基于VMD和SVDD结合的滚动轴承性能退化程度定量评估  

Performance degradation quantitative assessment method for rolling bearings based on VMD and SVDD

作　　者：姜万录[1,2] 雷亚飞[1,2] 韩可 张生[1,2] 苏晓[1,2] JIANG Wanlu;LEI Yafei;HAN Ke;ZHANG Sheng;SU Xiao(Hebei Provincial Key Laboratory of Heavy Machinery Fluid Power Transmission and Control,Yanshan University,Qinhuangdao 066004,China;Key Laboratory of Advanced Forging ＆ Stamping Technology and Science（Yanshan University）,Ministry of Education of China,Qinhuangdao 066004,China;CRRC Nanjing Puzhen Co.,Ltd.,Nanjing 210031,China)

机构地区：[1]燕山大学河北省重型机械流体动力传输与控制重点实验室,河北秦皇岛066004 [2]燕山大学先进锻压成形技术与科学教育部重点实验室,河北秦皇岛066004 [3]中车南京浦镇车辆有限公司,南京210031

出　　处：《振动与冲击》2018年第22期43-50,共8页Journal of Vibration and Shock

基　　金：国家自然科学基金(51875498; 51475405);河北省自然科学基金重点项目(E2018203339);河北省博士研究生创新(CXZZBS2018045);国家重点基础研究发展计划(973计划)(2014CB046405)

摘　　要：提出了一种基于变分模态分解(VMD)和支持向量数据描述(SVDD)相结合的滚动轴承性能退化程度定量评估方法。针对采样时间长、采集到的信号数据点多时,信号中某些部分可能受到异常信号干扰的问题,首先提出了一种基于VMD和SVDD结合的特征提取新方法,将长信号分为多帧短信号,分别使用VMD方法分解短信号并提取各分量的奇异值组成特征向量,得到一组特征向量集,然后使用SVDD方法找到并剔除其中的异常样本点,求出剩余特征向量的平均值便可作为原信号的特征。特征提取完毕后,使用SVDD方法进行性能退化评估,以待检样本到训练得到的超球体模型球心的距离描述性能退化程度,并使用隶属函数将距离指标转化为与正常状态的隶属度作为性能退化指标,实现设备的性能退化程度的定量评估。使用轴承全寿命数据,并与以传统时域无量纲指标作为特征的分析结果进行了对比,验证了所提出方法的优越性。A performance degradation degree quantitative assessment method for rolling bearings was proposed, which integrates the methods of Variational Mode Decomposition（VMD） and Support Vector Data Description （SVDD）. Aiming at solving the problem that some parts of a signal may be disturbed by abnormal signals if the sampling time is long and the collected signal contains too many data points. A new feature extraction method based on VMD and SVDD was proposed, in which the long signal was segmented into several short frame signals, and the short signals were decomposed by VMD to obtain several components. The singular value of each component was extracted respectively to form a feature vector, and then a set of feature vectors was obtained. After finding and removing outliers by SVDD, the average value of the remained feature vectors was used as the feature of the original signal. Following the feature extraction, SVDD was used to assess the performance degradation. The degree of performance degradation was described by the distance from the test sample to the center of the hypersphere model, and the membership function was used to transform the distance index into the membership degree to the normal state and taken as the performance degradation index, which quantitatively assesses the performance degradation degree. The proposed method was tested with the complete life data of a rolling bearing, and the result was compared with the analysis result by the traditional time-domain index feature extraction method. The superiority of the proposed method was verified.

关 键 词：变分模态分解 支持向量数据描述 性能退化评估 异常点检测 

分 类 号：TH212[机械工程—机械制造及自动化] TH213.3