汽轮机叶片强度可靠性分析的响应面方法研究

《汽轮机叶片强度可靠性分析的响应面方法研究》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、华北电力大学（保定）博士学位论文汽轮机叶片强度可靠性分析的响应面方法研究姓名：段巍申请学位级别：博士专业：热能工程指导教师：王璋奇20090401华北电力大学博士学位论文摘要AbstractTherearemanystochasticparametersinsteamturbinebladedesign，manufacturing，installationandoperation，whichresultintherandomnessofstructuralresponse．Inthetraditionala

2、nalysismethod，itissupposedthattheparametersofthebladearedeterministic．Itisdifficulttoexplainwhythebladeisfailedinnormaloperationwhenitisdesignedcorrectlybythetraditionaldeterministicmethodandevenmoredifficulttoevaluatequantitativelyhowmuchthebladeissafe．So

3、，itisnecessarytotaketherandomparametersintoaccountandcarryoutthestrengthreliabilityanalysisanddesignforsteamturbineblade．AstheperformancefunctioninbladestrengthreliabilityanalysisCannotbeexpressedasananalyticalformintermsofbasicrandomvariables，thetradition

4、alprobabilityanalysisapproachcannotbedirectlyapplied．So，allapproachwhichcombinesfiniteelementmethod(FEM)，responsesurfacemethod(RSM)andMonteCarlosimulation(MCS)isputforwardtosolvethebladestrengthreliabilityanalysis、Ⅳitllimplicitperformancefunction．Basedonth

5、ebladefiniteelementparametriealmodelandexperimentaldesign,thetwokindsofresponsesurfacemethods，thatalemultinomialresponsesurfacemethod(MRSM)andartificialneuralnetwork(aNN)，alerespectivelyappliedtoconstructtheapproximateanalyticalexpressionbetweenthebladestr

6、uctureresponsesandrandomvariables，whichacts弱asurrogateofthefiniteelementsolverforestimatingtheperformancefunction．1f1豫nthesurrogate，whichiSobtainedbyMRSMorANN，isusedformostofthesamplesneededinMonteCarlosimulationmethod．Furthermore，thestatisticalparametersa

7、ndcumulativedistributionfunctionsofthebladeresponses，such醛maximumdeflection，maximumstress，staticfrequencyanddynamicfrequency,areobtainedbyMonteCarlosimulation．BasedonFEM·RSM-MCSapproach,thestatisticstrengthreliabilityanalysisandvibrationreliabilityanalysis

8、oftheezlualCroSS—sectionstraightbladeandthevariablecross-sectiontorsionbladearecarriedoutrespectively．Meanwhile，theanalysisresetsinducedbythetwodifferentresponsesurfacemethodsMRSMandANNarecomparedrespectively