基于IEWT-IDWAE的滚动轴承故障识别  

Fault Identification of Rolling Bearings Based on Improved Empirical Wavelet Transform and Improved Deep Wasserstein Auto Encoder

作　　者：苏珉 袁朴 陈高杰 SU Min;YUAN Pu;CHEN Gaojie(Department of Mechanical and Electrical Engineering,Sichuan Engineering Technical College,Deyang 618000,China;School of Mechanical&Automotive Engineering,South China University of Technology,Guangzhou 510640,China;Zhongshan Maret CNC Technology Co.,Ltd.,Zhongshan 528437,China)

机构地区：[1]四川工程职业技术学院机电工程系,四川德阳618000 [2]华南理工大学机械与汽车工程学院,广州510640 [3]中山迈雷特数控技术有限公司,广东中山528437

出　　处：《轴承》2023年第1期69-75,共7页Bearing

摘　　要：针对噪声导致深度学习识别滚动轴承故障时深层网络收敛速度变慢以及识别率降低的问题,提出一种改进经验小波变换(IEWT)结合改进深层Wasserstein自动编码器(IDWAE)的故障识别模型。首先,针对经验小波变换的过分解问题,提出一种振动信号频谱有效边界划分方法,进而将信号自动分解为不同频段的调幅-调频分量;然后,利用一种新的AM-FM分量筛选指标选择主要分量进行重构,实现对信号的有效降噪;最后,针对变分自编码器训练困难的缺陷,引入Wasserstein自编码器,根据中间层神经元的激活值对神经元大小进行自动增减以构造IDWAE,将经IEWT降噪后的信号输入IDWAE进行自动特征提取和故障识别。试验结果表明:IEWT-IDWAE在一定程度上缓解了工程人员对繁琐的特征提取和特征选择的依赖,对噪声的鲁棒性高,故障识别率达到了99.57%,标准差仅0.12,故障识别能力优于其他组合模型方法。Aimed at slow convergence speed of deep network and low identification rate caused by noise when deep learning is used to identify the faults of rolling bearings,a fault identification model of improved empirical wavelet transform(IEWT)combined with improved deep Wasserstein auto encoder(IDWAE)is proposed.Firstly,in view of overdecomposition problem of empirical wavelet transform,a method is proposed for dividing the effective boundary of vibration signal spectrum,and then the signal is automatically decomposed into amplitude modulation-frequency modulation components of different frequency bands.Secondly,a new screening index for AM-FM components is employed to select the main components for reconstruction,and the effective noise reduction of signal is realized.Finally,in view of defects of difficult training of variational auto encoder,the Wasserstein auto encoder is introduced,the size of neurons is automatically increased and decreased to construct IDWAE according to activation values of middle-layer neurons,and the signals after IEWT noise reduction are input into IDWAE for automatic feature extraction and fault identification.The experimental results show that IEWT-IDWAE alleviates the dependence of engineers on cumbersome feature extraction and feature selection to a certain extent,and has high robustness to noise,the fault identification rate reaches 99.57%,the standard deviation is only 0.12,and the fault identification ability is better than other combined model methods.

关 键 词：滚动轴承 故障诊断 小波变换 编码器 深度学习 噪声 调幅 调频 

分 类 号：TH133.33[机械工程—机械制造及自动化] TP18[自动化与计算机技术—控制理论与控制工程]