基于半监督邻域自适应正交判别投影的转子故障诊断  

Rotor fault diagnosis based on semi-supervised neighborhood adaptiveorthogonal discriminant projection

作　　者：常书源 赵荣珍[1] 石明宽 CHANG Shuyuan;ZHAO Rongzhen;SHI Mingkuan(School of Mechanical and Electronical Engineering,Lanzhou University of Technology,Lanzhou 730050,China)

机构地区：[1]兰州理工大学机电工程学院,兰州730050

出　　处：《振动与冲击》2021年第10期159-165,共7页Journal of Vibration and Shock

基　　金：国家自然科学基金(51675253);兰州理工大学红柳一流学科建设项目。

摘　　要：针对故障特征集维数过高导致故障难以辨识的问题,提出一种基于半监督邻域自适应正交判别投影(SSNA-ODP)的转子故障诊断方法。提取原始振动信号的时域、频域、时频域特征构造混合域特征集;利用SSNA-ODP方法对混合域特征集进行维数约简,提取出有利于实施分类的低维特征子集;输入到支持向量机(SVM)中进行模式识别。典型故障数据样本的应用验证情况表明,该方法能够改善ODP方法在有标记样本较少时的泛化能力和使用全局统一邻域参数的数据流形特征,从而有效提高了故障识别的准确率。A rotor fault diagnosis method based on semi-supervised neighborhood adaptive orthogonal discriminant projection(SSNA-ODP)was proposed to solve the difficulty in fault identification due to the high dimension of the fault feature set.The features in time domain,frequency domain and time frequency domain of the original vibration signal were extracted to construct a feature set in mixed domain.The SSNA-ODP method was used to reduce the dimension of the feature set in mixed domain and extract the low-dimensional feature subset which was beneficial to classification and then the set was input into a support vector machine(SVM)for pattern recognition.The application verification by using typical fault data samples shows that the method can improve the generalization ability of the ODP method when there are fewer labeled samples and make use of the data manifold characteristics of global uniform neighborhood parameters,so as to effectively improve the accuracy of fault identification.

关 键 词：维数约简 半监督 正交判别投影(ODP) 故障诊断 

分 类 号：TH165.3[机械工程—机械制造及自动化] TP18[自动化与计算机技术—控制理论与控制工程]