China Mechanical Engineering

Previous Articles     Next Articles

Fault Feature Extraction Method for Gears Based on ISSD and SVD

TANG Guiji;LI Nannan;WANG Xiaolong   

  1. Department of Mechanical Engineering,North China Electric Power University,Baoding,Hebei,071003
  • Online:2020-12-25 Published:2020-12-28



  1. 华北电力大学机械工程系,保定,071003
  • 基金资助:

Abstract: Aiming at the problems that the gear fault features were weak and difficult to extract effectively under strong background noises, a fault feature extraction method for gears was proposed based on ISSD and SVD.  Considering the defects that the modal parameters needed to be selected by experience in the SSD algorithm, the SSD algorithm was improved based on dispersion entropy optimization algorithm. On the basis of getting a set of singular spectrum component(SSC), the optimal SSC was selected according to the kurtosis maximum criterion and the SVD processing was carried out. The singular value energy standard spectrum was used to adaptively determine the signal reconstruction order to restore the signals and improve the noise reduction effectiveness. Finally, the gear fault features were extracted by using envelope demodulation. The proposed method was applied in simulated signals and gear measured signals, and compared with the traditional envelope spectrum, SSD envelope spectrum and empirical mode decomposition combined with SVD(EMD-SVD) methods. The results show that the proposed method has better effectiveness of noise reduction and feature extraction, and may realize the identification of gear faults more effectively.

Key words: improved singular spectrum decomposition(ISSD), singular value decomposition(SVD), dispersion entropy, gear, fault feature extraction

摘要: 针对齿轮故障特征微弱,在强背景噪声下难以有效提取的问题,提出了一种改进奇异谱分解(ISSD)结合奇异值分解(SVD)的齿轮故障特征提取方法。针对奇异谱分解(SSD)算法中模态参数需凭经验选取的缺陷,基于散布熵优化算法对SSD算法进行了改进,在得到既定的一组奇异谱分量的基础上,根据峭度值最大准则筛选出了最佳奇异谱分量并进行了SVD处理,采用奇异值能量标准谱自适应地确定了信号重构阶数以还原信号和提高降噪效果。最后对信号进行包络解调以提取齿轮故障特征,将所提方法运用到仿真信号和齿轮实测信号中,并同传统包络谱、SSD包络谱以及经验模态分解结合SVD(EMD-SVD)方法进行了对比分析,结果表明,所提方法的降噪和特征提取效果更佳,能够更加有效地实现齿轮故障的判别。

关键词: 改进奇异谱分解, 奇异值分解, 散布熵, 齿轮, 故障特征提取

CLC Number: