欢迎来到天天文库
浏览记录
ID:39402400
大小:657.00 KB
页数:62页
时间:2019-07-02
《基于动态递归模糊神经网络的微生物发酵过程软测量方法研究》由会员上传分享,免费在线阅读,更多相关内容在行业资料-天天文库。
1、江苏大学硕士学位论文基于动态递归模糊神经网络的微生物发酵过程软测量方法研究姓名:张瑶申请学位级别:硕士专业:控制理论与控制工程指导教师:孙玉坤20100612江苏大学硕士学位论文ABSTRACTMicroorganismfermentationengineeringisacomplexbiochemicalreactionprocesswhichhashighnonlinear,time—varyingandhysteretic,andwhichinternalmechanismiSverycomplicated.Thetr
2、aditionalmeasurementmethodsaredi硒culttomeasuresomekeyvariablesonline,suchaseellconcentration,substrateconcentration.productionconcentration,whichmakesthewholefermentationprocesstooptimizecontrolbecomesverydifficult.soft-sensortechniqueiSoneofthemosteffectivewaystos
3、olvethisproblem.Thisarticletakelysinefermentationprocessasresearchobject,dynamicrecurrentfuzzyneuraInetworksoft—sensormodeliSbuildtopredictthethreeimportantvariables(eellconcentration,substrateconcentration,productionconcentration)infermentationprocessonthebasisoft
4、hesoft.sensortheory.Thesimulationresultsshowthatthesoft-sensormodelCanaccuratelypredictthekeyvariables.anditiSwithpreferablestabilityandCanpredictcorrectlyunderthesituationofdisturbance,whichprovidethepreconditionforoptimizingcontrolinthefermentationprocess.Theconc
5、reteworkiSasfollowing.Firstly,onthebasisofreadingliteraturesandstudyinglysinefermentationexperiment,accordingtoacurveofmicrobemetabolizinginthefactualfermentingprocess,thesoft-sensormodelsofbasedonFuzzyNeuralNetworkanddynamicfuzzyneuralnetworkarerespectivelyestabli
6、shed.Thenthevariablesinthefermentingprocessoflysinesuchaseellconcentration,substrateconcentration,productionconcentrationarepredicted,andtheperformanceoftwomodelswerestudiedandcompared.Secondly,accordingtotheresearchofassistantvariableselectionanddatapretreatmentme
7、thodinsoftsensor,thekemelprinciplecomponentanalysisisadoptedt0fixassistantvariableandexperimentaldataareprocessedwiththemodifiedmedianminimumdistancealgorithm(MMMD)basedonMahalanobisdistance.Thirdly,谢t11thedeficiencyofinitialvaluethatisextremelysensitive,easytofall
8、intolocalminimum,afuzzyCmeansclusteringalgorithmbasedondatafieldisproposed,thusconquertheblindnessofrandomselectionofinitialcentersSOastoquickent
此文档下载收益归作者所有