基于优化VMD和MHA-DenseNet的滚动轴承故障诊断  

Fault Diagnosis of Rolling Bearing Based on Optimized VMD and MHA-DenseNet

作　　者：安冬[1] 梁彬彬 叶井启 须颖[1] 邵萌[1] 刘振鹏[1] AN Dong;LIANG Bin-bin;YE Jing-qi;XU Ying;SHAO Meng;LIU Zhen-peng(School of Mechanical Engineering,Shenyang Jianzhu University,Shenyang 110168,China;Liaoning Jiangyang Technology Co.,Ltd.,Shenyang 110000,China)

机构地区：[1]沈阳建筑大学机械工程学院,沈阳110168 [2]辽宁江扬科技有限公司,沈阳110000

出　　处：《组合机床与自动化加工技术》2022年第6期114-118,共5页Modular Machine Tool & Automatic Manufacturing Technique

基　　金：国家自然科学基金(51975130);辽宁省重点研发计划项目(2017225016)。

摘　　要：针对传统特征提取的故障诊断技术不能充分表征振动信号故障特征导致故障识别精度不高的问题,提出一种优化VMD和MHA-DenseNet的滚动轴承故障诊断方法。首先,采用麻雀搜索算法(SSA)对变分模态分解算法(VMD)的相关参数组合进行寻优;其次,采用优化VMD分解滚动轴承故障信号,获得的本征模态函数分量(IMF)作为神经网络输入数据;最后,构建多头注意密集神经网络(MHA-DenseNet)故障诊断模型来有效学习故障数据中的特征信息并完成滚动轴承的准确诊断。实验结果表明,提出的故障诊断方法识别率高达99.03%,相较于对比实验该方法提高了故障诊断的准确率。Aiming at the problem that the fault diagnosis technology of traditional feature extraction can’t fully characterize the fault characteristics of vibration signals and the fault recognition accuracy is not high,an optimized VMD and MHA-DenseNet fault diagnosis method for rolling bearings is proposed.Firstly,the sparrow search algorithm(SSA)is used to optimize the relevant parameter combination of the variational modal decomposition algorithm(VMD);Secondly,the optimized VMD is used to decompose the fault signal of the rolling bearing,and the eigenmode function component(IMF)obtained is used as the nerve network inputs the data;Finally,a multi-head attention dense neural network(MHA-DenseNet)fault diagnosis model is constructed to effectively learn the characteristic information in the fault data and complete the accurate diagnosis of the rolling bearing.Experimental results show that the recognition rate of the proposed fault diagnosis method is as high as 99.03%,which improves the accuracy of fault diagnosis compared with the comparative experiment.

关 键 词：滚动轴承 变分模态分解 麻雀搜索算法 多头注意密集神经网络 故障诊断 

分 类 号：TH133.3[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]