基于EEMD和多元多尺度熵的风力发电机组滚动轴承故障特征提取  

Feature extraction of rolling bearing for wind generator based on EEMD and multivariate multiscale entropy

作　　者：韩龙[1,2] 李成伟[1] 王丽[2] 朱显辉[2] 苏勋文[2] 

机构地区：[1]哈尔滨工业大学电气工程及自动化学院,哈尔滨150001 [2]黑龙江科技大学电气与控制工程学院,哈尔滨150022

出　　处：《工业仪表与自动化装置》2016年第1期23-26,共4页Industrial Instrumentation & Automation

基　　金：国家自然科学基金项目(51107015);黑龙江省教育厅科学技术研究项目(12543057)

摘　　要：为了降低风力发电机组滚动轴承信号的噪声和进行多信道数据处理,提出了一种基于EEMD和多元多尺度熵的特征提取方法。利用EEMD算法对多信道的原始声发射信号进行分解获取无模式混淆的IMF,通过敏感度评估算法选取反应故障特征敏感的IMF进行多元多尺度熵分析,由单因素方差分析选择最优尺度对应的多元样本熵作为各种故障的特征值。通过从实验台采集得到正常、轻微损伤和断裂3种状态的样本数据,与多种特征提取方法相比较和SVM算法分类分析,证明了所选择故障特征量的准确性,同时也验证了所提出的滚动轴承故障特征提取方法的有效性和准确性。In order to reduce the noise of the rolling bearing of wind turbines,and process multi-channel data,a method of feature extraction based on EEMD and multivariate multiscale entropy was proposed in this paper. The original acoustic emission signal of multi channel was decomposed with EEMD to obtain free mode confusion IMF,multiple variables were constituted by the response fault feature sensitive IMF that was selected by the sensitivity evaluation algorithm,and it was processed by multivariate multiscale entropy analysis,the optimal multiple scale sample entropy was selected by analysis of variance as eigenvalues of various faults. The data of normal,slight damage and fracture can be acquired from the test platform. The accuracy of the characteristics of fault can be proved by the comparison with multiple feature extraction methods and the analysis of the SVM classification algorithm. Meanwhile,it showed that the method of fault feature extraction of rolling bearing is effective.

关 键 词：风力发电机组 滚动轴承 特征提取 EEMD 多元多尺度熵 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]