摘要
在滚动轴承故障诊断中,算法难以学习所有负载下的健康状态特征,为有效诊断滚动轴承在变负载下的健康状态,算法需要较强的负载域适应能力。针对上述问题,提出了基于多通道时频域信号的卷积神经网络算法。不同的小波提取不同的特征,算法采用多种小波可以提供多样的健康状态特征。并且全局最大池化替换每一空洞卷积之后的最大池化,从全局范围内提取最大激活。因此,算法只需在源域下训练,即可在目标域下得到良好的诊断效果。为验证该算法的有效性,利用公共数据集进行实验。实验结果表明,该算法在不同负载下的分类精度较其他算法有明显提高,从而可以有效识别滚动轴承的健康状态。
In the fault diagnosis of rolling bearings,it is difficult for the algorithm to learn the features of the health state under all loads.Therefore,the algorithm needs strong load domain adaptability to effectively diagnose the health state of rolling bearings under varying load.To solve the above problem,a convolutional neural network based on multi-channel time-frequency signals was proposed.Different wavelets extract different features,and the algorithm used multiple wavelets to provide various health state features.Max pooling was replaced global max pooling after each dilated convolution to extract the max activation from the global scope.The algorithm only needed to be trained in the source domain to get a good diagnosis effect in the target domain.Experiments were carried out using the public data set to verify the effectiveness of the algorithm.Experimental results show that the classification accuracy of this algorithm is obviously improved compared with other algorithms under varying load,which can effectively recognize the health state of rolling bearings.
作者
孙仕鑫
高洁
王伟
杜劲松
杨旭
SUN Shi-xin;GAO Jie;WANG Wei;DU Jin-song;YANG Xu(Shenyang Institute of Automation, Chinese Academy of Sciences, Shenyang 110016, China;Institutes for Robotics and Intelligent Manufacturing, Chinese Academy of Sciences, Shenyang 110169, China;University of Chinese Academy of Sciences, Beijing 100049, China;Key Laboratory on Intelligent Detection and Equipment Technology of Liaoning Province, Shenyang 110179, China)
出处
《科学技术与工程》
北大核心
2021年第15期6386-6393,共8页
Science Technology and Engineering
基金
中国科学院战略性先导科技专项(C类)(XDC04030200)
工信部智能制造综合标准化与新模式应用项目(2018-15)
中国科学院STS项目(KFJ-STS-QYZD-107)。
关键词
负载域适应能力
空洞卷积
全局最大池化
多通道时频域信号
load domain adaptability
dilated convolution
global max pooling
multi-channel time-frequency signals
