摘要
为了解决多变量系统的各个变量之间往往相互影响,且一般不能严格服从高斯分布的问题,采用ICA方法时正常状态下观测的数据进行分析处理,从中提取出统计独立的独立分量,为简化后续分析,对得到的独立分量进行筛选、划分,并分别计算两类统计量:I2统计量和SPE统计量,确定其控制限,与在线数据进行对比,用于监控系统运行.通过一多变量过程仿真实例,证明了这种方法的可靠性,这为ICA应用于监控多变量系统的运行、检测故障的发生提供了有益的思路.
The variable of the multivariate system influences each other and cannot strictly follow a Gaussian distribution in common,so independent components which are statistical independence are extracted from the observed data under normal situation with independent component analysis.In order to simplify the subsequent analysis the obtained independent components can be selected and divided,and two types of statistics,I2 and SPE,are respectively calculated whose confidence bounds need to be identified.,then we can compare them with the statistic values of the online data and monitor the running system.Finally,a given simulation of multivariate process testifies its reliability.It provides a new effective approach to the application of the independent component analysis in process monitoring and fault detection for multivariate system.
出处
《电脑知识与技术(过刊)》
2007年第14期501-502,共2页
Computer Knowledge and Technology
关键词
独立分量分析
过程监控
统计量
Independent component analysis
Process monitoring
Statistics