[1]王海军,许飞云.基于非负Tucker 3分解的稀疏分量分析在故障信号提取中的应用[J].东南大学学报(自然科学版),2013,43(4):758-762.[doi:10.3969/j.issn.1001-0505.2013.04.016]
 Wang Haijun,Xu Feiyun.Sparse component analysis based on nonnegative Tucker 3 decomposition for fault signal extraction[J].Journal of Southeast University (Natural Science Edition),2013,43(4):758-762.[doi:10.3969/j.issn.1001-0505.2013.04.016]
点击复制

基于非负Tucker 3分解的稀疏分量分析在故障信号提取中的应用()
分享到:

《东南大学学报(自然科学版)》[ISSN:1001-0505/CN:32-1178/N]

卷:
43
期数:
2013年第4期
页码:
758-762
栏目:
机械工程
出版日期:
2013-07-20

文章信息/Info

Title:
Sparse component analysis based on nonnegative Tucker 3 decomposition for fault signal extraction
作者:
王海军许飞云
东南大学机械工程学院, 南京211189
Author(s):
Wang Haijun Xu Feiyun
School of Mechanical Engineering, Southeast University, Nanjing 211189, China
关键词:
非负Tucker 3分解 稀疏分量分析 更新算法 交替最小二乘法
Keywords:
nonnegative Tucker 3 decomposition sparse component analysis updating algorithm alternative least squares method
分类号:
TH17;TP206
DOI:
10.3969/j.issn.1001-0505.2013.04.016
摘要:
针对初始故障信号不稀疏难于判断的问题,在非负Tucker 3分解(NTD)的基础上,提出了一种基于NTD的稀疏分量分析(SCA)处理二次特征信号的方法.同时,为了克服NTD算法收敛慢、易陷入过拟合等局限性,对分解因子增加了非负约束,并提出了对分解因子一次更新的算法.对比传统的最小交替二乘法,该更新算法能一次性地计算所有分解因子,避免了计算大规模的Jacobian矩阵,从而较大地提高了算法的效率.实验结果表明:NTD和SCA相结合的方法(SCA_NTD)只需迭代约150步可达到收敛,而且在频谱稀疏性处理方面优于NTF等传统的方法;在分解相同维数张量的条件下,SCA_NTD的最高精度达到了97.16%.因此,SCA_NTD不仅能够改善信号特征的稀疏性,而且对提高算法的收敛速度和精度也具有重要的意义.
Abstract:
Aiming at the problem of non-sparseness of original diagnosis signal and being difficult to distinguish, a method of sparse component analysis(SCA)based on nonnegative Tucker 3 decomposition(NTD)is proposed to process the quadratic feature of faults for improving the sparseness. Meanwhile, a new updating algorithm with nonnegative constraints is put forward to overcome the limitation of slow convergence and data overfitting in NTD. Compared with the conventional alternative least squares(ALS), the updating algorithm can simultaneously compute all the factors in one time that avoids calculating the large-scale Jacobian matrix and therefore improves the efficiency. The experimental results show that the method of combining NTD and SCA(SCA_NTD)only needs 150 steps to achieve convergence which is superior to other typical methods like NTF, etc; SCA_NTD also gets the highest accuracy of 97.16% under the same dimension of a tensor. Therefore, SCA_NTD not only improves the sparseness but also is significant to improve the convergence and efficiency.

参考文献/References:

[1] Tucker L R. Some mathematical notes on three-mode factor analysis[J]. Psychometrika, 1966, 31(3): 279-311.
[2] Paatero P, Tapper U. Positive matrix factorization: a non-negative factor model with optimal utilization of error estimates of data values [J]. Environmetrics, 1994, 5(2): 111-126.
[3] Lee D D, Seung H S. Learning the parts of objects by non-negative matrix factorization[J]. Nature, 1999, 401(6755): 788-791.
[4] Hazan T, Polak S,Shashua A. Sparse image coding using a 3D non-negative tensor factorization[C]//10th IEEE International Conference on Computer Vision. Beijing, 2005: 50-57.
[5] Morup M, Hansen L K, Arnfred S M. Algorithms for sparse nonnegative Tucker decompositions [J]. Neural Computation, 2008, 20(8): 2112-2131.
[6] Cichocki A, Zdunek R, Phan A H, et al. Alternating least squares and related algorithms for NMF and SCA problems in nonnegative matrix and tensor factorizations[M]. Chichester, UK: John Wiley & Sons, Ltd, 2009: 203-266.
[7] Peng S, Xu F, Jia M, et al. Sparseness-controlled non-negative tensor factorization and its application in machinery fault diagnosis[J]. Journal of Southeast University:English Edition, 2009, 25(3): 346-350.
[8] Karoui M S, Deville Y, Hosseini S, et al. Blind spatial unmixing of multispectral images: new methods combining sparse component analysis, clustering and non-negativity constraints[J]. Pattern Recognition, 2012, 45(12): 4263-4278.
[9] Asaei A, Davies M E, Bourlard H, et al. Computational methods for structured sparse component analysis of convolutive speech mixtures[C]//IEEE International Conference on Acoustics, Speech and Signal Processing. Kyoto, Japan, 2012: 2425-2428.
[10] Kolda T G. Multilinear operators for higher-order decompositions[M]. California, USA: Sandia National Laboratories, 2006.
[11] Cai X J, Chen Y N, Han D R. Nonnegative tensor factorizations using an alternating direction method[J]. Frontiers of Mathematics in China, 2013, 8(1): 3-18.
[12] Jiang L L, Yin H Q. Bregman iteration algorithm for sparse nonnegative matrix factorizations via alternating l(1)-norm minimization [J]. Multidimensional Systems and Signal Processing, 2012, 23(3): 315-328.
[13] Albright R, Cox J, Duling D, et al. Algorithms, initializations, and convergence for the nonnegative matrix factorization[R]. Raleigh, USA: Carolina State University, 2006.
[14] Khoromskij B, Khoromskaia V, Chinnamsetty S, et al. Tensor decomposition in electronic structure calculations on 3D Cartesian grids[J]. Journal of Computational Physics, 2009, 228(16): 5749-5762.
[15] 刘海林,姚楚君.欠定混叠稀疏分量分析的超平面聚类算法 [J].系统仿真学报,2009(7):1826-1828.
  Liu Hailin, Yao Chujun. Hyperplane clustering algorithm of underdetermined mixing sparse component analysis[J]. Journal of System Simulation,2009(7):1826-1828.(in Chinese)

备注/Memo

备注/Memo:
作者简介: 王海军(1982—),男,博士生;许飞云(联系人),男,博士,教授,博士生导师,fyxu@seu.du.cn.
基金项目: 国家自然科学基金资助项目(50875048,51175079,51075069).
引文格式: 王海军,许飞云.基于非负Tucker 3分解的稀疏分量分析在故障信号提取中的应用[J].东南大学学报:自然科学版,2013,43(4):758-762. [doi:10.3969/j.issn.1001-0505.2013.04.016]
更新日期/Last Update: 2013-07-20