envelope spectrum. The new method can de-noise the signal and enhance the periodic impact components feature. It is shown that the lesser powerful fault induced by Single accidental impact and nonimpact components is restrained in de-noised signal. The effectiveness of the proposed method has been validated by both simulated and experimental bearing vibration signals. It's shown that fault character extracted by the proposed method is more clearly and believable than the fast kurtogram algorithm and wavelet envelope analysis.

Key words: empirical wavelet transform, fast kurtogram algorithm, maximum correlated kurtosis deconvolution, wavelet envelope analysis, rolling bearing

中图分类号: