流形模糊C均值方法及其在滚动轴承性能退化评估中的应用  

Fuzzy C-means Using Manifold Learning and Its Application to Rolling Bearing Performance Degradation Assessment

作　　者：王奉涛[1] 陈旭涛[1] 闫达文[2] 李宏坤[1] 王雷[1] 朱泓[1] 

机构地区：[1]大连理工大学振动工程研究所,大连116023 [2]大连理工大学数学科学学院,大连116023

出　　处：《机械工程学报》2016年第15期59-64,共6页Journal of Mechanical Engineering

基　　金：国家自然科学基金(51375067);航空科学基金(20132163010);中央高校基本科研业务费专项资金资助项目

摘　　要：滚动轴承全寿命周期性能退化监测是设备主动维修技术重要的组成部分,对损伤状态进行有效评估可以实现设备接近零停机运行,发挥机器的最大生产力。为有效描绘滚动轴承性能退化趋势,提出一种基于流形学习的模糊C均值(Fuzzy C-means algorithm,FCM)方法。首先提取监测信号的时域、频域特征及小波包时频域特征组成高维特征集,然后按确定的本征维数提取高维特征集的低维流形特征,进而建立基于局部线性嵌入流行学习(Locally linear embedding,LLE)的模糊C均值模型评估轴承当前运行状态。通过IMS滚动轴承全寿命试验,验证了该方法能够有效描绘滚动轴承性能退化阶段,为预知维修提供了重要信息。Entire life cycle degradation monitoring of rolling bearing is an important part of equipment initiative maintenance technology.Assessing the damage state effectively can realize near-zero downtime for equipment and exert its maximum productivity.In order to depict rolling bearing degradation trends effectively,the fuzzy C-means Algorithm（FCM） based on manifold learning is proposed.First of all,the time domain features,frequency domain features and wavelet packet time-frequency domain characteristics extracted from monitoring signals are used to constitute high-dimensional feature set.After then,the low-dimensional manifold features of the high-dimensional feature set are extracted according to the certain intrinsic dimension.In this sense,the FCM model based on locally linear embedding（LLE） manifold learning is built to evaluate current operating status of the rolling bearing.Finally,entire life cycle experiment of IMS rolling bearing is used to evaluate the efficiency of the proposed method for describing performance degradation stage of the rolling bearing.

关 键 词：LLE流形 模糊C均值 滚动轴承 性能退化 

分 类 号：TH113[机械工程—机械设计及理论] TN911[电子电信—通信与信息系统]