基于BP神經(jīng)網(wǎng)絡(luò)的冷軋軋制壓力預(yù)報(bào).pdf

1、分類號UDC：密緞：夠777723武漢科技大學(xué)碩士學(xué)位論文基予BP神綴網(wǎng)絡(luò)的冷乳軋制壓力預(yù)報(bào)ColdRollingForcePredictionBaseOilBPNeuralNetwork談芬芳指導(dǎo)教』『ii姓名：申請學(xué)位綴鄹：論文定稿日期：學(xué)佼授予革位：學(xué)霞授予囂翳：趙剛教授武漢科技大學(xué)材料與冶禽學(xué)院答辯婁員會主席；徐光教授評闋人：董素梅商工余馳斌教授第1I頁武漢科技大學(xué)碩士學(xué)位論文AbstraetRollingprocessisco

2、mplicatedandnon—linearformingprocessInfact，themodelingofrollingforceisanon—linearmappingproblemfrommanyfactorsaffectingrollingforcetollingforceTraditionaltheoreticalformulasofrollingforce，orreturningmodel，aresimplifiedan

3、ddeducedundermanyassumptionsSoitsdifficulttodefinethepropertiesofrollingprocessandcomputetherollingforceaccuratelyBecauseofitsuniqueabilitytoprocessinformation，thetechnologyofneuralnetworksisveryappropriationforthecomput

4、ingofrollingforceHere，amathematicalmodelofrollingforceisfoundedusingthetechnologyofrollingforceSomedevelopmentsaremadeinthemodeling：1Analyzingthepropertiesofrollingprocessofstrip，selectingthemainfactorsaffectingtherollin

5、gforce2PresentinganewmathematicalmodelofaluminumrollingprocessaccordingtoBPneuralnetworks。AfterthestudyingandtestingofthissystemusingVAIproductiondata，thefactisthatthemaximumerrorisbelow6percentIt’sbetterthanthemaximumer

6、roroftraditionalrollingforceformula，whichismorethan40percentTheaccuracyofthemodelisfitforengineeringapplicationTheBPneuralnetworkwhichadoptedLevenberg—Marquardtoptimizedalgorithmhasbeenusedtopredictrollingforceoncoldroll

7、ingmillInthisalgorithm，theparameter塾carlbeadaptivelyadjusted，thenetworkconvergencespeedishigherTheprecisionofrollingforcepredictionhasbeenimprovedIthasprovidedanewpreciseandhighefficiencywayforrollingforcepredictiononcol