基于XGBoost-LSTM的数控机床主轴轴承故障预测方法研究  

Research on fault prediction method of CNC machine tool spindle bearing based on XGBoost-LSTM

作　　者：赵恒喆 杨晓英[1,2] 石岩 杨逢海 杨欣 ZHAO Hengzhe;YANG Xiaoying;SHI Yan;YANG Fenghai;YANG Xin(School of Mechanical Engineering,Henan University of Science and Technology,Luoyang 471003,China;Henan Collaborative Innovation Center of Advanced Manufacturing of Machinery and Equipment,Luoyang 471003,China;School of Business,Henan University of Science and Technology,Luoyang 471023,China)

机构地区：[1]河南科技大学机电工程学院,洛阳471003 [2]机械装备先进制造河南省协同创新中心,洛阳471003 [3]河南科技大学商学院,洛阳471023

出　　处：《现代制造工程》2023年第8期155-160,共6页Modern Manufacturing Engineering

基　　金：山东省重点研发计划资助项目(2020CXGCO11001)。

摘　　要：为实现数控机床主轴轴承的故障准确预测,提出一种将灰色关联度分析法、深度学习和残差滑动窗口分析相结合的故障预测方法。采取灰色关联度分析法对采集的设备状态变量进行特征筛选,在此基础上建立基于极端梯度提升树(eXtreme Gradient Boosting,XGBoost)和长短期记忆(Long Short-Term Memory,LSTM)神经网络(XGBoost-LSTM)加权融合的轴承温度预测模型,通过设定报警阈值和规则,利用滑动窗口法对轴承温度预测模型的预测残差进行分析,实现对主轴轴承故障的准确预测,并通过实例验证了该方法的有效性。In order to realize accurate fault prediction of spindle bearings of CNC machine tools,a fault prediction method combining grey relational analysis,deep learning and sliding residual window analysis was proposed.This method adopts grey relational analysis method to screen the features of the collected equipment state variables.On this basis,the bearing temperature prediction model based on eXtreme Gradient Boosting(XGBoost)and Long Short-Term Memory(LSTM)neural network weighted fusion was established.By setting the alarm threshold and rules,and the prediction residual of the bearing temperature prediction model was analyzed using the sliding window method,the accurate prediction of the spindle bearing fault was realized.The effectiveness of this method was verified by an example.

关 键 词：数控机床 主轴轴承 灰色关联度分析 XGBoost-LSTM 故障预测 

分 类 号：TH133.33[机械工程—机械制造及自动化]