摘要
【目的】为了解决滚动轴承故障特征提取困难、诊断性能偏低的问题,提出了一种基于时频域特征和朴素贝叶斯的故障诊断方法。【方法】首先,通过局部均值分解方法对原始振动信号进行处理,获得多个乘积函数分量。其次,基于原始振动信号和各个乘积函数分量,提取时频域特征,并采用主成分分析实现特征降维,获得低维敏感特征。最后,依据低维敏感特征集,结合朴素贝叶斯模型,实现对江南大学—机械工程学院滚动轴承数据集的分析。【结果】实验结果表明,该方法相较于传统朴素贝叶斯准确率高39.49%,相较于主成分分析准确率高5.94%,由此得出该方法对滚动轴承故障的诊断表现较好。【结论】对于传统的单一的故障诊断模型,基于时频域特征和朴素贝叶斯的故障诊断模型具有更高的准确率,解决了滚动轴承故障特征提取困难、诊断性能偏低的问题。
[Purposes]In order to solve the problems of difficult feature extraction and low diagnostic performance of rolling bearings,a fault diagnosis method based on time-frequency domain features and naive Bayes is proposed.[Methods]This method first processes the original vibration signal through local mean decomposition to obtain multiple product function(PF)components.Secondly,based on the original vibration signal and various PF components,time-frequency domain features are extracted,and principal component analysis is used to achieve feature dimension reduction,obtaining low dimensional sensitive features.Finally,based on the low dimensional sensitive feature set and combined with the naive Bayesian model,the analysis of the rolling bearing dataset from Jiangnan University School of Mechanical Engineering is achieved.[Findings]The experimental results show that the accuracy of this method is 39.49%higher than that of traditional naive Bayes,and 5.94%higher than that of principal component analysis.Therefore,it can be concluded that this method performs well in diagnosing rolling bearing faults.[Conclusions]Compared to traditional single fault diagnosis models,fault diagnosis models based on time-frequency domain features and naive Bayes have higher accuracy and solve the problems of difficult feature extraction and low diagnostic performance in rolling bearing faults.
作者
温翔采
张清华
胡勤
刘迪洋
WEN Xiangcai;ZHANG Qinghua;HU Qin;LIU Diyang(Jilin Institute of Chemical Technology,Jilin 132000,China;Guangdong University of Petrochemical Technology,Maoming 525000,China)
出处
《河南科技》
2024年第7期18-24,共7页
Henan Science and Technology
关键词
滚动轴承
时频域特征
局部均值分解
主成分分析
朴素贝叶斯
rolling bearings
time-frequency domain features
local mean decomposition
principal com⁃ponent analysis
naive bayes