摘要
针对传统方法识别变速器制造故障的局限性,引入深层神经网络进行故障识别。在使用深层神经网络进行故障识别时,故障数据的数量、特征多样性和特征吻合度等都对识别结果产生较大影响,因此对变速器故障数据集进行研究。首先,通过分析选择时频域统计分析法和转速谱阵图分析法;然后,基于变速器的典型制造故障数据建立3种形式的故障数据集,每种数据集包含6类标签;最后,搭建深层卷积神经网络和深度信念网络模型,利用模型进行故障数据集的分类识别。实验结果表明:建立的制造故障数据集可以较好地代表变速器制造故障。
In view of the limitations of traditional methods to identify transmission manufacturing faults,this paper introduces deep neural network for fault identification.However,the quantity,representation and diversity of the data have a great impact on the network prediction results.Therefore,the transmission fault data sets need to be studied.Firstly,we analyze and select timefrequency domain statistical analysis method and rotational speed spectrum analysis method.Next,based on the typical manufacturing failure data,three types of fault data sets are established,each of which contains six categories.Through subjective evaluation,the established data setsare consistent with the real situation.Finally,deep convolutional neural network and deep belief network are used to classify and identify fault data sets.The experimental results show that the established fault data sets can better represent the fault problem of the transmission online detection.
作者
施全
贾书曼
陈绮丹
陈婷
于中喜
刘骄
SHI Quan;JIA Shuman;CHEN Qidan;CHEN Ting;YU Zhongxi;LIU Jiao(Vehicle Engineering Institude,Chongqing University of Technology,Chongqing 400054,China)
出处
《重庆理工大学学报（自然科学）》
CAS
北大核心
2020年第10期74-82,共9页
Journal of Chongqing University of Technology：Natural Science
基金
节能与新能源汽车动力传动系统关键试验技术及装备(cstc2017zdcy-zdzxX0008)
重庆市研究生创新基金项目(CYS18302)。
关键词
变速器
数据集
故障识别
非稳态振动
transmission
data set
fault identification
unsteady vibration
