基于CUDA的接觸碰撞問(wèn)題的并行算法研究.pdf

1、南京航空航天大學(xué)碩士學(xué)位論文基于CUDA的接觸碰撞問(wèn)題的并行算法研究姓名：李楊申請(qǐng)學(xué)位級(jí)別：碩士專(zhuān)業(yè)：機(jī)械設(shè)計(jì)及理論指導(dǎo)教師：陳國(guó)平201012基于CUDA的接觸碰撞問(wèn)題的并行算法研究IIABSTRACTContactimpactproblemsunderlargedefmationareinvolvedwithgeometricnonlinearitymaterialnonlinearitycontactnonlinearity.Be

2、causeofthecomplexitythesolutionprocessisverytimeconsuming.Thusitisdifficulttoobtainsatisfactyresultsintraditionalserialcomputers.IndertoimprovetheaccuracyefficiencyofthesolvingthispaperpresenttheCUDAbasedparallelalgithmo

3、fthecontactimpactproblemsresearchtheapplicationoncontactimpactproblemintransientanalysis.Contactimpactproblemsaregenerallycarriedoutwithfiniteelementmethod.Thispaperstudiesthesolutionprocessasfollows:1.Establishthefmatof

4、nonlinearfiniteelementsolutionofcontactimpactproblemsunderthefullLagrangefmat.Inthecalculationfmatallphysicalparameterstakeinitialconfigurationasreference.TherefethefullLagrangemethodtakeslessamountofcalculationcomparedw

5、ithupdatedLagrangemethod.Thesolvingprocessusesexplicitintegrationwhichavoidsthestiffnessmatrixinversionthusresultsinhighersolutionefficiencycomparedwiththeimplicitintegration.2.Studyisconductedoncontactsearchingalgithmco

6、ntactfcealgithm.Indertocomputethecontactfcecontactsearchmustbeimplemented.Thispaperfocusesontheapplicationofpositioncodealgithminwhichthechaintablesareciteddevelopedintheglobalsearch.Thetechniqueofchaintablesisaneffectiv

7、emethodtosearchthecontactpairs.VectwhichisakindofcontainerinthelanguageofCisintroducedtodecidethehittingpointestablishtheappropriatedatastructure.Aftercontactsearchingcontactfcewillbecalculatedaccdingthecheckconditiondef

8、initecondition.Lagrangemultipliermethodpenaltyfunctionmethodaretwobasicmethodsfthecomputationofcontactfce.Duetothecompatibilityeasinessofprogrammingpenaltyfunctionismostwidelyused.Thispaperalsodiscussestheapplicationofde

9、fensenodealgithminthecomputationofcontactfce.Defensenodealgithmwhichiscompatiblewiththeexplicitsolutioncanimprovethesolutionprecision.Howeverduetothecomplexityofthesolutionfmitishardtoimplementwithparallelmethod.3.CUDAis

10、akindofnewparallelcomputingarchitecturesoftwarehardwarebasedonGPU.Threadsareusedtocontroltheimplementoftheprogram.BecauseofthegoodparallelismoffiniteelementmethodparallelcomputingtechnologybasedonCUDAcanbewellappliedonfi

11、niteelementanalysis.Whatthepaperfocusesonareasfollows:TheestablishingofCUDAbasedparallelcomputingplatfmParallelgenerationprocessofstiffnessmatrixmassmatrixofthesystemParallelsolvingoftheequationofthesystem.Theoptimizatio

12、nofparallelcodeEvaluationcriteriaofparallelsolvingefficiency.IndertofacilitatetheinteractionbetweenusersthefiniteelementprograminterfacedesignbasedonMFCisalsostudiedtointegratevisualizethefiniteelementprogram.Theinterfac