ISSN10003762 CN41 1148/TH
轴承2016年11期 Bearing 2016 , No. 11
广义S变换时频谱SVD降噪的滚动轴承故障
冲击特征提取方法
朱怡，蒋思源
（中国民航大学，天津
OOOOE
5357
摘要：针对强背景噪声下滚动轴承故障冲击特征难以提取的特点，提出了基于广义S变换时频谱SVD降噪的滚动轴承故障冲击特征提取方法。利用广义S变换时频分辨率高，时频谱能量集中，适合处理与分析非平稳冲击信号的特点，将广义S变换时频谱系数矩阵作为SVD的Hankel矩阵，以奇异值差分谱峰值群最后一个峰值点对应的奇异值作为置零阅值，最后对降噪后的数据矩阵进行广义S逆变换得到时域冲击特征信号。仿真及实际研究表明，该方法能够有效地提取出低信噪比信号中的周期性冲击特征，并能够有效地提取轴承故障振动信号中的冲击特征频率。
关键词：滚动轴承：广义S变换；奇异值分解；冲击特征
中图分类号：TH133.33;TN911.7
文献标志码：B
文章编号：10003762(2016)11005305
RollingBearingDefaultImpactFeatureExtractingMethodBased on Generalized STransformTime-Frequency SpectrumDenoisedby SvD
Zhu Yi,Jiang Siyuan
( Civil Aviation University of China, Tianjing 300300, China)
Abstract: Based on the characteristic that the feature extraction of rolling bearings weak fault is very hard under strong background noise, a new impact feature extracting method based on generalized S transfom time frequency spectrum denoised by SVD is proposed. Generalized S transform time frequency resolution is high, the frequency spectrum en ergy concentrated and suitable for processing and analyzing non stationary characteristics of impact signal. The Hankel matrix is composed of generalized S transform spectrum coefficient. Singular values difference spectrum peak of the last peak point corresponding singular value is set as zero threshold. In the last, the time domain feature signal is got from the denoised data matrix of generalized S after inverse transformation. The simulation results show that the method based on generalized S transform time frequency spectrum denoised by SVD can elfectively extract the periodic impact characteristic in the low SNR signals. Finally this method is applied to impact rolling bearing fault feature extraction successfully, efectively extracting the characleristic frequency.
Key words : rolling bearing: generalized S transfom; singular value decomposition; impact featurt
滚动轴承的故障常常表现为轴承工作表面的麻点、裂纹和剥落等局部损伤[1-2]，这些局部损伤在工作过程中会被撞击，从而产生周期性的冲击振动信号，该振动信号的频率即轴承故障特征
收稿日期：2016-06-06；修回日期：2016-06-29
基金项目：中央高校基本科研业务费（3122013x003， 3122015F003)
作者简介：朱怡（1982一），女，讲师，博士，主要研究方向为虚
拟仿真技术、自动化控制等，Email：kyyx0416@163.com 万方数据
率，可用于对轴承故障的诊断。然而，旋转机械设备结构复杂目工作环境多样化，故障振动信号的冲击特征常被淹没在背景信号及噪声中。因此，有效地提取出故障冲击特征，是对滚动轴承相关故障进行诊断的关键[3-5]
奇异值分解（SingularValueDecomposition， SVD)降噪方法是一种高效的非线性滤波方法，对宽带噪声信号的去噪效果优异，适用于包含强噪声信号的滚动轴承振动信号去噪，然而，将一维的故障冲击时域信号有效地构造成Hankel矩阵是