基于CEEMDAN-Hilbert-CNN的滚动轴承故障诊断

全文下载： 202102014.pdf

文章编号： 1672-6987（2021）02-0097-07； DOI： 10.16351/j.1672-6987.2021.02.014

李雪颖， 朱哈娜， 刘慧明*(青岛科技大学 自动化与电子工程学院，山东 青岛 266061)

摘要： 在故障诊断方面，针对希尔伯特黄变换信号分解存在模态混叠和人工神经网络面对高维数据收敛性差、特征分类误差大等问题，提出一种改进的希尔伯特黄变换与卷积神经网络相结合的滚动轴承故障诊断的新方法：将希尔伯特黄变换的经验模态分解替换为自适应白噪声的完整经验模态分解，并与卷积神经网络相结合。提出的滚动轴承故障诊断新方法进行故障诊断过程：首先应用改进的希尔伯特黄变换对数据进行处理得到一个瞬时频率矩阵，然后将瞬时频率矩阵进行重构，最后搭建卷积神经网络，将重构数据输入卷积神经网络进行分类。经反复实验并与已有方法进行比较，可验证所提出的方法是合理且行之有效的。

关键词： 滚动轴承； 故障诊断； 自适应白噪声的完整经验模态分解; 希尔伯特黄变换； 卷积神经网络

中图分类号： TH 133.3; TP 18文献标志码： A

引用格式： 李雪颖， 朱哈娜， 刘慧明. 基于CEEMDANHilbertCNN的滚动轴承故障诊断［J］. 青岛科技大学学报(自然科学版), 2021, 42(2): 97103.

LI Xueying，ZHU Hana， LIU Huiming. Fault diagnosis of rolling bearing based on CEEMDANHilbertCNN［J］. Journal of Qingdao University of Science and Technology（Natural Science Edition), 2021， 42(2)： 97103.

Fault Diagnosis of Rolling Bearing Based on CEEMDANHilbertCNN

LI Xueying， ZHU Hana， LIU Huiming

(College of Automation and Electronic Engineering, Qingdao University of Science and Technology， Qingdao 266061, China)

Abstract: In terms of fault diagnosis, HilbertHuang transform signal decomposition, modal aliasing and artificial neural networks have problems such as poor convergence of highdimensional data and large classification errors. This paper proposes a new method for fault diagnosis of rolling bearings combined with an improved Hilbert yellow transform and convolutional neural networks: Empirical mode decomposition is replaced with complete ensemble empirical mode decomposition with adaptive noise, which is combined with convolutional neural networks. The new method of rolling bearing fault diagnosis proposed in this paper is used to perform the fault diagnosis process: first, the improved HilbertHuang transform is used to process the data to obtain an instantaneous frequency matrix, then the instantaneous frequency matrix is reconstructed, and finally a convolutional neural network is built. The reconstructed data is input into a convolutional neural network for classification. After repeated experiments and comparison with existing methods, it can be verified that the method proposed in this paper is reasonable and effective.

Key words: rolling bearing；fault diagnosis； complete ensemble empirical mode decomposition with adaptive noise；HilbertHuang transform； convolution neural network

收稿日期： 20200412

基金项目： 青岛科技大学科研启动基金资助项目(010022586).

作者简介： 李雪颖(1995—),女,硕士研究生.*通信联系人.