摘要
预测轴承的剩余使用寿命虽然已经引入了许多数据驱动的方法,但很少有研究考虑不同传感器之间的时空相关性,这种相关性可以用来识别不同传感器之间的特征,以提高预测模型的鲁棒性。因此,针对多传感器的复杂性特征,构建了一种动态加权图卷积网络(DW-GCN),进行轴承的剩余寿命预测。首先利用不同节点间的空间相关性和时序相关性构建动态加权图,其次利用GCN对样本关联图的空间特征进行提取,最后将提取的特征输入到预测网络中进行剩余使用寿命预测。对轴承数据集进行测试,DW-GCN方法能够有效预测轴承剩余使用寿命,具有较高的预测精度。
Although many data-driven methods for predicting the remaining useful life of bearings have been introduced,few studies have considered the spatio-temporal correlation between different sensors,which can be used to identify features between different sensors to improve the robustness of the predictive model.Therefore,a dynamically weighted graph convolutional network(DW-GCN)was constructed to predict the residual service life of bearings for the complexity of multi-sensor.Firstly,the dynamic weighted graph was constructed by using the spatial and temporal correlations between different nodes.Secondly,the GCN was used to extract the spatial features of the sample correlation graph.Finally,the extracted features were input into the prediction network to predict the remaining service life.Tested on the bearing data set,the DW-GCN method can effectively predict the remaining service life of bearing,and the prediction accuracy is higher.
作者
张洪滔
赵舜
李雅婧
江鹏伟
Zhang Hongtao;Zhao Shun;Li Yajing;Jiang Pengwei(School of Mechatronic Engineering,North University of China,Taiyuan 030051,China)
出处
《煤矿机械》
2024年第3期163-165,共3页
Coal Mine Machinery
基金
国家自然科学基金项目(51905496)
山西省自然科学基金项目(201801D221237)
中北大学研究生科技立项(20221813)。
关键词
GCN
轴承剩余寿命预测
动态加权图
GCN
prediction of bearing remaining useful life
dynamic weighted grap
