基于半监督PCA-LPP流形学习算法的故障降维辨识

来源期刊:中南大学学报(自然科学版)2016年第5期

论文作者:唐力伟 张晓涛 王平 邓士杰

文章页码:1559 - 1565

关键词:流形学习;局部保持投影;主元分析;故障诊断;故障辨识

Key words:manifold learning; locality preserving projection; principal component analysis; fault diagnosis; pattern recognition

摘    要:提出一种基于半监督思想PCA-LPP的流形学习维数约简故障辨识方法,兼顾PCA的全局结构和LPP的局部结构保持以及样本的类别信息,构造新的投影矩阵目标函数,给出PCA-LPP流形学习算法的计算原理。采用UCI中wine数据集验证半监督PCA-LPP方法的维数约简性能,并就齿轮箱故障声发射实验信号,以小波包能量熵作为特征向量,并将特征向量的降维结果输入支持向量机进行故障类型辨识。研究结果表明:半监督PCA-LPP方法的降维结果,能够充分考虑不同故障特征向量的差异信息,相应的故障类型辨识精度高于PCA及LPP方法。

Abstract: A novel fault identification dimensionality reduction method based on semi-supervised PCA-LPP manifold learning algorithm was proposed. The objective function of projection matrix of semi-supervised PCA-LPP was constructed by global structure and local structure, the global structure was described by PCA, the local structure described by LPP and category information of samples, and the calculation principle of semi-supervised PCA-LPP manifold learning algorithm was given. The processing results of wine dataset of UCI show that the semi-supervised PCA-LPP method has a good ability of dimensionality reduction. Aiming at the gearbox acoustic emission signals, its eigenvectors is constructed by wavelet packet energy entropy, and the dimensionality reduction results of eigenvectors are given to the support vector machine, the fault identification of semi-supervised PCA-LPP method obtains higher identification rate than that of LPP and PCA, because the method considers the similarities and differences between all eigenvectors.

相关论文

  • 暂无!

相关知识点

  • 暂无!

有色金属在线官网  |   会议  |   在线投稿  |   购买纸书  |   科技图书馆

中南大学出版社 技术支持 版权声明   电话:0731-88830515 88830516   传真:0731-88710482   Email:administrator@cnnmol.com

互联网出版许可证:(署)网出证(京)字第342号   京ICP备17050991号-6      京公网安备11010802042557号