迁移学习在轴承保持架故障诊断中的研究  

Research on Transfer Learning in Fault Diagnosis of Bearing Cage

作　　者：李继光 苏燕辰[1] 任继炜 李艳萍[1] LI Ji-guang;SU Yan-chen;REN Ji-wei;LI Yan-ping(School of Mechanical Engineering,Southwest Jiaotong University,Sichuna Chengdu 610031,China)

机构地区：[1]西南交通大学机械工程学院,四川成都610031

出　　处：《机械设计与制造》2023年第11期17-20,共4页Machinery Design & Manufacture

摘　　要：针对滚动轴承保持架由于故障频率太小容易被噪声干扰,振动分析等传统故障检测方法检测困难,特征较难提取的问题,提出一种基于迁移学习的轴承保持架的故障诊断算法。利用数据量较多但缺少保持架故障相关数据的凯斯西储大学的轴承振动加速度数据集进行模型训练,提取出重要的模型参数信息,然后在此基础上,利用此模型参数在少量的齿轮箱轴承保持架振动加速度数据上进行迁移学习,实现对齿轮箱轴承保持架的故障识别,实验表明该迁移学习方法在识别滚动轴承保持架上是有效的。Because the fault frequency of rolling bearing cage is too small,it is easy to be disturbed by noise.Traditional fault detection methods such as vibration analysis are difficult to detect and feature extraction is difficult.A fault diagnosis algorithm of bearing cage based on transfer learning is proposed.The bearing data set of Case Western Reserve University has a large amount of data but lacks cage fault related data,is used for model training,and important model parameter information is extracted.On this basis,the model parameters are used to transfer learning on a small number of cage data to realize fault identification of rolling bearing cage.The experimental results show that the transfer learning method is effective in identifying the cage of rolling bearing.

关 键 词：故障诊断 保持架 卷积神经网络 长短时记忆网络 迁移学习 

分 类 号：TH16[机械工程—机械制造及自动化] U270[机械工程—车辆工程] TH133.3[交通运输工程—载运工具运用工程]