资源描述:
《efg法与其拓扑优化的gpu并行计算分析》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、AbstractElement-freeGalerkin(EFG)methodnotonlyhastheadvantageofgettingridoftheshacklesofthegrid,buthasgoodstabilitycalculationandfastconvergenceandhighprecision,etc.Another,EFGmethodtopologyoptimizationcaneffectivelysolvethereconfigurablemeshandotherissues,whicharearousedb
2、yfiniteelementmethodtopologyoptimization.Moreimportantly,EFGmethodtopologyoptimizationcaneffectivelyeliminatecheckerboardphenomenon.Therefore,moreandmoreattentionofscholarshasbeenattractedbyEFGmethodanditstopologyoptimization.However,inlargescaleproblems,theresearchandappl
3、icationofEFGmethodanditstopologyoptimizationisseverelyconstrainedbythedrawbackthatEFGmethodhaslowcomputationefficiency.Atthesametime,GraphicProcessingUnit(GPU)hasbeenwidelyappliedinmanyfieldswiththerapiddevelopmentofparallelingcomputing.Therefore,accordingtotheshortcomingo
4、flowefficiencyandtime-consuming,thispapercarriesouttheGPUparallelcomputingresearchofEFGmethodanditstopologyoptimization.Themainresearchcontentsaresummarizedasfollows:Firstly,thethree-dimensionEFGstructureanalysismethodwasstudiedandapplied.Inthetwo-dimensional,weintroducedt
5、riangularelementtoinsteadoftheEFGtraditionalalgorithm’squadrilateralelements.Meanwhile,weintroducedtetrahedralelementtoinsteadofhexahedralelementinthethree-dimensional.Ontheotherhand,inordertoachievetheintegralcalculationofEFGmethod,theHammerintegralwasusedtoreplacetheGaus
6、sintegralwhichistheintegralmethodoftheEFGtraditionalalgorithm.Andtherelevantformulawasdetailedlyderived.Another,thefeasibilityofthealgorithmwasverifiedbytheanalysisofnumericalexamples.Secondly,theGPUparallelcomputingofEFGmethodwasinvestigated.Basedonthethoughtofinteracting
7、nodespairs,ajointassemblyschemeanditsGPUparallelalgorithmofthegeneralstiffnessmatrixandthegeneralpunishmentstiffnessmatrixandthepreprocessingmatrixwasraised.Moreover,anacceleratedalgorithmofEFGmethodwhichcouplesGPUandPreconditionedConjugateGradient(PCG)methodwaspresented.T
8、hederivationprocessofcorrelationformulaandtheflowchartofthisacceleratedalgorithmweregiven