中国机械工程

• 机械基础工程 • 上一篇    下一篇

基于FSWT细化时频谱SVD降噪的冲击特征分离方法

何志坚1,2;周志雄1   

  1. 1.湖南大学,长沙,410082
    2.湖南信息职业技术学院,长沙,410200
  • 出版日期:2016-05-10 发布日期:2016-05-05
  • 基金资助:
    国家科技重大专项(2012ZX04003041);国家自然科学基金资助项目(51475158)

Impact Feature Separation Method Based on FSWT Zoom Time-frequency Spectrum De-noised by SVD

He Zhijian1,2;Zhou Zhixiong1   

  1. 1.Hunan University,Changsha,410082
    2.Hunan College of Information,Changsha,410200
  • Online:2016-05-10 Published:2016-05-05
  • Supported by:

摘要: 为有效提取滚动轴承故障振动信号的故障冲击特征,提出了基于FSWT细化时频谱SVD降噪的冲击特征分离提取方法。首先对原始信号进行频率切片小波变换得到全频带下的时频分布,然后根据时频谱能量分布特点选择出感兴趣的时频区域,再以较高的时频分辨率对感兴趣的时频区域进行细化分析得到细化的时频谱,从而分割出含有故障特征时频区域。为克服噪声对细化时频谱精度的影响,FSWT细化分析过程融入SVD降噪,通过对FSWT细化时频谱系数矩阵进行奇异值差分谱阈值降噪,使得FSWT细化时频谱的冲击特征更加明显,最后通对降噪后的细化时频谱进行FSWT逆变换重构,分离出故障冲击信号。仿真分析和故障诊断实例表明,基于FSWT细化时频谱SVD降噪的冲击特征分离提取方法能够成功从低信噪比信号中提取出周期性的冲击特征,有效地实现对滚动轴承各种故障的诊断。

关键词: 频率切片小波变换, 奇异值分解, 滚动轴承, 故障诊断

Abstract: In order to extract the impact features of rolling bearings effectively,an impact feature separation method was proposed based on FSWT zoom time-frequency spectrum de-noised by SVD. Firstly, the original signals were analyzed by FSWT to get their whole time-frequency distributions. Then the interesting time-frequency region could be selected according to the time-frequency energy distribution characteristics, furthermore the fault characteristics region could be separated by a zoom analysis to the interested time-frequency region with higher resolutions. So as to inhibit the influences of the noise on the accuracy of zoom time-frequency distribution maps, FSWT zoom analysis was integrated with the SVD de-noising process, the FSWT zoom analysis time-frequency distribution matrix was de-noised by the SVD singular value difference spectrum threshold de-noising method to make the impact features outstanding. Finally, the fault impulse signals were separated by applying the FSWT inverse transform to the de-noised zoom analysis time-frequency spectrum. The simulated analysis and actual fault diagnosis example results demonstrate the impact feature separation method based on FSWT zoom time-frequency de-noised by SVD may extract the periodic impact features from low SNR signals and accomplish the fault diagnosis of rolling bearings.

Key words: frequency slice wavelet transform (FSWT), singular value decomposition(SVD), rolling bearing, fault diagnosis

中图分类号: