有监督不相关正交局部保持映射故障辨识 .pdfVIP

学兔兔 第34卷 第5期 仪 器 仪 表 学 报 Vo1．34 No．5 2013年5月 Chinese Journal of Scientific Instrument Mav．2013 有监督不相关正交局部保持映射故障辨识 米 李 锋 ，王家序 ，杨荣松 (1．四川大学制造科学与工程学院 成都 610065；2．四川大学空天科学与工程学院 成都 610065) 摘 要：提出基于有监督不相关正交局部保持映射(SUOLPP)维数化简的故障辨识方法。首先构造全面表征不同故障特征的 时频域特征集，再利用SUOLLP将高维时频域特征集自动约简为具有更好区分度的低维特征矢量，并输入到Littlewood—Paley小 波核支持向量机中进行故障模式辨识。时频域特征融集可较全面准确地反映旋转机械的故障特征；SUOLPP同时利用流形局 部几何结构和类标签来设计相似加权矩阵，并使输出基向量统计不相关和相互正交，提高了故障辨识精度。深沟球轴承故障诊 断和空间轴承寿命状态辨识实例验证了该方法的有效性。 关键词：时频域特征集；有监督不相关正交局部保持映射；维数化简；流形学习；故障辨识 中图分类号：TH165．3 TN911．2 文献标识码：A 国家标准学科分类代码：510．40 Fault identification method based on supervised uncorrelated orthogonal locality preserving projection Li Feng ，Wang Jiaxu ，Yang Rongsong ( ．School ofManufacturing Science and Engineering，Sichuan University，Chengdu 610065，China； 2．School of Aerospace Science and Engineering，Sichuan University，Chengdu 610065，China) Abstract：A novel fault diagnosis method based on dimensionality reduction with supervised uncorrelated orthogonal Locality Preserving Projection(SUOLPP)is proposed in this paper．The time—frequency domain feature set is firstly constructed，which completely characterize the properties of different faults．Then，SUOLPP is used to automatically reduce the high--dimensional time··frequency domain feature sets of training and test samples to the low--dimensional eigenvectors that have better discrimination．Finally，the low—dimensional eigenveetors of training and test samples are inputted into Littlewood—Paley wavelet support vector machine(LPWSVM)to carry out fault identification．The time- frequency domain feature set can more comprehensively and accurately refle