基于SDP和改进SAM⁃MobileNetv2的滚动轴承故障诊断方法研究  

FAULT DIAGNOSIS OF ROLLING BEARINGS BASED ON SDP AND IMPROVED SAM⁃MobileNetv2

作　　者：张天缘 孙虎儿[1] 朱继扬 赵扬 ZHANG TianYuan;SUN HuEr;ZHU JiYang;ZHAO Yang(School of Mechanical Engineering,North University of China,Taiyuan 030051,China)

机构地区：[1]中北大学机械工程学院,太原030051

出　　处：《机械强度》2024年第4期787-794,共8页Journal of Mechanical Strength

基　　金：山西省自然科学基金项目(201801D121186)资助。

摘　　要：针对传统的滚动轴承故障诊断方法难以准确高效的实现故障分类,提出了一种融合对称点模式(Symmetrized Dot Pattern,SDP)和改进SAM⁃MobileNetv2的滚动轴承故障分类方法。首先,将轴承振动信号通过SDP算法转化为含有丰富特征信息的二维图像。然后,将二维图像输入到改进SAM⁃MobileNetv2网络模型中,对故障特征信息进行提取和分类。在改进SAM⁃MobileNetv2网络中,使用自适应激活函数ACON(Activate or not)对SAM⁃MobileNetv2中的ReLU6激活函数进行替换,提高模型分类性能。最后,将本模型与多种网络模型做对比。试验结果表明,本模型可以准确高效地实现对滚动轴承故障的分类,使用凯斯西储大学轴承故障数据的准确率为99.5%,使用渥太华大学轴承故障数据的准确率为97.2%。Traditional fault diagnosis methods for rolling bearings are difficult to accurately and efficiently achieve fault classification.A method of rolling bearing fault classification based on symmetrized dot pattern(SDP)and improved SAM⁃MobileNetv2 was proposed.Firstly,the bearing vibration signal was transformed into two⁃dimensional images with rich characteristic information by SDP algorithm.Secondly,the two⁃dimensional images were fed into the SAM⁃MobileNetv2 network model,which extracted and classified fault feature information.Improved SAM⁃MobileNetv2 networks used the adaptive activation function ACON to replace the ReLU6 activation function in SAM⁃MobileNetv2 to improve model classification performance.Finally,this model was compared with various models.The experimental results show that this model can accurately and efficiently realize the classification of rolling bearing faults,using Case Western Reserve University bearing fault data with an accuracy rate of 99.5%,using the University of Ottawa bearing failure data with an accuracy rate of 97.2%.

关 键 词：滚动轴承 对称点模式 SAM⁃MobileNetv2模型 故障诊断 

分 类 号：TH165.3[机械工程—机械制造及自动化] TH17