基于多尺度动态卷积和GRU的轴承故障诊断  

Bearing Fault Diagnosis Based on Multi-Scale Dynamic Convolution and GRU

在线阅读下载全文

作  者:董绍江[1] 彭银山 邹松 黄翔 DONG Shaojiang;PENG Yinshan;ZOU Song;HUANG Xiang(School of Mechatronics and Vehicle Engineering,Chongqing Jiaotong University,Chongqing 400074,China;School of Traffic&Transportation,Chongqing Jiaotong University,Chongqing 400074,China)

机构地区:[1]重庆交通大学机电与车辆工程学院,重庆400074 [2]重庆交通大学交通运输学院,重庆400074

出  处:《组合机床与自动化加工技术》2025年第3期150-154,共5页Modular Machine Tool & Automatic Manufacturing Technique

基  金:国家自然科学基金资助项目(51775072);重庆市科技创新领军人才支持计划项目(CSTCCCXLJRC201920);重庆市教委科学技术研究项目(KJZD-K202300711)。

摘  要:针对传统轴承故障诊断过程中忽略轴承振动信号的关联时间维度信息的问题,提出了基于多尺度动态扩张卷积神经网络(MSDDCNN)和门控循环单元网络(GRU)的轴承故障诊断方法。首先利用不同尺寸宽卷积核从各个维度捕捉振动信号多维特征以增大感受野;其次引入动态加权层自适应选择卷积核尺度的大小并自动地给予特征序列中的不同部分不同的权重,更加充分提升特征表示的能力;最后利用门控循环单元充分提取振动信号中不同尺度的时序特征,以加强各个维度间前后时间维度关联信息。实验结果表明,所提方法在PU和JNU公开数据集上平均准确率分别为98.79%和98.65%。为验证所提网络模型诊断有效性,所提方法在某公司自制的轴承故障数据集(CME)也表现出较高的准确率和较大抗噪声能力,为有效诊断旋转部件故障提供了实际依据。A bearing fault diagnosis method based on multi-scale dynamic dilated convolutional neural network(MSDDCNN)and gated recurrent unit network(GRU)is proposed to address the problem of ignoring the correlation time dimension information of bearing vibration signals in traditional bearing fault diagnosis processes.This method first uses wide convolutional kernels of different sizes to capture multidimensional features of vibration signals from various dimensions to increase the receptive field;Secondly,a dynamic weighting layer is introduced to adaptively select the size of the convolutional kernel and automatically assign different weights to different parts of the feature sequence,which further enhances the ability of feature representation;Finally,the gated loop unit is utilized to fully extract temporal features of different scales from the vibration signal,in order to enhance the correlation information between the front and back time dimensions of each dimension.The experimental results show that the proposed method has an average accuracy of 98.79%and 98.65%on the PU and JNU public datasets,respectively.To verify the diagnostic effectiveness of the proposed network model,the method also demonstrated high accuracy and strong noise resistance in the self-made bearing fault dataset(CME)of a company,providing practical basis for effective diagnosis of rotating component faults.

关 键 词:多尺度 动态卷积 扩张卷积 GRU 故障诊断 

分 类 号:TH133.33[机械工程—机械制造及自动化] TG66[金属学及工艺—金属切削加工及机床]

 

参考文献:

正在载入数据...

 

二级参考文献:

正在载入数据...

 

耦合文献:

正在载入数据...

 

引证文献:

正在载入数据...

 

二级引证文献:

正在载入数据...

 

同被引文献:

正在载入数据...

 

相关期刊文献:

正在载入数据...

相关的主题
相关的作者对象
相关的机构对象