摘要
针对传统的故障诊断方法依赖经验进行人工提取特征的问题,基于变分模态分解(VMD)和平滑伪魏格纳分布(SPWVD)并结合卷积神经网络(CNN)提出了VMD-SPWVD-CNN模型用于滚动轴承故障诊断。首先,利用VMD对轴承振动信号进行处理,分解为多个具有不同中心频率的模态分量;其次,对信号的每个模态分量分别进行SPWVD计算,将每个分量的计算结果累加得到轴承信号的二维时频图;最后,将时频图作为ResNet18卷积神经网络的输入,自动提取图像的深层特征完成滚动轴承的故障诊断。对10类轴承故障进行多次故障识别,该方法的平均准确率提升至99.56%,能有效地完成滚动轴承故障类别以及损伤程度的精确判定。
Aiming at the problem that traditional fault diagnosis methods rely on experience to extract features manually, this paper proposed a VMD-SPWVD-CNN model for rolling bearing fault diagnosis based on variational mode decomposition(VMD) and smooth pseudo-Wegener distribution(SPWVD) combined with convolutional neural network(CNN).Firstly, VMD was used to process the bearing vibration signals, which were decomposed into several modal components with different center frequencies.Secondly, SPWVD calculation was carried out for each modal component of the signal, and the calculation results of each component were accumulated to obtain the two-dimensional time-frequency diagram of the bearing signal.Finally, the time-frequency graph is used as the input of ResNet18 convolutional neural network to automatically extract the deep features of the image to complete the fault diagnosis of rolling bearings.For 10 kinds of bearing faults, the average accuracy of this method is improved to 99.56%,which can effectively determine the classification and damage degree of rolling bearing faults.
作者
刘世林
陈里里
LIU Shi-lin;CHEN Li-li(School of Mechatronics and Vehicle Engineering,Chongqing Jiaotong University,Chongqing 400074,China)
出处
《组合机床与自动化加工技术》
北大核心
2022年第4期62-65,69,共5页
Modular Machine Tool & Automatic Manufacturing Technique
基金
重庆市重点实验室开放课题(CKLURTSIC-KFKT-202006)
重庆市工程实验室2020年度开放课题(CELTEAR-KFKT-202003)
重庆市研究生教育教学改革项目(yjgl182027)。
