摘要
滚动轴承是旋转机械最主要的零部件之一,针对滚动轴承故障类型的有效识别问题,提出了一种基于卷积神经网络(CNN)和极端梯度提升(XGBoost)的滚动轴承故障诊断方法。对滚动轴承数据进行预处理,利用训练集对CNN进行训练和调参,运用训练好的CNN模型进行特征的提取,并使用XGBoost模型进行故障类型的分类。采用德国帕德博恩大学滚动轴承进行模型的测试,结果表明:该模型交叉验证得分接近满分,并且在测试集上准确率达到了99%,优于仅仅使用CNN、支持向量机(SVM)、XGBoost以及三层神经网络。
Rolling bearing is one of the most important parts of rotating machinery.Aiming at the problem of identifying the fault type of rolling bearing effectively,a fault diagnosis method based on convolutional neural networks(CNN)and extreme gradient boosting(XGBoost)is proposed.Firstly,the data of the rolling bearing is preprocessed,the CNN is trained and adjusted by using the training set,and the features are extracted by using the trained CNN model.Finally,the fault type is classified by using the XGBoost model.The model is tested using rolling bearings from the university of Paderborn in Germany,and the results show that the model scores close to full marks for crossvalidation and achieved 99%accuracy in the test set,which is better than just using CNN,support vector machine(SVM),XGBoost and a three-layer neural network.
作者
马怀祥
冯旭威
李东升
齐澍椿
MA Huaixiang;FENG Xuwei;LI Dongsheng;QI Shuchun(School of Mechanical Engineering,Shijiazhuang Tiedao University,Shyiazhuang 050043,Hebei,China;Wuhu Yangtze River Tunnel Construction Headquarters,China Railway No.14 Bureau group Co.,Ltd.,Wuhu 241000,Anhui,China)
出处
《中国工程机械学报》
北大核心
2021年第3期254-259,共6页
Chinese Journal of Construction Machinery
关键词
卷积神经网络
极端梯度提升
故障诊断
滚动轴承
convolutional neural networks(CNN)
eXtreme gradient boosting(XGBoost)
fault diagnosis
rolling bearing