欢迎来到天天文库
浏览记录
ID:36455017
大小:2.54 MB
页数:64页
时间:2019-05-10
《基于MICA方法的间歇过程监控研究》由会员上传分享,免费在线阅读,更多相关内容在学术论文-天天文库。
1、中国石油大学(华东)硕士学位论文基于MICA方法的间歇过程监控研究姓名:张晓玲申请学位级别:硕士专业:控制理论与控制工程指导教师:田学民20080501独立成分,并利用,2和SPE统计图监测过程中是否有故障发生。FS.MKICA方法不仅能提取间歇过程中的非线性特性,而且减少了基于全部样本建模的计算代价,对青霉素发酵过程的监控结果显示,该非线性算法比线性MICA方法检测故障时更灵敏。关键词:间歇过程,故障检测与诊断,多向独立成分分析,自适应算法,非线性BatchProcessMonitoringBased011MICAMethodsZhangXiaoling(ControlTheoryandCo
2、ntrolEngineering)DirectedbyProf.TianXueminAbstractprocesseshavebecomemoreandmoreimportantinmodemindustrialprocesses.Inensuringthesafetyandstabilityofbatchprocessesandhighqualityfmalproduct,on-linemonitoringandfaultdiagnosisinbatchprocessesemergeasanessentialandimportanttask.Asthedevelopmentofon-line
3、measurementinstrurnentsandcomputertechnology,largeamountsofprocessvariables’datacanbecollectedmoreeasilythanbefore.ThedataCanbeanalyzedtosupervisetheprocessbehavior,byminingthevaluableinformationandresources.Multi-wayprincipalcomponentanalysis(MPCA)andmulti-waypartialleastsquares(MPLS),whichassumeth
4、atthevariablesmustsubjecttothenormaldistributionconditionandonlyutilizethesecond—orderstatisticalinformation,areusedmostwidelymultivariatestatisticaltechniqueinbatchprocessesmonitoring.Multi-wayindependentcomponentanalysis(MICA),onetypeofmultivariatestatisticalmethodbasedonICAtechnique,isrecentlydev
5、elopedtoapplytothebatchprocessesmonitoring.ThismethodCantreat、析ththree—waydataofbatchprocessesmoreeffectivelybecauseitutilizesthehiIgh—orderstatisticalinformationandavoidstheassumptionofGaussiandistribution.Inaddition,theextractedlatentvariablesbyMICAarestatisticallyindependentwhileprincipalcomponen
6、tsgeneratedfromMPCAalemerelyde-correlated.Therefore,theindependentvariablesorcomponentscandescribetheprocessescharacteristicmoreintrinsicallythanMPCAorMPLS.Inthiswork,MICAbatchmonitoringmethodisdiscussedandconsideringthecharacteristicsofbatchprocesses,twonewkindsofmonitoringmethodsareproposedbasedon
7、MICA.Inviewofbatch—to—batchvariationinmostindustrialbatchprocesses,anadaptiveMICAmethodisproposedtocapturethedynamicvariationamongdifferentbatches.ThisapproachfirstestablishesanMICAmodelbasedonthehist
此文档下载收益归作者所有