基于VMD-IMDE-PNN的滚动轴承故障诊断方法  

Rolling Bearing Fault Diagnosis Method Based on VMD-IMDE-PNN

作　　者：刘备 蔡剑华[1] 彭梓齐 LIU Bei;CAI Jianhua;PENG Ziqi(College of Mathematics and Physics,Hunan University of Arts and Science,Changde 415000,Hunan,China)

机构地区：[1]湖南文理学院数理学院,湖南常德415000

出　　处：《噪声与振动控制》2022年第5期96-101,133,共7页Noise and Vibration Control

基　　金：国家自然科学基金资助项目(U2031112);湖南省自然科学基金资助项目(2021JJ30469);湖南省教育厅科学研究基金资助项目(20B405);博士科研启动基金资助项目(E07021011)。

摘　　要：为了提高滚动轴承故障诊断的准确性,提出一种变分模态分解(Variational Mode Decomposition,VMD)、改进粗粒化多尺度散布熵(Improved Coarse-grained Multi-scale Dispersion Entropy,IMDE)和概率神经网络(Probabilistic Neural Network,PNN)相结合的滚动轴承故障诊断方法。首先对振动信号进行VMD处理,根据互相关系数准则筛选最佳模态分量,突显振动信号的故障特性;然后针对多尺度散布熵(Multi-scale Dispersion Entropy,MDE)不稳定的缺点,对MDE的粗粒化过程进行改进,提出IMDE的非线性分析方法。模拟信号分析结果表明,相比于MDE方法,IMDE方法降低了熵值波动,提高了熵值稳定性。将两种方法运用于实际滚动轴承实验数据,发现相比于MDE,IMDE熵值曲线更平滑稳定,不同滚动轴承状态下的IMDE熵值曲线区分更加明显。最后采用PNN对提取的特征进行识别,与MPE-PNN,MDE-PNN以及VMD-MDE-PNN方法相比,所提的VMD-IMDE-PNN方法能精确地识别滚动轴承的故障类型,且识别率更高。In order to improve the accuracy of rolling bearing fault diagnosis, a fault diagnosis method based on variational mode decomposition(VMD), improved coarse-grained multi-scale dispersion entropy(IMDE) and probabilistic neural network(PNN) was proposed. Firstly, the vibration signals were processed by VMD, and the optimal modal components were selected according to the cross-relation number criterion to highlight the fault characteristics of the vibration signals. Then, aiming at the disadvantage of instability of multi-scale dispersion entropy(MDE), a nonlinear analysis method of the IMDE was proposed to improve the coarse-granulation process of the MDE. Simulation results show that the IMDE method can reduce entropy fluctuation and improve entropy stability compared with the MDE method. By applying the two methods to analyzing actual rolling bearing experimental data, it is found that IMDE entropy curves are smoother and more stable compared with MDE, and the differences of IMDE entropy curves under different rolling bearing conditions are more distinct. Finally, PNN was used to identify the extracted features. Compared with MPE-PNN, MDEPNN, and VMD-MDE-PNN methods, the proposed VMD-IMDE-PNN method can more accurately identify the fault types of rolling bearings, and has higher recognition rate.

关 键 词：故障诊断 变分模态分解 改进粗粒化多尺度散布熵 概率神经网络 滚动轴承 

分 类 号：TH133.3[机械工程—机械制造及自动化] TN911.7[电子电信—通信与信息系统]