作者：周林春¹, 陈春俊^1,2

作者单位：1. 西南交通大学机械工程学院，四川 成都 610031;
2. 西南交通大学 轨道交通运维技术与装备四川省重点实验室，四川 成都 610031

关键词：滚动轴承;故障诊断;连续小波变换;复数卷积神经网络

摘要：

针对基于实数卷积神经网络的滚动轴承故障诊断方法对振动信号幅相信息利用不充分的问题，提出一种基于复数卷积神经网络的故障诊断模型。该模型以一维振动信号经连续小波变换得到的时频复数矩阵为输入，通过复数卷积神经网络独有的复数卷积方式提取和融合信号的幅值和相位特征，并通过全连接层和Softmax实现故障诊断结果输出。结果表明：采用复数卷积神经网络模型的故障诊断方法具有更强的抗噪声鲁棒性，在添加信号噪声的不同转速工况之间能保持更好的泛化性能，可提高滚动轴承故障诊断的准确率。

Fault diagnosis of rolling bearing based on complex-valued convolutional neural network
ZHOU Linchun¹, CHEN Chunjun^1,2
1. School of Mechanical Engineering, Southwest Jiaotong University, Chengdu 610031, China;
2. Technology and Equipment of Rail Transit Operation and Maintenance Key Laboratory of Sichuan Province, Southwest Jiaotong University, Chengdu 610031, China
Abstract: In order to solve the problem of rolling bearing fault diagnosis method based on real-valued convolutional neural network(RV-CNN) makes insufficient use of the amplitude and phase information of vibration signals, a fault diagnosis model based on complex-valued convolutional neural network(CV-CNN) was proposed. The model takes the time-frequency complex matrix obtained by continuous wavelet transform(CWT) of one-dimensional vibration signal as input, extracts and fuses the amplitude and phase characteristics of the signal through the complex-valued convolutional method unique to CV-CNN, and through the fully connected layer and Softmax to achieve diagnostic results output. The results show that the fault diagnosis method with CV-CNN model has stronger anti-noise robustness, can maintain better generalization performance between different speed conditions with signal noise added and improved the accuracy of diagnosis of rolling bearing faults.
Keywords: rolling bearing;fault diagnosis;continuous wavelet transform;complex-valued convolutional neural network
2020, 46(11):109-115  收稿日期: 2020-02-26;收到修改稿日期: 2020-04-28
基金项目: 国家自然科学基金资助项目(51975487)
作者简介: 周林春(1996-),男,四川成都市人,硕士研究生,专业方向为信号处理与故障诊断
参考文献
[1] 张超, 陈建军, 徐亚兰. 基于EMD分解和奇异值差分谱理论的轴承故障诊断方法[J]. 振动工程学报, 2011, 24(5): 539-545
[2] HINTON G, OSINDERO S, THE Y. A fast learning algorithm for deep belief nets[J]. Neural Computation, 2006, 18(7): 1527-1554
[3] 杨宇, 于德介, 程军圣, 等. 经验模态分解(EMD)在滚动轴承故障诊断中的应用[J]. 湖南大学学报(自然科学版), 2003, 30(5): 25-28
[4] 艾树峰. 基于双树复小波变换的轴承故障诊断研究[J]. 中国机械工程, 2011(20): 56-61
[5] 王硕, 唐建明, 徐荣华. 使用PCA及SVM算法进行滚动轴承故障检测[J]. 机电工程技术, 2005(2): 31-32,99
[6] 乔文生, 陈兴辉, 艾士娟, 等. 基于小波包和BP神经网络的滚动轴承故障诊断[C]//全国设备监测与诊断学术会议, 2008.
[7] 雷文平, 吴小龙, 陈超宇, 等. 基于自动编码器和SVM的轴承故障诊断方法[J]. 郑州大学学报(工学版), 2018, 39(5): 68-72
[8] 单外平, 曾雪琼. 基于深度信念网络的信号重构与轴承故障识别[J]. 电子设计工程, 2016, 24(4): 67-71
[9] 杨平, 苏燕辰. 基于卷积门控循环网络的滚动轴承故障诊断[J]. 航空动力学报, 2019, 34(11): 2432-2439
[10] MA P, ZHANG H L, FAN W H, et al. A novel bearing fault diagnosis method based on 2D image representation and transfer learning–convolutional neural network[J]. Measurement Science and Technology, 2019, 30(5): 055402
[11] ISLAM M, KIM J. Automated bearing fault diagnosis scheme using 2D representation of wavelet packet transform and deep convolutional neural network[J]. Computers in Industry, 2019, 106: 142-153
[12] 袁建虎, 韩涛, 唐建, 等. 基于小波时频图和CNN的滚动轴承智能故障诊断方法[J]. 机械设计与研究, 2017, 33(2): 93-97
[13] TRABELSI C, BILANIUK O, ZHANG Y, et al. Deep complex networks[C]//ICLR, 2018.
[14] POPA C A, CERNĂZANU-GLĂVAN C. Fourier transform-based image classification using complex-valued convolutional neural networks[C]//International Symposium on Neural Networks, 2018.
[15] 何晓霞, 沈玉娣, 张西宁. 连续小波变换在滚动轴承故障诊断中的应用[J]. 机械科学与技术, 2001, 20(4): 571-572
[16] IOFFE S, SZEGEDY C. Batch normalization: accelerating deep network training by reducing internal covariate shift[C]// Proc. ICML, 2015.
[17] HIROSE A. Complex-valued neural networks[M]. Berlin: Springer-Verlag, 2006.
[18] SUKSMONO A B , HIROSE A. Beamforming of ultra-wideband pulses by a complex-valued spatio-temporal multilayer neural network[J]. International Journal of Neural Systems, 2005, 15(1/2): 85-91
[19] LAURENS V D M, HINTON G. Visualizing data using t-SNE[J]. Journal of Machine Learning Research, 2008, 9(2605): 2579-2605