摘要
生化工过程中存在大量测量变量,这些变量一般不是相互独立的,而是由少数必要的潜隐变量驱动,这些潜隐变量通过独立成分分析方法(ICA)抽取出来;针对现有的Infomax(信息极大)ICA算法收敛速度慢的问题,引入四阶统计去相关的混合学习规则,结合加权协方差阵的非对角元素最小化,提出了一种改进Infomax算法,将其用于生化工过程故障的提取,并通过在TE(TennesseeEastman)模型上仿真,结果表明该方法改善了原有算法的收敛性能,盲源分离效果良好。
There is a great deal of geodesic variable in the bio-chemical process, These variable isn't independent mutually generally, But drive by a handful of potential variable of necessities, These potential variable is sampled pass Independent Component Analysis method (ICA) . Aim at the problem of convergent speed slowly to existing Infomax (information maximum) ICA algorithm, Import the mixture study rule of statistics into four ranks go to related, Combine to the non--cross element minimum of weight covariance matrix, Put forward a kind of improved Infomax algorithm, Using for distill fault of the bio--chemical process, and pass on the TE (Tennessee Eastman) model to imitate. The result indicate that the method improved astringency of original arithmetic, The blind source separation's effect is good.
出处
《计算机测量与控制》
CSCD
2006年第12期1649-1651,共3页
Computer Measurement &Control
基金
国家科技攻关计划项目(2001BA204B0103)
关键词
独立分量分析
改进Infomax算法
统计过程监控
故障检测
independent component analysis
improved Infomax arithmetic
statistics process supervision
fault detection