ANSYS-R15流体设计-培训资料

ANSYS流动仿真产品的最新动态王晓ANSYS日本技术本部总经理ANSYS亚洲专家组组长1.建模和网格2.高性能计算3.求解器的最新功能4.各种物理模型5.旋转机械仿真的最新功6.使用的便利性内容SimulationDrivenProductDevelopment仿真驱动产品开发建模和网格高效的前处理工具是”仿真引导产品开发”成功的关键建模工具的性能•更快的CAD形状导入–Multi-threading/multi-processingimport–Parallelcontactdetection•更快的建模–SharedTopologies–Modelgenerationtime–Selectivepartupdating–DatabaseLoadandSave(ANSYSCADInterfaces&DesignModeler)形状复制•复制功能可加快建模和网格生成–Instancingsupportformulti-bodyparts•Allpartsofapatterncansharespecifications–Meshingofbodyinstances•CreatepatterninDM•Askwholepatterntoinheritspecificationsoffirstpart(ANSYSCADInterfaces&DesignModeler)形状的拓扑共享-TopologySharing•ImprovedflexibilitywithcustomSharedTopologyforbodies(ANSYSDesignModeler)Threebodies,onepartSharedTopologyforconformalmeshSlidingcontactsurface(nosharedtopology)Mappedfacemeshspecificationforcyclicsymmetrysimulation网格生成的性能•更少的时间和内存forpatch-conformingtetmeshing–Significantimprovementsin•Advancingfronttrianglesurfacemeshing•ProximityadvancedsizefunctionBaselinecaseshowingtimeandmemoryreductionsfortypicalfluidsmodel–moreimprovementexpectedbyfinalrelease(ANSYSMeshing)使用便利性的改善•确定网格生成出错的位置–Betterindicationtouseroflocationoffailure–ReturnofmeshuponfailureforallmethodsWhenmeshingfailsinsurfacemeshing,validsurfacemeshisreturned.Edgecoloringcanindicatewhereproblemis.Easierinspectionofproblemswithdifferentlycoloredmeshforregionsinwhichmeshingfailedorisout-of-date(ANSYSMeshing)六面体网格的生成•六面体网格生成的一大改善Mappedmeshingofcircularloops(allsidefacesneedtobemapped)Useofinflationallowsmappedcylinderwithgoodmeshquality(ANSYSMeshing)FluentMeshing助力高效仿真流程①②⑥③④⑤⑥•Newmodeldiagnosticsandgapclosing/holefixingtools–Guidancetofindassemblyproblems–Multipletoolsprovidedtoaddressproblems…usingnewfastgraphictoolstofixthegapGapsbetweenobjectsareidentifiedandcanbeclosedwithstandardgapclosingtoolor…(FluentMeshing)网格的诊断和修改Newholefixingtoolguidestheusertotheholesinthemodel,whichtheusercaneitherauto-close,ormanuallyclosewithnewfastgraphictools网格质量•网格和形状的诊断(withwrapper)–Addressvariousproblems•Deviationfromoriginalgeometry•Acuteangle•局部的网格再生成–Addressproblemsin•Meshquality•Preparationforprismmeshing(FluentMeshing)体网格性能的提高•PrismLayers–2-3xspeed-upfor10-20layers–PostIgnoreincludedinworkflow•Tetmeshing–Commondatastructuresforfurtherperformanceimprovements•HexCore–~2xspeed-up•ParallelMeshing–TetandPrismusingpre-decomposeddomains(sew),90%scalability•FasttransfertoFluentSolver–FullParallelsupportVirtualCFDVolumeMeshFluentSolver(FluentMeshing)高性能计算ANSYSworkscontinuouslytooptimizeparallelscalabilityandperformanceofourCFDsolvers,allowingcustomerstoleveragethelatesthardwareadvancementsandarriveatanaccuratesolutionfast.并行计算-CFX•SignificantefforttoimprovesolverHPCscalinginCFX–Excellentscalabilitynowseenforequationassembly•Upto90%efficiencyover2048cores–ImprovementshavealsobeenmadetotheAMGlinearsolverSolverwallclockspeed-upon150Mnodeductcaseover10iterationson2048cores,bottomimagesspeed-upbrokendownintoequationassemblyandlinearsolverWallclocktimereductiononsix-stageaxialcompressor(14domain90Knodesperdomain)并行计算-FLUENTHighsolverscalabilityatlargecorecounts•~89%efficiencyfor150Mcellcaseat15360cores•OtherbenchmarkcasesshowsimilartrendsMorescalablediscretephaseparticletracking•Over2xfor512-wayparallel010020030040050060070003072614492161228815360RatingNumCores150Millioncells14.5.015.0.0Ideal01000200030004000500060007000163264128256384512Rating246,000cells,1millionparticlesHybridMPI2Domain并行的数据I/O和计算的开始•MoreefficientparallelI/Oandstartup–Casereadtimereducedsignificantlyathighcorecounts–Start-uptimefor8192-wayparallelreducedfrom30minutesto30secondsEffectiveconfigurationofparallelprocesses–Usedifferentnumberofprocessesformeshingandsolvemodes0100020003000400050006000700010242048409661448192921610240Timeinseconds150Mcellcasereadtime15.0.014.5.0(Fluent)求解器·计算方法的改进Robust,accurate,andadvancedCFDsolvertechnologyfromANSYSprovidesthebasisforfastassessmentandoptimizationofdesignsandtheirperformance.AdjointSolver(形状优化)•Expandedadjointsolvercapabilities–Supportforlargerscaleproblems•Upto30millionscells–Abilitytosolvetheadjointequationforenergy•Observablescanbedefinedasintegralsofheatfluxandtemperature(Fluent)BeforeAfterMeshMorpherOptimizer(形状优化)•Improvedcontrolpointselection–SelectmultiplecontrolpointswithRMBclick–SelectcontrolpointbasedonI,J,Kcoordinates•Significantproductivitygains(β)–NewNEWUOAalgorithmrequiresfewerdesigniterations(Fluent)移动·变形网格(CFXMoving&DeformingMesh)•Improveddefaultsformeshstiffnessforbetterrobustnessingeneral–Stiffnessexponentdefaultchanged,forlessabruptchangeinstiffness–Referencevolumeincludedbydefault•Improvedslidingmeshonsurfacesofrevolution–Improvedrobustnessforradialmachines–ReducedparallelmemoryoverheadExampleshowingmeshstiffnessresultingwitholddefaults(top)vs.newdefaults(bottom)•Increasedtemporalaccuracy–2ndordertemporaldiscretizationwithlayeringandremeshing•Improvedaccuracyandrobustnessformeshsmoothing–Node-basedsolverfordiffusionsmoothing–Linearlyelasticsolidsmoothing•Increasedflexibilitywithlocalre-meshing–Detectionandre-meshingofattachedboundaryla