基于MTF和AM-MSCNN的滚动轴承故障诊断方法  

Fault Diagnosis Method for Rolling Bearings Based on MTF and AM-MSCNN

作　　者：范佳鹏 陈曦晖 李勇 陈志帮 邢子豪 FAN Jiapeng;CHEN Xihui;LI Yong;CHEN Zhibang;XING Zihao(College of Mechanical and Electrical Engineering,Hohai University,Changzhou 213200,China;School of Mechanical and Electrical Engineering,China University of Mining and Technology,Xuzhou 221116,China)

机构地区：[1]河海大学机电工程学院,江苏常州213200 [2]中国矿业大学机电工程学院,江苏徐州221116

出　　处：《轴承》2024年第12期74-79,共6页Bearing

基　　金：国家自然科学基金资助项目(52204178,51905147);常州市应用基础研究计划资助项目(CJ20220208)。

摘　　要：针对传动系统中多振源耦合导致滚动轴承故障特征被噪声和其他故障特征掩盖的问题,提出基于马尔可夫转移场(MTF)和含有注意力机制的多尺度卷积神经网络(AM-MSCNN)的滚动轴承故障诊断方法。首先,利用MTF将原始一维振动信号转换为具有时间相关性的二维特征图像,保留完整的时间信息,更好地反映轴承故障特征;然后,搭建AM-MSCNN模型,有效提取全局特征并自适应调整权重;最后,将MTF提取的二维特征图输入AM-MSCNN进行特征提取和故障诊断。搭建滚动轴承故障模拟试验台对模型的有效性进行验证,结果表明所提模型的故障识别率达到95%以上,高于MSCNN和CNN的故障识别率。Aimed at the problem that the fault features of rolling bearings are masked by noise and other fault features due to multi-vibration source coupling in transmission system,a fault diagnosis method for rolling bearings is proposed based on Markov transition field(MTF)and multi-scale convolutional neural network with attention mechanism(AM-MSCNN).Firstly,MTF is used to convert the original one-dimensional vibration signal into a two-dimensional feature image with temporal correlation,preserving complete temporal information and better reflecting the bearing fault features.Then,the AM-MSCNN model is built to extract the global features effectively and adjust the weights adaptively.Finally,the two-dimensional feature maps extracted by MTF are input into AM-MSCNN for feature extraction and fault diagnosis.A fault simulation test bench for rolling bearings is built to verify the validity of the model.The results show that the fault recognition rate of the proposed model is above 95%,which is higher than that of MSCNN and CNN.

关 键 词：滚动轴承 故障诊断 特征提取 信号处理 马尔可夫转移场 注意力机制 卷积神经网络 

分 类 号：TH133.33[机械工程—机械制造及自动化]