一种旋转机械振动信号的盲源分离消噪方法  

De-noising Vibration Signal of Rotating Machinery with Blind Sources Separation

作　　者：周晓峰[1] 杨世锡[1] 甘春标[1] 

机构地区：[1]浙江大学液压传动及控制国家重点实验室,杭州310027

出　　处：《振动．测试与诊断》2012年第5期714-717,858-859,共4页Journal of Vibration,Measurement & Diagnosis

基　　金：国家高技术研究发展计划("八六三"计划)资助项目(编号:2009AA04Z410);国家自然科学基金资助项目(编号:50675194)

摘　　要：为了消除旋转机械振动信号中不同类型的噪声,提出了一种基于虚拟信号(virtual signal,简称VS)的多级独立分量分析(independent component analysis,简称ICA)的消噪方法。通过引入与测量噪声匹配的虚拟噪声通道,将单通道观测信号扩展为多通道观测信号,用独立分量分析方法消除与数据采集系统相关的测量噪声。将振源信号的组合(有用信号)视为一个虚拟源,对消除了测量噪声的两通道观测信号再次用独立分量分析方法实现有用信号和背景噪声的分离,从而达到消除背景噪声的目的。试验表明,该方法可以得到很好的消噪结果,有效提高信号的信噪比。A new denoising method based on VS-ICA (Virtual Signals Independent Component Analysis) is proposed to remove noise in vibration signals of rotating machinery. First, the one-dimensional observed signal is transformed into multi-dimensional signal vector by incorporating virtual noise channel into original observed signal. ICA is applied to extend observed signal, thus the noise related to data acquisition system is reduced. Secondly, virtual source, which is a composition of vibrating sources, is introduced. Thus, only two observed signals, which have been removed from measurement noise, are needed for separating background noise and virtual source through ICA. The experiments indicate that excellent de-noised results can be obtained and the signal noise ratio can be improved effectively with the proposed method.

关 键 词：虚拟噪声通道 虚拟源 独立分量分析 噪声消除 盲源分离 

分 类 号：TN911.7[电子电信—通信与信息系统] TH165.3[电子电信—信息与通信工程]