文献类型：期刊文献

中文题名：基于全变量信息的ICA子空间故障检测方法

英文题名：ICA Subspace Fault Detection Method Based on Full Variable Information

作者：杨文翠[1];王帆[2];谭帅[1];侍洪波[1]

机构：[1]华东理工大学化工过程先进控制和优化技术教育部重点实验室,上海200237;[2]北京科技大学自动化学院,北京100083

年份：2019

卷号：26

期号：1

起止页码：12

中文期刊名：控制工程

外文期刊名：Control Engineering of China

收录：CSTPCD;;北大核心:【北大核心2017】；CSCD:【CSCD_E2019_2020】；

基金：国家自然科学基金(61374140);国家自然科学基金青年基金(61403072)

语种：中文

中文关键词：独立主元分析;子空间;信息缺失;故障检测

外文关键词：ICA;subspace;information loss;fault detection

摘要：独立主元分析(Independent Component Analysis,ICA)应用于化工过程故障监测,首先是对原始空间进行降维得到低维空间,但有些微小故障不能被检测,造成信息缺失,影响监测结果。针对这一问题,提出一种新的子空间监控方法。根据变量和独立主元空间(Independent Component Subspace,ICS),残差空间(Residual Subspace,RS)关联程度,把相似的变量划分到同一个空间,保留全部的过程变量,并分别建立监控模型。最后通过数值例子及田纳西伊斯特曼过程验证该方法的有效性。
Independent component analysis has been widely used in chemical processes,but after the dimension reduction for the original data space is done,as a consequence,some minor faults cannot be monitored by the main space,or cannot be detected by the residual space,leading to information loss,which will greatly affect the performance of process monitoring.To solve this problem,in this paper,a novel method based on ICA is proposed.According to the similarity extent between each variable and dependent principal component space ICS or residual space RS,the similar variables can be divided into the same space,in this way the whole process variables can be preserved without loss of information.Then the monitoring models are established respectively in each subspace.At last,the proposed method is applied to a numerical example and Tennessee Eastman process to evaluate the monitoring performance.