基于典型独立元分析方法的带钢热连轧过程故障检测  

FAULT DETECTION OF HOT STRIP MILL PROCESS BASED ON CV-ICA METHOD

作　　者：张瑞成 李禹亭 梁卫征 熊伟 Zhang Ruicheng;Li Yuting;Liang Weizheng;Xiong Wei(College of Electrical Engineering,North China University of Science and Technology,Tangshan 063000,Hebei,China)

机构地区：[1]华北理工大学电气工程学院,河北唐山063000

出　　处：《计算机应用与软件》2023年第8期87-91,共5页Computer Applications and Software

基　　金：河北省自然科学基金项目(F2018209201)。

摘　　要：针对带钢热连轧过程中故障检测不准确、容易出现错误报警的问题,提出一种典型独立元分析(CV-ICA)的故障检测方法。该方法利用典型变量分析(CVA)求出观测数据的典型变量,有效地解决了数据自相关和互相关性的问题;对典型变量进行独立元(ICA)分解,弥补了ICA不能准确捕捉动态关系数据的缺点,从而有效地提高了故障检测的准确率。通过带钢热连轧过程(Hot Strip Mill Process, HSMP)案例的仿真研究,证明该方法对故障数据的识别精度达到100%,误判率降到0.6%以下。Aimed at the problems of inaccurate fault detection and false alarm in the hot strip mill process(HSMP),a canonical independent component analysis fault detection method(CV-ICA)is proposed.This method used canonical variable analysis(CVA)to find out the typical variables of observation data,which effectively solved the problem of data autocorrelation and cross-correlation.Independent component analysis(ICA)was used to decompose the typical variables,which made up for the disadvantage that ICA could not accurately capture the dynamic relationship data,so as to effectively improve the accuracy of fault detection.Through the simulation of the hot strip mill process case,the result shows that the recognition accuracy of the method for fault data is 100%,and the error rate is reduced to less than 0.6%.

关 键 词：故障检测 典型变量分析 独立元分析 带钢热连轧 动态系统 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术] TP27