优化VMD和改进残差网络的齿轮箱故障诊断研究
Optimizing VMD and Improving Residual Network Research on Gearbox Fault Diagnosis
郭梓良 1郝如江 1杨文哲 1王一帆 1张建超1
作者信息
- 1. 石家庄铁道大学机械工程学院,石家庄 050043
- 折叠
摘要
为提高齿轮箱故障诊断的准确率,采用变分模态分解(Variational mode decomposition,VMD)和一维残差网络(Resnet)相结合的齿轮箱故障识别方法.首先使用磷虾群算法对VMD中的惩罚因子α和分解层数K进行寻优,其次使用VMD对故障信号进行分解产生若干条本征模函数(Intrinsic mode function,IMF)并通过使用相关系数对IMF分量进行筛选,并将筛选后的分量作为改进残差网络的输入数据进行网络训练和测试,得到故障识别结果.通过与VMD-随即森林、VMD-CNN等方法的平均准确率进行横向比较,结果表明,本文的方法用来识别故障类型更为准确.
Abstract
In order to improve the accuracy of gearbox fault diagnosis,this paper adopts a gearbox fault identification method combining VMD(Variational mode decomposition)and 1D-Resnet(One-dimensional residual network).Firstly,the krill swarm algorithm is used to optimize the penalty factor α and the number of decomposition layers K in the VMD.Secondly,the VMD is used to decompose the fault signal to generate several IMF(Intrinsic mode functions),and the correlation coefficient is used to screen the IMF component.Finally,the screened component is used as the input data of the improved residual network for network training and testing,and the fault identification result is obtained.Through the horizontal comparison with the average accuracy of VMD-random forest,VMD-CNN and other methods,the results show that the method in this paper is more accurate for identifying gearboxfault types.
关键词
磷虾群算法/VMD/改进残差网络/齿轮箱故障Key words
Krill swarm algorithm/VMD/improved residual network/gearbox failure引用本文复制引用
出版年
2024
国家科技期刊平台
NSTL
万方数据