利用GS优化SM-SVM的滚动轴承故障诊断方法研究  

Research on Rolling Bearing Fault Diagnosis Method Based on GS Optimized SM-SVM

作　　者：周超[1] 曹春平[1] 孙宇[1] ZHOU Chao;CAO Chun-ping;SUN Yu(School of Mechanical Engineering,Nanjing University of Science and Technology,Jiangsu Nanjing210094,China)

机构地区：[1]南京理工大学机械工程学院,江苏南京210094

出　　处：《机械设计与制造》2020年第6期16-19,共4页Machinery Design & Manufacture

基　　金：江苏省重点研发计划—产业前瞻工艺性关键技术(BE2015011-3)。

摘　　要：针对滚动轴承常见故障,提出利用网格搜索(GS)优化序列最小支持向量机(SM-SVM)的故障诊断方法.首先,对提取的滚动轴承振动信号进行预处理,并对其分别提取峭度指标、偏度系数、方均根值等时域统计量和小波包分解节点能量等特征,并对特征向量进行归一化和PCA降维处理.其次,利用GS算法对SM-SVM的核函数参数和惩罚因子进行优化,以提高滚动轴承故障模式识别的正确率.最后,利用MATLAB LIBSVM工具箱对滚动轴承不同故障进行模式识别,并将本方法与SM-SVM和LS-SVM方法进行了比较.结果发现,改进方法的模式识别正确率比原方法的高出5%.For several common faults of rolling bearings,a diagnostic method using grid search(GS)to optimize sequence minimum support vector machine(SM-SVM)was proposed.Firstly,the extracted vibration signal of rolling bearing was prepr-ocessed,and the kurtosis index,the skewness coefficient,root mean square value of time-domain statistics and the node energy of the wavelet packet decomposition were extracted respectively from the preprocessed vibration signal.The time-domain statistics and the node energy of the wavelet packet decomposition was taken as the feature to form the feature vector,and the feature vector was normalized and its’dimensions were reduced by PCA.Secondly,GS algorithm was used to optimize SM-SVM kernel function parameters and penalty factors so as to improve the accuracy of SM-SVM method for fault pattern recognition of rolling bearing.Finally,MATLAB LIBSVM toolbox was used to perform pattern recognition of different faults for rolling bearing,and the method proposed was compared with SM-SVM and LS-SVM methods.The results indicates that the improved method of pattern recognition accuracy rate is higher than the original method of 5%.

关 键 词：故障诊断 滚动轴承 模式识别 SM-SVM GS 

分 类 号：TH16[机械工程—机械制造及自动化] TP206.3[自动化与计算机技术—检测技术与自动化装置]