摘要
针对内燃机轴承故障诊断中精度较低问题,提出基于EMD转速波动信号特征提取的内燃机轴承故障诊断研究。引入经验模态分解(EMD)技术将复杂的转速波动信号分解为多个本征模态函数(IMF)分量,基于本征模态函数(IMF)分量能量比变异系数,优选转速波动信号特征。利用卷积神经网络(CNN)对优选的转速波动信号特征深度挖掘,识别诊断内燃机轴承故障,实现基于EMD转速波动信号特征提取的内燃机轴承故障诊断。经实验证明,设计方法错诊比例和漏诊比例均为超过1%,可以实现对内燃机轴承故障精准诊断。
In response to the problem of low accuracy in the diagnosis of internal combustion engine bearing faults,a research on internal combustion engine bearing fault diagnosis based on EMD speed fluctuation signal feature extraction is proposed.Introducing Empirical Mode Decomposition(EMD)technology to decompose complex speed fluctuation signals into multiple Intrinsic Mode Function(IMF)components,and based on the energy ratio coefficient of IMF components,optimizing the characteristics of speed fluctuation signals.Using Convolutional Neural Networks(CNN)to deeply mine the selected speed fluctuation signal features,identify and diagnose internal combustion engine bearing faults,and achieve internal combustion engine bearing fault diagnosis based on EMD speed fluctuation signal feature extraction.Experimental results have shown that the design method has a misdiagnosis rate and missed diagnosis rate of over 1%,which can achieve accurate diagnosis of internal combustion engine bearing faults.
作者
杨光
Yang Guang(Geely Automobile Research Institute,Zhejiang Ningbo 315336)
出处
《内燃机与配件》
2025年第14期61-63,共3页
Internal Combustion Engine & Parts
关键词
EMD
波动信号
轴承故障
诊断
本征模态函数
卷积神经网络
EMD
Fluctuation signal
Bearing malfunction
Diagnosis
Intrinsic mode function
Convolutional neural network
作者简介
杨光(1983-),男,辽宁沈阳人,汉族,本科,中级工程师,吉利汽车大项目组组长,研究方向:发动机设计及开发。
