基于多重降噪的滚动轴承声信号故障特征提取  

Fault Feature Extraction of Rolling Bearing Sound Signals Based on Multiple Noise Reduction

作　　者：王涛 胡定玉[1] 廖爱华[1] 师蔚[1] 丁亚琦 陶涛 WANG Tao;HU Dingyu;LIAO Aihua;SHI Wei;DING Yaqi;TAO Tao(School of Urban Rail Transportation,Shanghai University of Engineering and Technology,Shanghai 201620,China;Vehicle Branch,Shanghai Metro Maintenance and Guarantee Co.,Ltd.,Shanghai 200235,China)

机构地区：[1]上海工程技术大学城市轨道交通学院,上海201620 [2]上海地铁维护保障有限公司车辆分公司,上海200235

出　　处：《噪声与振动控制》2021年第3期95-100,119,共7页Noise and Vibration Control

基　　金：国家自然科学基金资助项目(51605274)。

摘　　要：针对滚动轴承故障诊断中声信号信噪比较低、特征提取困难的问题,提出多重降噪轴承故障特征提取方法。该方法首先用最小熵解卷积对故障轴承声信号进行预处理来提高信噪比,然后利用局部特征尺度分解将处理后的信号分解为多个内禀尺度分量,进一步利用相关系数-峭度值原则,筛选出最佳内禀尺度分量进行重构,最后通过1.5维Teager能量谱提取轴承故障特征。仿真及实验结果表明,相较于单一使用最小熵解卷积或局部特征尺度分解等降噪方法,多重降噪方法可以在信噪比极低的情况下有效提取故障特征。Aiming at the problem of low signal-to-noise ratio(SNR)of acoustic signal and the difficulty in feature extraction in fault diagnosis of rolling bearings,a feature extraction method for multiple noise reduction bearing faults is proposed.In this method,using the minimum entropy convolution for fault bearing acoustic signal preprocessing,the signalto-noise ratio is improved.Then,based on the advantage of local characteristic scale decomposition,the processed signal is decomposed into several intrinsic scale weight components.Furthermore,using the correlation coefficient and kurtosis value principle,the best intrinsic scale components are selected and reconstructed.Finally,through the 1.5D Teager energy spectrum,the bearing fault feature is extracted.Simulation and experimental results show that compared with the single use of noise reduction method of minimum entropy convolution or local characteristic scale decomposition,the multiple noise reduction method can effectively extract the fault features in the case of extremely low SNR.

关 键 词：故障诊断 滚动轴承 最小熵解卷积 局部特征尺度分解 1.5维Teager能量谱 

分 类 号：TH133.3[机械工程—机械制造及自动化] T206[一般工业技术]