基于ASL-Isomap流形学习的滚动轴承故障诊断方法  

A Rolling Bearing Fault Diagnosis Method based on ASL-Isomap Manifold Learning

作　　者：王振亚[1] 戚晓利[1] 吴保林 WANG Zhenya;QI Xiaoli;WU Baolin(School of Mechanical Engineering, Anhui University of Technology, Ma'anshan 243032, Anhui China)

机构地区：[1]安徽工业大学机械工程学院,安徽马鞍山243032

出　　处：《噪声与振动控制》2019年第2期167-174,共8页Noise and Vibration Control

基　　金：国家自然科学基金资助项目(51505002);安徽省自然科学基金资助项目(1808085ME152);安徽省高校自然科学研究重点资助项目(KJ2017A053);研究生创新研究基金资助项目(2017012)

摘　　要：针对滚动轴承故障特征集维数高及冗余问题,提出一种基于自适应自组织增量学习神经网络界标点的等度规映射(Adaptive self-organizing incremental neural network landmark Isomap,ASL-Isomap)流形学习的滚动轴承故障诊断方法。首先,从时域、频域、时频域以及复杂域提取振动信号的故障特征,构建高维混合域故障特征集;其次,采用ASL-Isomap方法对高维混合域故障特征集进行维数约简,提取出低维、敏感特征子集;最后,应用核极限学习机(Kernel extreme learning machine,KELM)分类器对低维特征进行故障识别。ASL-Isomap方法集成自适应邻域构建和SOINN界标点选取的优势,能够更有效挖掘数据的低维本质流形。圆柱滚子轴承故障诊断实验验证该故障诊断方法的有效性和优越性。Aiming at the problem of over-high dimensions and redundancy in the mixed fault feature set of rolling bearings, a fault diagnosis method for the rolling bearings based on adaptive self-organizing incremental neural network landmark Isomap (ASL-Isomap) is proposed. Firstly, the fault features of vibration signals are extracted from time domain, frequency domain, time-frequency domain and complex domain to construct the high-dimensional hybrid domain fault feature set. Secondly, the ASL-Isomap method is used to reduce the dimension of the high-dimensional hybrid domain fault feature set, and the low-dimensional and sensitive feature subset is extracted. Finally, the low-dimensional features are input into a kernel extreme learning machine (KELM) classifier to recognize the fault types. The ASL-Isomap method integrates the advantages of adaptive neighborhood construction and SOINN landmarks selection, and can more effectively explore the low-dimensional essential manifold of the data set. Experimental results of fault diagnosis of cylindrical roller bearings show the effectiveness and advantage of the proposed method.

关 键 词：故障诊断 滚动轴承 流形学习 ASL-Isomap 核极限学习机 

分 类 号：TP206.3[自动化与计算机技术—检测技术与自动化装置] TN911.7[自动化与计算机技术—控制科学与工程]