基于多块修正ICA算法的分散式非高斯过程监测方法  

Decentralized Non-Gaussian Process Monitoring Scheme Based on Multi-block Modified ICA Algorithm

作　　者：万新春 童楚东[1] 史旭华[1] WAN Xinchun;TONG Chudong;SHI Xuhua(School of Information Science&Enginnering,Ningbo University,Ningbo,315211,China)

机构地区：[1]宁波大学信息科学与工程学院,浙江宁波315211

出　　处：《信息与控制》2020年第4期464-471,共8页Information and Control

基　　金：国家自然科学基金资助项目(61773225,61803214)。

摘　　要：针对分散式非高斯过程监测方法通常都忽略测量变量间整体性的问题,为了同时提取测量变量间的局部特征和全局特征,提出了一种基于多块修正ICA(multi-block modified ICA, MBMICA)算法的分散式非高斯过程监测方法.首先,通过修正FastICA迭代算法得到MICA(modified ICA)算法;然后,利用整体测量变量的解混向量引导各个子块中独立成分的分解,得到能够考虑测量变量间整体性的MBMICA算法,并利用该算法实施分散式非高斯过程监测;最后,通过仿真对比实验验证了MICA算法的可行性及该方法相比于其它分散式过程监测方法的优越性.Given that decentralized non-Gaussian process monitoring approaches usually don't take into account the integrality of measured variable as a whole,we present a novel decentralized non-Gaussian process monitoring method to extract simultaneously the local feature and the global feature of the measured variables.Firstly,we modify the original FastICA iterative algorithm to derive a modified ICA algorithm(MICA);And then the extraction of block independent components can be oriented by the de-mixing vectors calculated from the measured variables as a whole,a multi-block MICA(MBMICA)algorithm that considers the integrality of the measured variables is thus obtained,based on which decentralized non-Gaussian process monitoring can then be implemented;Finally,we validate the feasibility of the MICA algorithm,and the superiority of the proposed method over other decentralized monitoring approaches through comparisons.

关 键 词：分散式过程监测 独立成分分析 MBMICA(multi-block modifiedICA) TE(Tennessee Eastman)过程 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]