ALIF-MMPE结合DAG-SVM的滚动轴承故障诊断  

Rolling Bearing Fault Diagnosis with ALIF-MMPE and DAG-SVM

作　　者：韩美东 张金豹[2] 赵永强[2] Han Meidong;Zhang Jinbao;Zhao Yongqiang(School of Aircraft Engineering,Nanchang Hangkong University,Nanchang 330063,China;School of Mechatronics Engineering,Harbin Institute of Technology,Harbin 150001,China)

机构地区：[1]南昌航空大学飞行器工程学院,南昌330063 [2]哈尔滨工业大学机电工程学院,哈尔滨150001

出　　处：《机械科学与技术》2020年第9期1358-1365,共8页Mechanical Science and Technology for Aerospace Engineering

基　　金：国家自然科学基金项目(51505100)资助。

摘　　要：针对滚动轴承故障诊断中非平稳振动信号下的有效故障特征提取问题,提出一种基于自适应局部迭代滤波、多元多尺度排列熵和有向无环图算法支持向量机的滚动轴承故障诊断方法。自适应局部迭代滤波通过构建自适应滤波函数,能够有效抑制噪声和模态混叠,经自适应分解后得到若干本征模态函数。仿真结果表明其效果优于经验模态分解。然后利用多元多尺度排列熵对包含显著故障信息的本征模态函数进行信息融合和特征提取,组成故障状态特征集。采用主成分分析对故障状态特征集进行降维,随机抽取部分样本带入有向无环图算法支持向量机中进行训练,其它则作为测试样本进行故障识别和诊断。试验故障诊断结果表明:自适应局部迭代滤波下多元多尺度排列熵优于多个本征模态函数下的多尺度排列熵和经验模态分解下的多元多尺度排列熵;本文方法能准确地识别滚动轴承不同的故障类型及故障程度。An approach based on adaptive local iterative filtering(ALIF),multivariate multiscale permutation entropy(MMPE)and directed acyclic graph support vector machine(DAG-SVM)is proposed in this paper for the problem of efficient fault feature extraction from the non-stationary vibration signals in the fault diagnosis of rolling bearings.ALIF could construct adaptive filtering functions to efficiently constrain noise and mode mixing,and a number of intrinsic mode functions(IMFs)are obtained with adaptive decomposition.The simulation result demonstrates that ALIF outperforms empirical mode decomposition(EMD).Then MMPE is employed to perform information infusion and feature extraction with the IMFs containing primary faulty information,and fault state feature sets are constructed with MMPE.After dimensionality reduction using principal component analysis(PCA),some of the new features are randomly selected to train the DAG-SVM,while the remaining ones are used as test samples for identification and fault diagnosis.Results of the experiment show that MMPE of ALIF is superior to the multiscale permutation entropy(MPE)of multiple IMFs and MMPE of EMD;the proposed method could efficiently identify different types of faults and severity levels of rolling bearings.

关 键 词：轴承故障诊断 自适应局部迭代滤波 多元多尺度排列熵 有向无环图算法支持向量机 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]