文献类型：期刊文献

中文题名：基于协方差ICA分析的多重振荡源分离方法

英文题名：ICA on Auto-covariance Approach to Multi-oscillation Isolation

作者：夏春明[1];郑建荣[1]

机构：[1]华东理工大学机械电子工程研究室,上海200237

年份：2005

卷号：20

期号：12

起止页码：1429

中文期刊名：控制与决策

外文期刊名：Control and Decision

收录：CSTPCD;;EI(收录号：2006069686846);Scopus;北大核心:【北大核心2004】；CSCD:【CSCD2011_2012】；

语种：中文

中文关键词：独立源分析;协方差;振荡检测与诊断;故障诊断

外文关键词：Independent component analysis; Auto-covariance; Oscillation detection and diagnosis; Fault diagnosis

摘要：提出一种新的基于协方差独立源分析(ICA)的多重振荡源分离定位方法.把控制系统中受到振荡干扰的过程数据变换到协方差函数,利用ICA分析的方法进行多重振荡源分离.通过仿真实验对比分析,指出其他时域主元分析(PCA)、时域ICA、协方差PCA等方法的不足,而协方差ICA分析能够准确地分离并定位多重振荡干扰源.仿真结果表明该方法是可行的.
A novel method based on auto-covariance independent component analysis （ICA） for multi-oscillation isolation and localization is proposed. The auto-covariance dataset, calculated from perturbed oscillatory operation data in process control system, is analyzed for the aimed tasks. Simulation test and comparison analysis between simulated sources and analysis results show that ICA on auto-covariance is capable of isolating and localizing multiple oscillatory sources accurately, whilst other approaches, such as those based on time-domain principal component analysis （PCA）, time-domain ICA or PCA on auto-covariance, are lack of such capabilities. Simulation resalos. demonstrate the feasibility of the proposed approach.