期刊文献+

基于改进1DCNN-GRU的滚动轴承故障诊断 被引量:2

An Improved 1DCNN-GRU for Rolling Bearing Fault Diagnosis
下载PDF
导出
摘要 针对传统滚动轴承故障诊断模型无法充分利用信号的空间及时间特征,需要大量专业知识等问题,提出一种改进一维卷积神经网络(1DCNN)与门控递归神经网络(GRU)结合的故障诊断方法。首先,利用具有不同卷积核的卷积层最大化提取信号的空间特征信息;其次,提出改进的线性修正单元(IReLU)增强网络的特征提取能力;然后,引入堆叠GRU模块进一步提取1DCNN模块输出数据的时间特征,完成空间及时间特征融合;最后,对融合后的特征进行识别。实验表明所提方法故障识别率高达99.96%,对不同负载下的数据均具有较高的识别率及较强的泛化性能。 To solve the issue that most traditional models of rolling bearing fault diagnosis can not fully exploit the spatial and temporal characteristics of the signals and require lots of expert knowledge,a novel fault diagnosis method based on one-dimensional convolutional neural networks(1DCNN)and gated recurrent neural networks(GRU)was proposed.Firstly,the convolution layers with different convolution kernels were used to maximize the extraction of spatial features in the signal.Secondly,the IReLU was proposed to enhance the feature extraction ability of the network.Then,the stacked GRU was introduced to further extract the temporal features in the output data of the 1DCNN module and complete the fusion of the spatial and temporal features.Finally,the fused features were recognized.The experimental results showed that the fault recognition accuracy of the proposed method is up to 99.96%,and the proposed method has a high identification accuracy and strong generalization performance for datasets under different loads.
作者 金海龙 马吴旭 孟宗 孙登云 曹伟 樊凤杰 JIN Hai-long;MA Wu-xu;MENG Zong;SUN Deng-yun;CAO Wei;FAN Feng-jie(College of Electrical Engineering,Yanshan University,Hebei,Qinhuangdao 066000,China)
出处 《计量学报》 CSCD 北大核心 2023年第9期1423-1428,共6页 Acta Metrologica Sinica
基金 国家自然科学基金(52075470) 中央引导地方科技发展资金(206Z4301G) 河北省高水平人才支持计划(A202102001) 河北省省级重点实验室绩效补助经费项目(21567608H)。
关键词 计量学 故障诊断 滚动轴承 特征融合 一维卷积神经网络 门控递归神经网络 metrology fault diagnosis rolling bearing feature fusion 1DCNN GRU
  • 相关文献

参考文献7

二级参考文献54

共引文献1310

同被引文献27

引证文献2

二级引证文献1

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部