摘要
针对当前滚动轴承剩余使用寿命(remaining useful life,RUL)预测准确率不高且未考虑到预测实时性的问题,提出一种云边协同计算模式下(cloud-edge collaborative computing,CECC)双数据源融合(data sources fusion,DSF)的滚动轴承剩余使用寿命实时预测方法。首先在离线阶段通过专家先验知识对训练集数据进行分析并进行网络预训练,然后通过边缘设备实时采集轴承水平与垂直两种数据源的振动信号并构建测试集,最后测试集数据实时上传到强大计算能力的云端进行融合预测。方法引入并行计算的Transformer模型在PHM2012数据集上进行试验,结果表明云边协同计算模式下轴承RUL预测的实时性得到显著提高,同时DSF预测方法与单一数据源预测方法相比MAE与RMSE两项指标分别降低了42.1%和40.9%。在XJTU-SY数据集上验证结果表明,DSF预测方法与其他文献中单一数据源预测方法相比MAE与RMSE分别降低了38.1%与38.8%;且云边协同预测方法相较于云计算预测,其时效性提升了80%左右,进一步证明了方法的可行性,并为其他领域寿命实时预测提供了解决方案。
In view of low accuracy and the lack of real-time in the RUL prediction of the rolling bearings,this paper proposes a real-time prediction method for RUL of rolling bearings based on the fusion of dual data sources in the cloud-edge collaborative computing mode.Firstly,this method analyzed the training set through the expert prior knowledge and performed network pre-training in the offline stage.Then,the vibration signals of the horizontal and vertical data sources of the bearing were collected by the edge device in real-time to construct the test sets.Finally,the test sets were uploaded to the cloud in real time for fusion prediction.The Transformer model that can be computed in parallel was applied on the PHM2012 dataset.The results show that the real-time performance of bearing RUL prediction is significantly improved under the cloud-edge collaborative computing mode.At the same time,compared with the single data source prediction method,the DSF prediction method reduces the MAE and RMSE by 42.1%and 40.9%respectively.Compared with the single data source prediction method in other literatures,the verification results on the XJTU-SY dataset show that the DSF prediction method reduces MAE and RMSE by 38.1%and 38.8%,respectively.Moreover,compared with cloud computing prediction,the cloud-edge collaborative prediction method improves the timeliness by about 80%,which further proves the feasibility of the method and provides a solution for real-time prediction in other fields.
作者
刘汝迪
唐向红
陆见光
刘方杰
柳鹏飞
LIU Rudi;TANG Xianghong;LU Jianguang;LIU Fangjie;LIU Pengfei(Key Laboratory of Modern Manufacturing Technology of the Ministry of Education,Guizhou University,Guiyang 550025,China;State Key Laboratory of Public Big Data,Guizhou University,Guiyang 550025,China;Chongqing Industrial Big Data Innovation Center Co.,Ltd.,Chongqing 400707,China;不详)
出处
《组合机床与自动化加工技术》
北大核心
2023年第11期89-94,共6页
Modular Machine Tool & Automatic Manufacturing Technique
基金
贵州省科技计划项目(黔科合基础ZK[2021]一般271/QKHJC-ZK〔2021〕YB271)
贵州省科技支撑计划项目(黔科合支撑[2022]一般074/QKHZC〔2022〕YB074)。
关键词
滚动轴承
剩余使用寿命
云边协同计算
双数据源融合
实时预测
rolling bearings
remaining useful life
cloud-edge collaboration
dual data sources fusion
real-time prediction
作者简介
刘汝迪(1998-),男,硕士研究生,研究方向为故障诊断与寿命预测,(E-mail)gs.rdliu20@gzu.edu.cn;通信作者:唐向红(1979-),男,教授,博士生导师,博士,研究方向故障诊断、图神经网络,(E-mail)xhtang@gzu.edu.cn。
