基于模糊神经网络的斜轧建模研究

《基于模糊神经网络的斜轧建模研究》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、的模糊神经网络，及其输入参数的隶属函数。训练数据通过有限元建模和实验数据得到。因此，主要工作包括以下几个方面：(1)在深入了解斜轧无缝钢管生产工艺和理解斜轧理论的基础上，运用ANSYS有限元软件对其进行了建模仿真，采集到了模糊神经网络所需的样本数据；(2)建立无缝钢管斜轧过程的模糊神经网络，选取了隶属函数，将样本数据分成两组：训练数据和检验数据。(3)N用有限元模拟得到的数据对模糊神经网络进行训练。(4)通过用铝管代替钢管进行斜轧规律研究，搭建了实验平台，验证了有限元模拟斜轧的准确性和可靠性。通过基于模糊神经网络斜轧建模研究，建立了工艺参数与质量参数的关系，对斜轧工艺具有指导意义。

2、同时，得到的模糊神经网络的数学模型，为实现斜轧过程自动控制奠定基础。关键词：斜轧；模糊控制；模糊神经网络；建模ABSTRACTCross—rollingisanimportantproductiontechnologyintheproductionofseamlesspipe．Ithasbecomeabasictechnologyandakeymeansforallkindsofseamlesspipeproductionafter100years’development．Butitisoftendifficultyforpeopletobuildtheexactmathmodel

3、byusingclassicmechanicsapproach．ByusingFEM，thecalculatingprecisionincross—rollingprocessishigherthanusingothermethods．Butthesimulationtimeistoolongtobeon-linecontrolmathematicsmodels．ItiSenforcedtocarryonthedevelopmentandresearchonmathematicsmodelsofartificialneuralnet．Thisprojectisastudyofcro

4、ss—rollingprocesstopipe—elongatingoperationintheproductionofseamlesspipe．Themappingrelationshipbetweentechnicalparametersandqualityparametersisobtainedbycalculatingparametersincross．rollingprocess。suchasdeformationstate，mechanicalstate，pipequalityect．Andthefuzzyneuralnetworkandmembershipfuncti

5、onofinputparametersincross．rollingprocessareestablished．TrainingdataaregainedthroughFEMmodelingandexperimentaldata．Therefore，themainworkincludesthefollowingaspects：(1)Basedonadeepknowledgeofproductiontechnologyofcross‘rollingseamlesspipeandcross．rollingtheory,sampledatafortheneedoffuzzyneuraln

6、etworkarecollectedbyANSYS／LS—DYNAformodelingandSimulation．(2)Establishfuzzyneuralnetworkaboutcross—rollingprocessofseamlesspipe；selectmembershipfunction；anddividesampledateintotwogroups：trainingdateandtestingdate．(3)TrainfuzzyneuralnetworkbyusingexperimentaldataandFEMsimulationdata．(4)Buildane

7、xperimentalplatformandverifytheveracityandreliabilityoffinite．elementmodelingofcross．rollingthroughstudyingtheprincipleofcross—rollingbyusingaluminumpipeinsteadofsteelpipe．Basedonthemodelingresearchonfuzzyneuralnetworkofcross—rolling，th