基于改进CNN-BiGRU的电机轴承故障识别  被引量:2

Motor Bearing Fault Identification Based on Improved CNN-BiGRU

在线阅读下载全文

作  者:陈玉球[1] CHEN Yu-qiu(Department of Mechanical and Electrical Engineering,Hunan Nonferrous Metals Vocational and Technical College,Zhuzhou 412007,China)

机构地区:[1]湖南有色金属职业技术学院机电工程系,株洲412007

出  处:《组合机床与自动化加工技术》2022年第7期75-80,共6页Modular Machine Tool & Automatic Manufacturing Technique

基  金:2020年度湖南省教育厅科学研究一般项目(20C1373)。

摘  要:针对基于深度学习的电机轴承故障识别方法易受环境噪声干扰的问题,提出了一种改进卷积神经网络(convolutional neural network, CNN)双向门控循环单元(bidirectional gated recurrent unit, BiGRU)的电机轴承故障识别方法。首先,使用CNN和BiGRU提取电机轴承故障振动信号的空间和时间特征;其次,引入动态选择和自注意力机制,依据不同轴承的故障状态自适应定位相关特征信息,实现故障特征精准有效提取;最后,利用t分布随机近邻嵌入方法,将动态选择和自注意力机制层的特征信息段可视化,进一步提高网络模型的可解释性。试验结果表明,改进CNN-BiGRU网络模型可以有效地对轴承的不同故障类型和故障程度进行识别,在不同背景噪声干扰下的特征学习能力和故障识别准确率显著优于其他典型的深度学习模型。Deep learning-based rolling bearing fault identification methods were vulnerable to environmental noise,a method based on improved convolutional neural network(CNN)with bidirectional gated recurrent unit(BiGRU)was proposed.Firstly,the CNN and BiGRU were used to extract the spatial and temporal characteristics in the vibration signal during the motor bearing failure.Secondly,the dynamic selection and self-attention mechanism were introduced to realize accurate and effective fault feature extraction based on relevant feature information of different bearing fault states.Finally,using t-distribution stochastic neighbor embedding,the feature information segment of dynamic selection and self-attention mechanism layer was visualized to further improve the interpretability of the model.The experimental results show that the improved CNN-BiGRU network model can effectively identify different fault types and fault degrees of bearings,and the feature learning ability and fault identification accuracy under different background noise interference are significantly better than other typical deep learning models.

关 键 词:电机轴承 故障识别 卷积神经网络 双向门控循环单元 

分 类 号:TH133.3[机械工程—机械制造及自动化] TG66[金属学及工艺—金属切削加工及机床]

 

参考文献:

正在载入数据...

 

二级参考文献:

正在载入数据...

 

耦合文献:

正在载入数据...

 

引证文献:

正在载入数据...

 

二级引证文献:

正在载入数据...

 

同被引文献:

正在载入数据...

 

相关期刊文献:

正在载入数据...

相关的主题
相关的作者对象
相关的机构对象