全矢KPCA和AR模型结合的滚动轴承故障预测方法  

Research on Rolling Bearing Fault Prediction Based on Full Vector KPCA and AR Model

作　　者：高亚娟 陈磊[1] 林辉翼 韩捷[1] GAO Ya-juan;CHEN Lei;LIN Hui-yi;HAN Jie(Research Institute of Vibration Engineering,Zhengzhou University,He’nan Zhengzhou450001,China)

机构地区：[1]郑州大学振动工程研究所

出　　处：《机械设计与制造》2019年第11期20-24,共5页Machinery Design & Manufacture

基　　金：河南省教育厅科学技术研究重点项目指导计划(13B603970.0);河南省高校重点学科开放实验室项目(PMTE201302A)

摘　　要：由于单一传感器获取的振动信号具有片面性,采用全矢谱信息融合技术对滚动轴承信号进行特征提取,并与KPCA模型和AR时序预测方法相结合进行故障预测。首先,采用全矢谱技术提取实验数据中的特征主振矢;然后,采用KPCA方法对得到的特征主振矢进行融合,消除数据冗余,并建立全矢KPCA监控模型;最后,将测试样本输入全矢KPCA监控模型并输出T2和SPE统计量,将其值作为AR预测模型的输入,预测其变化情况,并根据其预测值超出KPCA监控模型的控制限与否来判断设备是否出现故障。实验结果表明,该方法既能较好地预测出滚动轴承的运行状态,又能进一步追踪故障发展趋势,为及时做好维修措施提供理论依据。Due to the fact thatthe vibration signal obtained by a single sensorhas one-sidedness,this paperuses full vector spectrum information fusion technology to extract the vibration feature of rolling bearing signal,and combines KPCA model and AR time series prediction method for fault prediction.Firstly,the main vector spectrum of the experimental data is extracted by using the full spectrum technique;then,the KPCA method is used to fuse the characteristic main vector,eliminate redundant data and establish the full vector KPCA monitoring model;finally,inputs the test samples to full vector KPCA monitoring model to obtain T2 and SPE statistics,and their values are used as the input of the AR forecasting model to predictthe operating status of rolling bearing.And determine whether the device is in fault statebased on whether the predicted value exceeds the control limit of the KPCA monitoring model.The experimental result shows that the method can not only predict the running state of the rolling bearing well,but also further track the development trend of the fault,and provide the theoretical basis for the timely maintenance measures.

关 键 词：故障预测 核主元分析 全矢谱 AR模型 滚动轴承 信息融合 

分 类 号：TH16[机械工程—机械制造及自动化] TH133.3