采用时间序列突变点检测的滚动轴承性能退化评价方法  

A Method for Evaluating Performance Degradation of Rolling Bearings Using Detection of Time Series Mutation Point

作　　者：刘弹[1] 李晓婉 梁霖[1] 吴杰 徐光华[1] 沈强 LIU Dan;LI Xiaowan;LIANG Lin;WU Jie;XU Guanghua;SHEN Qiang(School of Mechanical Engineering,Xi’an Jiaotong University,Xi’an 710049,China)

机构地区：[1]西安交通大学机械工程学院

出　　处：《西安交通大学学报》2019年第12期10-16,共7页Journal of Xi'an Jiaotong University

基　　金：国家科技重大专项资助项目(2014ZX040015061-05)

摘　　要：以滚动轴承为研究对象,根据性能退化曲线的现实意义定义了对性能退化特征提取具有指导意义的评价方法,该方法由初始退化点、初始敏感性、失效突变性和趋势一致性4个指标构成,并根据性能退化曲线的物理特性设计了性能退化特征评价的定量化指标;利用时间序列线性化突变点检测的方法给出了评价指标的计算过程,为性能退化特征的选取提供了更为直观的依据。使用国际通用的Swiss数据集进行验证,实验结果表明该评价方法可以全面评价各个性能退化特征对性能退化过程的表征效果。为了更好地保持原始高维数据的空间位置关系,提出了基于邻域参数自适应选取的局部线性嵌入(LLE)算法,通过自适应局部权值向量来保存高维空间数据的局部线性结构,将多个局部线性进行叠加来不断地逼近全局的非线性,有效提高直接使用高维数据反映原始状态性能的可靠性和稳定性。使用滚动轴承加速寿命试验数据,对比邻域参数自适应LLE算法和常规LLE算法所提取的性能退化特征,验证了邻域参数自适应LLE算法可以保留更多原始高维数据的信息。Taking rolling bearings as research object and according to its practical significance of performance degradation curve,four evaluation indexes are defined to guide the extraction of performance degradation characteristics.These indexes are initial degradation point,initial sensitivity,failure mutagenicity and trend consistency.A quantitative index for evaluating the performance degradation characteristic is designed based on the physical characteristics of the performance degradation curve,and a calculation process of the evaluation index is given by using the method of detecting time series linearization mutation point,which provides a more intuitive basis for the selection of performance degradation characteristics.The proposed method is verified by using the internationally available Swiss datasets.Experimental results show that the method can comprehensively evaluate the characterization effects of all performance degradation characteristics on performance degradation process.A local linear embedding(LLE)algorithm based on adaptive selection of neighborhood parameters is proposed to better maintain the spatial location relationship of the original high-dimensional data.The local linear structure of high-dimensional spatial data is preserved by an adaptive local weight vector,and multiple local linearities are superimposed to continuously approximate the global nonlinearity.The reliability and stability of directly using high-dimensional data to reflect the performance of the original state are hence effectively improved.The performance degradation characteristics extracted by the neighborhood parameter adaptive LLE algorithm and the conventional LLE algorithm from the accelerated life test data of rolling bearings are compared.Results show that the neighborhood parameter adaptive LLE algorithm retains more information of the original high-dimensional data.

关 键 词：性能退化评价 特征提取 时间序列突变点 局部线性嵌入 滚动轴承 

分 类 号：TH17[机械工程—机械制造及自动化]