基于神經網絡的軋制力預報研究.pdf

1、東北大學碩士學位論文基于神經網絡的軋制力預報研究姓名：張海淑申請學位級別：碩士專業(yè)：計算機應用技術指導教師：傅仲逑20050701查!!壟堂塑主蘭垡墮查!!!!!!旦ResearchofRollingForcePredictionbasedonNeuralNetworkABSTRACTThethesisadoptsthehighprecisionrollingmillasthebackgroundforresearch。BPneural

2、networkwasformerlyintroducedintollingforcepredictionsysteminordertomeetpracticaldemandonthecontrolprecisionofrollingCOUrSeThethesisonlyfocusesworkontheapplicationofIGLSneuralnetworkalgorithmtothefieldSOastoimprovetheprec

3、isionofrollingforcepredictionsystemFirstlyrelevantstudyiscarriedoutaroundtherollingforcemodelSomeconclusionsarederivedbyanalyzingsuchtheories011rollingasthebasicmathematicmodelforthethicknessCOntrolandPHcxtrveItiswellkno

4、wnthatrollingforceisoneofthecrucialparametersintheCOUrseofrolHngFurtherresearchononeofcommonrollingforcemodelsnamedSIMSindicatesthattheprecisionofrollingforcepredictionsystemisaffectedbymanyelementssuchastheinoutbeltwidt

5、h，rollerradius，thespeedofbeltatexitandtemperatureandSOonSecondlyaseriousofanalysesmadeonseveralcommonneuralnetworkmodelsandalgorithmslayafoundationforfurtherimprovementAimingatthelimimfionoftraditionalrollingforcemodel，w

6、eestablishIGLSmultilayerfeedforwardneuralnetworkmodelforsinglemillbydefiningthetopologyarchitectureandparametersofneuralnetwork，andaccordinglyselecting，pretreatingandrenewingappropriatespecimendataOnthebasisofpreviousres

7、olutionwithneuralnetwork，consideringthefeaturesofrollingforceshowedintheconrseofrolling，weputforwardakindoffavorableresolutionwithneuralnetworkandthereforetestandverifythefeasibifityofitThirdlyoveralldesignforarchitectur

8、eofsoftwareusedinthesystemisfulfilledwithUMLlanguageaccordingtotheprogrammingideaofobject—orientedThefunctionsofsuchmodulesasinitializationofnetworkmodel，selfstudyofnetworkandreal_timepredictionareexplicatedseriatimThewh