基于EEMD子带提取相关机械振动信号单通道盲源分离

发布时间：2018-01-29 02:55

本文关键词： 盲源分离相关机械源总体经验模态分解欠定　出处：《振动与冲击》2014年20期 　论文类型：期刊论文

【摘要】：针对传统独立分量分析难以解决机械故障诊断中存在的相关源信号盲分离、欠定盲分离等问题,在相关振源信号部分子带满足统计独立的假设前提下,提出基于总体经验模态分解子带提取相关机械源单通道盲源分离方法。该方法将单通道观测信号进行总体经验模态分解获得到子带观测信号,将单通道信号及子带观测信号组成新的多维信号,利用奇异值分解及贝叶斯准则估计源信号数目;据互信息标准与源信号数目选若干独立性较强的子带观测信号重构,实现信号升维;对重构的观测信号进行白化预处理及联合近似对角化,获得源信号估计。并仿真、实验验证该方法在机械故障诊断中的有效性。
[Abstract]:Traditional independent component analysis (ICA) is difficult to solve the problems of blind separation and under-determined blind separation of correlated source signals in mechanical fault diagnosis. Under the assumption that some sub-bands of correlated vibration source signals meet the assumption of statistical independence. A method of extracting single channel blind source based on global empirical mode decomposition subband is proposed, in which single channel observation signal is decomposed to subband observation signal. The single channel signal and the subband observation signal are used to form a new multi-dimensional signal, and the number of the source signal is estimated by using singular value decomposition and Bayesian criterion. According to the mutual information standard and the number of source signals, some independent sub-band observation signals are reconstructed to realize signal dimension elevation. The albino preprocessing and joint approximate diagonalization of the reconstructed observation signal are used to obtain the source signal estimation. The simulation results show that the proposed method is effective in mechanical fault diagnosis.
【作者单位】：燕山大学电气工程学院;
【基金】：国家自然科学基金(51105323) 河北省自然科学基金(E2012203166)
【分类号】：TH165.3;TN911.7
【正文快照】： 由于现代化工业生产不断发展,机械设备故障诊断技术备受关注。独立分量分析(Independent Compo-nent Analysis,ICA)为较新的信号处理方法[1]。而基于独立分量分析的盲源分离(Blind Signal Separation,BSS)方法在机械设备故障诊断及状态监测中已获得成功应用[2]。该方法大多局

【参考文献】