基于多尺度卷积神经网络的滚动轴承故障识别*

PDF(1077 KB)

集成电路与嵌入式系统 ›› 2023, Vol. 23 ›› Issue (12) : 88-91.

应用精选

基于多尺度卷积神经网络的滚动轴承故障识别^*

张志艺^1,2, 李光亚^1,2, 王子一², 李旭卿²

作者信息 +

Fault Identification of Rolling Bearing Based on Multi-scale Convolutional Neural Network

Zhang Zhiyi^1,2, Li Guangya^1,2, Wang Ziyi², Li Xuqing²

Author information +

文章历史 +

摘要

针对当前基于卷积神经网络的滚动轴承故障信号识别多采用单一尺度卷积核,对特征提取不充分、多尺度特征难提取,且超参数选取多依据人工经验等问题,提出一种多尺度卷积神经网络识别方法。首先,直接将原始时域信号作为输入,可有效保留原始信号特征且不需要人们对故障机理、信号处理专业知识的认识。其次,通过多尺度卷积神经网络对信号特征学习,并引入粒子群优化算法来寻求最优的尺度信息。最后,用滚动轴承数据集来验证模型性能,并与其他模型做对比。实验结果表明,本文模型平均识别准确率达到99.45%,高于其他模型,且抗噪声能力优于其他模型,证明所提方法具有可行性。

Abstract

Aiming at the current problem of using a single scale convolution kernel in time domain signal recognition based on convolutional neural networks,which leads to insufficient feature extraction,and the selection of hyperparameters is mostly based on artificial experience,a multi-scale convolution neural network (MCNN) recognition method is proposed.First,directly using the original time domain signal as input not only effectively preserves the original signal characteristics,but also does not require people's understanding and experience of the fault mechanism,as well as professional knowledge of signal processing.Then,the multi-scale convolutional neural network is used to learn the signal characteristics,and particle swarm optimization (PSO) is introduced to seek the optimal scale information.Finally,the performance of the model is verified using a rolling bearing dataset and compared with other models.The experiment results show that the average recognition accuracy reaches 99.45%,which is superior to other models,proving the feasibility of the proposed method.

导出引用

张志艺, 李光亚, 王子一, 李旭卿. 基于多尺度卷积神经网络的滚动轴承故障识别^*[J]. 集成电路与嵌入式系统. 2023, 23(12): 88-91

Zhang Zhiyi, Li Guangya, Wang Ziyi, Li Xuqing. Fault Identification of Rolling Bearing Based on Multi-scale Convolutional Neural Network[J]. Integrated Circuits and Embedded Systems. 2023, 23(12): 88-91

中图分类号： TP391

参考文献

[1] 李俊卿.滚动轴承故障诊断技术及其工业应用[D].郑州:郑州大学,2010.
[2] 邓佳林.基于CNN的滚动轴承变工况故障诊断方法研究[D].成都:西南交通大学,2020.
[3] 夏新涛,常振,叶亮,等.滚动轴承性能稳定性及可靠性分析[J].仪器仪表学报,2017,38(6):421-1431.
[4] 汪祖民,张志豪,秦静,等.基于卷积神经网络的机械故障诊断技术综述[J].计算机应用,2022,42(4):1036-1043.
[5] CHEN Xiaohan,ZHANG Beike,GAO Dong.Beang fault diagnosis base on multi-scale CNN and LSTM model[J].Journal of Intelligent Manufacturing,2021,32(4):971-987.
[6] WANG Huang,LlU Zhiliang,PENG Dandan,et al.Understanding and learning discriminant features based on multitaction.1DCNN for wheelset bearing fault diagnosis[J].IEEE Transactions on Industrial Informatics,2020,16(9):5735-5745.
[7] 王琦,邓林峰,赵荣珍.基于改进一维卷积神经网络的滚动轴承故障识别[J].振动与冲击,2022,41(3):216 -223.
[8] 熊红林,樊重俊,赵珊,等.基于多尺度卷积神经网络的玻璃表面缺陷检测方法[J].计算机集成制造系统,2020,26(4):900-909.
[9] 谭震,郭新蕾,李甲振,等.基于多尺度卷积神经网络的管道泄漏检测模型研究[J].水利学报,2023,54(2):220-231.
[10] PENG D,LIU Z,WANG H,et al.A novel deeper one dimensional CNN with residual learning for fault diagnosis of wheelset bearings in high-speed trains[J].IEEE Access,2019(7):10278-10293.
[11] JING L,ZHAO M,LI P,et al.A convolutional neural network based feature learning and fault diagnosis method for the condition monitoring of gearbox[J].Measurement,2017(111):1-10.
[12] 郭晨,简涛,徐从安,等.基于深度多尺度一维卷积神经网络的雷达舰船目标识别[J].电子与信息学报,2019,41(6):1302-1309.
[13] 杜坚,胡洁,仇芝.基于改进粒子群相关向量机电机轴承故障诊断[J].应用力学学报,2019,36(4):779-784,992.
[14] Zou Yisheng.A method for predicting the remaining usful life of rolling bearings under different working conditions based on multidomain adversarial networks[J].Measurement,2022(188).