简谐成分的盲源分离适用性研究

发布时间：2018-02-20 22:20

本文关键词： 盲源分离独立分量分析二阶盲辨识峭度模态识别简谐成分　出处：《哈尔滨工程大学学报》2014年04期 　论文类型：期刊论文

【摘要】：盲源分离问题(BSS)大多基于信源信号的独立性假设或者时间结构假设条件来展开研究,对信源的不当假设可能导致算法过学习,产生虚假的信源识别结果。针对机械系统中普遍存在的简谐成分,研究了BSS方法应用于简谐成分盲分离的适用性。简要介绍了2种典型的BSS方法——独立分量分析方法(ICA)和二阶盲辨识方法(SOBI),通过峭度分析简谐信号的非高斯性,发现当简谐信号构成傅里叶级数系时,有可能构成非高斯性更强的信号。应用FastICA算法和SOBI算法进行简谐信号盲分离的仿真研究以及简支梁结构模态识别的实验研究。结果表明:当简谐信号构成傅里叶级数系时,ICA方法会优先分离非高斯性更强的信号,导致方法过学习;而SOBI方法能确保简谐成分的盲分离过程准确可靠。
[Abstract]:The blind source separation problem (BSS) is mostly based on the independence hypothesis or time structure hypothesis of the source signal. The improper assumption of the source may lead to over-learning of the algorithm. Producing false source identification results. In view of the common harmonic components in mechanical systems, The applicability of BSS method in blind separation of simple harmonic components is studied. Two typical BSS methods, independent component analysis (ICA) and second order blind identification, are briefly introduced. The non-#china_person0# of simple harmonic signals is analyzed by kurtosis. It is found that when a simple harmonic signal constitutes a Fourier series system, FastICA algorithm and SOBI algorithm are used to simulate the blind separation of simple harmonic signals and the experimental study on modal identification of simple beam structures. The results show that when the simple harmonic signals form Fourier, the results show that: 1. The series system timing ICA method will give priority to the separation of non-#china_person0# signals. The SOBI method can ensure the accuracy and reliability of blind separation of harmonic components.
【作者单位】：哈尔滨工程大学动力与能源工程学院;
【基金】：国家自然科学基金资助项目(51375103)
【分类号】：TN911.7

【共引文献】