基于超声解调信号多特征融合的轴承故障识别  

Bearing fault recognition based on multi-feature fusion of ultrasonic demodulation signals

作　　者：姜浪朗 张敬超 江国乾[1] 苏连成[1] 李英伟[2] JIANG Langlang;ZHANG Jingchao;JIANG Guoqian;SU Liancheng;LI Yingwei(School of Electrical Engineering,Yanshan University,Qinhuangdao,Hebei 066004,China;School of Information Science and Engineering,Yanshan University,Qinhuangdao,Hebei 066004,China)

机构地区：[1]燕山大学电气工程学院,河北秦皇岛066004 [2]燕山大学信息科学与工程学院,河北秦皇岛066004

出　　处：《燕山大学学报》2022年第6期484-491,560,共9页Journal of Yanshan University

基　　金：国家自然科学基金资助项目(61827811);国防基础研究计划资助项目(JCKY2019407C002)。

摘　　要：针对当前振动监测对轴承初期微损状态监测难、精准度低的问题,提出了一种基于经验模态分解与多类熵值相结合的轴承状态监测方法,研究超声解调信号对故障诊断的可行性。首先将预处理后的超声解调信号进行经验模态分解得到若干本征模态分量,然后对各本征模态分量计算不同熵值特征,再将多特征融合后代入随机森林训练分类模型,利用混淆矩阵进行精度评价,最终对早期故障识别准确率高达97.92%。研究表明,超声解调信号对判别轴承早期故障效果更佳;经过多类熵值特征融合后,轴承状态分类具有更高的识别准确率。Aiming at the current problems of difficulty and low accuracy in monitoring the initial micro-damage condition of bearings by vibration monitoring,a bearing condition monitoring method based on the fusion of empirical modal decomposition and multi-class entropy values is proposed to study the feasibility of ultrasonic demodulation signals for fault diagnosis.Firstly,the pre-processed ultrasonic demodulation signal is empirically modal decomposed to obtain several eigenmodal components.Secondly,different entropy features are calculated for each eigenmodal component,followed by fusing multiple features and bringing them into the random forest to train the classification model,and using the confusion matrix to evaluate the accuracy.Finally,the accuracy of early fault identification is as high as 97.92%.It is shown that the ultrasonic demodulation signal is more effective in discriminating early bearing faults.After the fusion of multi-class entropy features,the bearing state classification has higher recognition accuracy.

关 键 词：轴承 状态监测 经验模态分解 超声解调信号 多特征融合 随机森林 

分 类 号：TH133.33[机械工程—机械制造及自动化]