摘要
针对电动后视镜驱动器振动信号非平稳非线性及信噪比低,提取传统特征难以有效识别样本故障状态的问题,提出了一种改进的集成经验模态分解算法(EEMD)。使用EEMD对振动信号进行了分解,利用相关系数与峭度系数筛选有效本征模态函数(IMF)分量。应用自回归模型(AR)功率谱估计方法,建立最佳阶次的AR模型,对有效IMF分量进行谱估计,并得到有效IMF分量的AR谱与AR累加谱。将AR累加谱的特征频率点与振幅作为特征向量,使用支持向量机(SVM)进行机器学习与分类。研究结果表明:EEMD-AR-SVM模型在实验中的分类准确率达到了93.9%,平均耗时46.1 s,达到了工业中自动检测的标准。
Aiming at the problem that vibration signals of the electric rearview mirror driver were nonstationary,nonlinear,low signal noise ratio(SNR),and difficult to effectively identify the fault state of the sample,an improved ensemble empirical mode decomposition(EEMD)algorithm was proposed.Taking the driver of electric rear-view mirrors for motor vehicle as the research object,EEMD was employed to decompose its vibration signal.Correlation coefficient and kurtosis index were used to filter the effective intrinsic mode function(IMF)components.Autoregressive(AR)model power spectrum estimation was used to establish the best order model,and AR spectrum and AR accumulated spectrum of the effective IMF components were obtained.The eigenfrequency and amplitude of the AR accumulation spectrum were treated as eigen vector and used to train support vector machines for classification.The results show that EEMD-AR-SVM model achieves a classification accuracy of 93.9%in the experiment,and takes an average time of 46.1 s,which reaches to the standard of automatic detection in the industry.
作者
高丰
朱少成
罗石
GAO Feng;ZHU Shaocheng;LUO Shi(School of Automotive&Traffic Engineering,Jiangsu University,Zhenjiang 212013,Jiangsu,China)
出处
《河南科技大学学报（自然科学版）》
CAS
北大核心
2021年第6期39-45,M0004,M0005,共9页
Journal of Henan University of Science And Technology:Natural Science
基金
国家重点研发计划基金项目(2019YFB1600500)。
关键词
机械故障诊断
集成经验模态分解算法
IMF筛选
AR功率谱估计
支持向量机
mechanical fault diagnosis
ensemble empirical mode decomposition
IMF filtering
AR spectrum estimation
support vector machines
