混合注意力机制的旋转机械故障诊断方法  

A fault diagnosis method for rotating machinery based on hybrid attention mechanism

作　　者：付忠广[1] 高玉才 王诗云 谢玉存 翟世臣 FU Zhongguang;GAO Yucai;WANG Shiyun;XIE Yucun;ZHAI Shichen(Key Laboratory of Power Station Energy Transfer Conversion and System of Ministry of Education,North China Electric Power University,Beijing 102206,China;College of Computer Science and Technology,Nanjing University of Aeronautics and Astronautics,Nanjing 210016,Jiangsu,China)

机构地区：[1]华北电力大学电站能量传递转化与系统教育部重点实验室,北京102206 [2]南京航空航天大学计算机科学与技术学院,江苏南京210016

出　　处：《中国工程机械学报》2022年第5期459-464,共6页Chinese Journal of Construction Machinery

基　　金：国家自然科学基金资助项目(50776029,51036002)。

摘　　要：针对旋转机械故障诊断问题,提出一种基于混合注意力机制的旋转机械故障诊断方法。首先将通道注意力模块和空间注意力模块进行串联构成混合注意力模块,并将其加入到LeNet5卷积神经网络中;然后将旋转机械振动信号进行连续小波变换,转换为时频图像;最后通过嵌有混合注意力模块的LeNet5网络对时频图像进行识别从而判定旋转机械故障类型。实验结果表明:提出的方法具有较高的故障识别准确率,能够有效判定旋转机械的运行状态。For the problem of rotating machinery fault diagnosis,this paper proposes a rotating machinery fault diagnosis method based on hybrid attention mechanism.First,the channel attention module and the spatial attention module are concatenated to form a hybrid attention module and added to the LeNet5 convolutional neural network.Then the rotating machine vibration signal is converted into a time-frequency image by continuous wavelet transform.Finally,the LeNet5 network embedded with a hybrid attention module identifies the time-frequency images to determine the type of rotating machine fault.The experimental results show that the method proposed has a high fault identification accuracy and can effectively determine the operation status of rotating machinery.

关 键 词：旋转机械 故障诊断 小波变换 注意力机制 卷积神经网络 

分 类 号：TH17[机械工程—机械制造及自动化] TK14[动力工程及工程热物理—热能工程]