结合LMS滤波和卷积盲分离的轴承故障诊断方法  

Bearing Fault Diagnosis Method Combining LMS Filtering and Blind Deconvolution

作　　者：陆建涛 殷齐涛 杜军 杨旷智 李舜酩[1] LU Jiantao;YIN Qitao;DU Jun;YANG Kuangzhi;LI Shunming(College of Energy and Power Engineering,Nanjing University of Aeronautics and Astronautics Nanjing,210001,China;AECC Sichuan Gas Turbine Establishment Mianyang,621000,China)

机构地区：[1]南京航空航天大学能源与动力学院,南京210001 [2]中国航发四川燃气涡轮研究院,绵阳621000

出　　处：《振动．测试与诊断》2024年第6期1120-1126,1246,共8页Journal of Vibration,Measurement & Diagnosis

基　　金：国家重点研发资助项目(2020YFB1709801);中国航发四川燃气涡轮研究院外委课题资助项目(GJLZ-2020-0056);中央高校基本科研业务费资助项目(NS2021010,YAH20008);江苏省双创博士计划资助项目(202030364)。

摘　　要：针对强噪声导致卷积盲源分离故障源信号估计精度较低的问题,提出一种结合最小均方算法(least mean square,简称LMS)滤波和卷积盲分离(robust multichannel blind deconvolution,简称RobustMBD)的滚动轴承复合故障诊断方法。首先,利用LMS滤波对含噪的轴承故障信号进行去噪预处理,降低噪声对故障信号的影响;其次,通过构建时滞关联模型将卷积混合模型转换为瞬时混合模型,并以归一化峭度为分离判据,采用精确线搜索替代迭代搜索,得到卷积盲分离方法鲁棒多通道盲解卷积;然后,对降噪后的复合故障信号采用鲁棒多通道盲解卷积进行盲源分离,得到轴承的独立故障信号;最后,通过仿真和滚动轴承试验数据对提出的滚动轴承复合故障诊断方法进行了验证。结果表明,与传统鲁棒多通道盲解卷积相比,在强噪声情况下,提出的方法能够有效分离出所有的故障信号。Strong noise will lead to low estimation accuracy of convolution blind source separation of fault source signals.To solve this problem,a composite fault diagnosis method for rolling bearings based on the least mean square(LMS)filtering and convolutional blind separation is proposed.This method uses LMS filtering to preprocess the noisy bearing fault signals.With the convolution blind separation method called robust multichannel blind deconvolution(RobustMBD)to separate the composite fault signals after noise reduction,the independent fault signals of the bearing are obtained.The RobustMBD constructs a time-delay correlation model to extend the convolution condition to the instantaneous condition,using the normalized kurtosis and the precise line searching.This method is verified by simulation and rolling bearing test data.The results show that com-pared with the traditional RobustMBD,the proposed method can effectively separate all fault signals under strong noise conditions.

关 键 词：滚动轴承 故障诊断 最小均方算法滤波 卷积盲源分离 

分 类 号：TH212[机械工程—机械制造及自动化] TH213.3