ELM-Adaboost分类器在轴承故障诊断中的运用  

Fault Diagnosis of Rolling Bearing Based on ELM-Adaboost Model

作　　者：胡超 沈宝国 杨妍 谢中敏 HU Chao;SHEN Bao-guo;YANG Yan;XIE Zhong-min(College of Aeronautical Engineering,Jiangsu Aviation Technical College,Jiangsu Zhenjiang 212134,China)

机构地区：[1]江苏航空职业技术学院航空工程学院,江苏镇江212134

出　　处：《机械设计与制造》2022年第2期111-115,共5页Machinery Design & Manufacture

基　　金：江苏省自然科学基金项目(BK20180863);镇江市科技计划资助(GY2018029);2017年度院级课题资助项目(JATC17010102)。

摘　　要：研究强分类器ELM-Adaboost模型在滚动轴承故障诊断中的运用。首先,从滚动轴承故障振动信号中提取时域特征参数,并采用因子分析法对变量进行降维处理;其次,对ELM-Adaboost模型中的关键参数进行详细分析,并选择最优的参数对ELM-Adaboost模型进行优化;最后,将ELM-Adaboost模型用于滚动轴承故障诊断中。实验研究结果表明:神经元数量和激活函数都能明显的影响到故障诊断准确率;激活函数采用Hardlim()函数的诊断效果比其它函数要好;ELM-Adaboost模型对神经元数量引起的诊断准确率波动性较小;在重复30次实验中,ELM-Adaboost模型对不同类型的轴承故障诊断准确率均在84%以上,而ELM模型则在78%以上,且ELM-Adaboost模型对重复诊断导致的结果波动性相比较低。The strong classifier ofELM-Adaboost model is studied,and it is used to diagnose rolling bearing faults. Firstly,the time-domain characteristic parameters are extracted from the rolling bearing fault vibration signal,and dimensionality of variables are reduced by the factor analysis method. Secondly,the key parameters of ELM-Adaboost model are analyzed in detail,and optimal parameters are selected to optimize the ELM-Adaboost model. Finally,the ELM-Adaboost model is used in rolling bearing fault diagnosis. The experimental results show that both the number of neurons and the activation function can significantly affect the accuracy of fault diagnosis of ELM-Adaboost model and ELM model. When the activation function of hidden layer of ELM model is Harlim()function,the diagnostic effect of the model is better. The ELM-Adaboost model has less fluctuation in the diagnostic accuracy caused by the number of neurons. In the repeated 30 experiments,the accuracy of the ELM-Adaboost model for different types of bearing fault diagnosis is above 84%,while the ELM model is above 78%. And the ELM-Adaboost model is less volatile than the results of repeated diagnosis.

关 键 词：滚动轴承 故障诊断 ELM-Adaboost 因子分析 

分 类 号：TH16[机械工程—机械制造及自动化] TH17