基于VMD样本熵和改进极限学习机的滚动轴承故障诊断  

Bearing failure diagnosis based on VMD sample entropy and improved extreme learning machine

作　　者：封成东 李玥[1] 封成智 FENG Chengdong;LI Yue;FENG Chengzhi(College of Information Science and Technology,Gansu Agricultural University,Lanzhou 730000,China;Gansu Productivity Promotion Center,Lanzhou 730000,China)

机构地区：[1]甘肃农业大学信息科学技术学院,甘肃兰州730000 [2]甘肃省生产力促进中心,甘肃兰州730000

出　　处：《甘肃农业大学学报》2023年第2期215-225,共11页Journal of Gansu Agricultural University

基　　金：国家自然科学基金项目(32060437,31360315);甘肃农业大学青年导师基金项目(GAU-QDFC-2020-12)。

摘　　要：【目的】为了更好地提取滚动轴承故障信号特征并提高故障的分类准确率,提出了基于变分模态分解(VMD)、样本熵和改进极限学习机(ELM)的滚动轴承故障诊断方法。【方法】首先,采用VMD算法对轴承信号进行分解,通过中心频率和频率混叠确定VMD分解层数,提取VMD分解后各模态分量的样本熵并构建特征样本;然后,建立ELM的诊断模型,引入麻雀搜索算法(SSA)和改进的麻雀搜索算法(ISSA)分别优化ELM输入层与隐含层神经元之间的连接权值及隐含层神经元的阈值;最后,将样本熵特征样本输入诊断模型进行滚动轴承的故障识别。【结果】仿真结果表明,ELM、SSA-ELM和ISSA-ELM 3种诊断模型的准确率平均值分别为93.33%、95.83%、97.50%,表明提出的优化模型ISSA-ELM相对ELM和SSA-ELM的诊断精度更高、泛化性能更优,对滚动轴承的故障分类效果较好。【结论】基于VMD样本熵和改进ELM的诊断方法不仅能有效地提取故障特征,而且实现较高准确率的故障识别效果。【Objective】In order to better extract signal features and improve the accuracy of fault classification,a rolling bearing fault diagnosis method based on Variational Mode Decomposition(VMD),Sample Entropy and an improved Extreme Learning Machine(ELM)has been proposed in this paper.【Method】First,the VMD algorithm was used to decompose the bearing signal,and the number of decomposition layers of VMD was determined by the center frequency and frequency aliasing,then the sample entropy of each modal component was extracted after VMD decomposition,and the characteristic samples were constructed.Secondly,the diagnosis model of ELM was established,and the Sparrow Search Algorithm(SSA)and the Improved Sparrow Search Algorithm(ISSA)were respectively introduced to optimize the connection weights between the ELM input layer and the hidden layer neurons and the threshold of the hidden layer neurons.Finally,the sample entropy charactristic samples were input into the diagnosis model to identify the rolling bearing fault.【Result】Simulation results show that the average accuracy of the three diagnostic models,ELM,SSA-ELM and ISSA-ELM were 93.33%,95.83%and 97.50%,respectively,showing that the optimized model ISSA-ELM proposed in this study has higher diagnostic accuracy than ELM and SSA-ELM.The generalization performance is better,and the fault classification effect of the rolling bearing is better.【Conclusion】The diagnosis method based on VMD sample entropy and improved ELM can not only effectively extract the fault features,but also achieve the high accuracy of fault dectation.

关 键 词：滚动轴承 VMD 样本熵 SSA ELM 

分 类 号：TH133.3[机械工程—机械制造及自动化] TH165