邻域自适应增量式PCA-LPP在齿轮箱故障诊断中的应用

发布时间：2018-01-13 13:19

本文关键词：邻域自适应增量式PCA-LPP在齿轮箱故障诊断中的应用　出处：《振动与冲击》2017年14期 　论文类型：期刊论文

【摘要】：针对流形学习算法的增量处理问题,提出一种邻域自适应增量式PCA-LPP流形学习算法,阐述了算法的基本原理以及增量样本处理方法。对新增样本的引入,首先根据已有样本对协方差矩阵和相似矩阵进行增量更新,而后结合已有样本降维结果对新增样本降维结果进行估计,最后采用子空间迭代法实现新旧样本降维结果的更新。采用齿轮箱故障信号特征向量对邻域自适应增量式PCA-LPP流形学习算法进行检验,结果表明,邻域自适应增量式PCA-LPP流形学习算法降维后特征具有良好的故障分类识别效果。
[Abstract]:A neighborhood adaptive incremental PCA-LPP manifold learning algorithm is proposed for the incremental processing of manifold learning algorithm. The basic principle of the algorithm and the method of incremental sample processing are introduced. Firstly, the covariance matrix and similarity matrix are incrementally updated according to the existing samples. Then the dimension reduction results of the new samples are estimated by using the existing sample dimensionality reduction results. Finally, the subspace iteration method is used to update the new and old sample dimensionality reduction results, and the gearbox fault signal eigenvector is used to test the neighborhood adaptive incremental PCA-LPP manifold learning algorithm, and the results show that. Neighborhood adaptive incremental PCA-LPP manifold learning algorithm has good performance in fault classification and identification after dimensionality reduction.
【作者单位】：军械工程学院火炮工程系;
【基金】：国家自然科学基金资助项目(50775219) 军队科研资助项目
【分类号】：TH132.41;TP181
【正文快照】： 流形学习是一类有效的非监督数据降维方法[1-2],能够提取嵌入在高维数据中的低维特征,为后续的数据分类提供有效的特征降维处理支持。He等[3]在拉普拉斯映射的基础上,提出一种典型的局部保持投影算法(Locality Preserving Projection,LPP),其能够在投影过程中保持数据样本局部

【参考文献】