摘要
针对滚动轴承故障样本不平衡和故障特征存在冗余性问题,提出了基于无监督判别投影(UDP)的滚动轴承故障诊断方法。该方法首先从时域和时频域提取多个特征参数,从而构造一个原始的高维特征集,随后运用UDP算法从该特征集中提取最敏感的低维流形特征,最后利用K-近邻分类器识别出滚动轴承的运行状态。将该方法分别应用于轴承故障类型和内圈故障严重性的识别,并与传统方法进行了比较,验证了该方法的可行性和优越性。
Aiming at the imbalanced fault samples and redundant fault features of rolling bearings,a rolling bearing fault diagnosis method was proposed based on UDP.The method firstly extracted several feature parameters from time domain and time-frequency domain.Thus,the raw high-dimensional feature set was constructed.Subsequently,the most sensitive low-dimensional manifold features were extracted from the feature set by employing UDP algorithm.Finally,K-nearest neighbor classifier was utilized to recognize the operating conditions of rolling bearings.The method was applied to the identification of bearing fault categories and inner fault severities separately.Compared with the traditional methods,the feasibility and superiority were validated.
出处
《中国机械工程》
EI
CAS
CSCD
北大核心
2016年第16期2202-2206,共5页
China Mechanical Engineering
基金
国家自然科学基金资助项目(71171154)
湖北省自然科学基金资助项目(2015CFB698)
湖北省科技支撑计划资助项目(2014BAA032
2015BAA063)
关键词
故障诊断
特征提取
流形学习
无监督判别投影
fault diagnosis
feature extraction
manifold learning
unsupervised discriminant projection（UDP）
