搜索
byRodSchollANSYS10.0introducestheabilitytomodelcohesion.Essentiallythisallowsfortheeasycreationofinitialzero-thicknesselementsbetweentheboundariesofsome“glued”components.Thesezero-thicknesselementswillthenopenupinpeelandshearandeven“letgo”whenamaximumstressisencountered.Whatmakesthissuperiortosolidelementswithfailurecriteria,orevencontactelementswithbirthanddeathbasedoncontactresultsandutilizingrestarts,isthattheformulationallowformodelingtheexponentialsurfacepotentialofanadhesiveindeformation.Thus,throughspecifiedcohesionelementmaterialpropertiestheamountofdeformation,andsubsequentload-sharingwithadjacentelementsiscaptured.Thehelpmanualdescribeshowtoinputmaterialpropertiesforthecohesiveregion—andcohesiveelementsaregeneratedprettyeasily:1.DefineelementtypeINTER20Xtomakethecohesiveelementsavailable.2.DefineINTER20Xmaterialpropertiesintermsofmaximumstressandseparationlengthatfailurestress(seematerialsmanual).3.Buildthejointregionline-on-lineandsharenodesalongadefinedseparationsurface.Surfacescanbeidentifiedbyanarea,orbyelementcomponentsoneitherside.4.EnsurethattherearenoMESH200,Shell,Contact,orotherelementssharingnodeswiththecohesivejoint.Theycanbeappliedlater,butcannotbepresentduringthecreationofthecohesiveelementswithoutanerrormessage.5.IssuetheCZMESHcommandreferringtotheelementcomponentsoneitherside,oranareanumberiftwovolumesshareit.ANSYSthenseparatesthenodesandpaintsthesurfaceswiththeINTER20Xelements.That’sit.Fromthere,youcansolvethemodelnormally.Thereareacouplepostprocessingtermsforepoxyseparationandepoxystress.Thehelpmanualisprettyclearontheirusage,socombinedwiththeattachedexamplescript,tchohesion.mac,youshouldfindimplementationsimple.ModelingStiffEpoxyBystiffepoxy,Irefertosomethingabout0.01thick,withastiffnessof105psiandmaximumstressof1,000psi.Suchanepoxyonlyexpands0.0001beforegivingway.Thisscenarioisdifficulttoachieveconvergencewithoutresortingtoaveryfinemesh.For2-Dmodels,themeshdensityrequirementsmaynotbesignificant,butfor3-Dmodelsprepareforlongsolvetimes.Attheloadorsubstepthatcohesiveseparationbeginstooccur,youmayfindnon-convergence.Increasingmeshdensitymaynothaveanyeffectforawhile,untilaminimummeshdensityisachieved.Atthispoint,thedebondinganalysiswillstrugglealong.Aslightlyincreaseddensitybeyondthis,andtheanalysisisquitestable.SeethegraphbelowcreatedbyvaryingMSIZEinthescripttcohesionhard.mac.Thus,a10%increaseinMeshDensity(MSIZE)resultedinabouta10%decreaseinelementedgelength.Alsonotethatforsomeproblems,increasingmeshdensityslightlybeyondtheminimumnecessaryforconvergencegreatlyreducedthenumberofequilibriumiterationsrequired(byafactorof5).However,forotherproblems,thebehaviorwasreversed,whereincreasingmeshdensityincreased(5X)thenumberofequilibriumiterationsrequired!Ifyouhaveanyunexpectedbehavior,trymovingthemeshdensityalittle.Alsonotethatdensityinthepeeldirectionisimportant,butedgeeffectsmayalsoslowconvergencegreatly.Asindicatedbelow,bewareofyourmeshdensityinthetransversedirectionaswell.IexpectconvergencetoimprovewithlaterreleasesofANSYS,andofcoursethismaynotbeanissueatallforyouifyourepoxyissoftandthickenough.Asasidenote,Itriedanalternativeapproachusingaveryfinelayerofelementsatthebondedregion.Thislayerwasjoinedtothesubstrateviabondedcontact.Foragivenfaceelementsize,thenodecountswereactuallysimilarbetweenasolidmeshwithtransitiontofine,andacoarsemeshwithfinesubstratelayer.Ihadhopedthatthenodecountwouldbesmallerforthesubstratelayerapproach,buttheywereinfactquitesimilar.Somegeometriesmightbemoreorlessconducivetonodecountreductionbyinclusionofasubstratebondinglayer.However,themodelwithbondedcontacthadlessdelaminationlength(20%)andanotablylargeregionofloaddistributionthroughtheepoxyintheunzippingdirectionwhencomparedtotheresultsfromthestraightforwardapproach.Soitseemsthesubstratebondinglayer,althoughaninterestingapproach,isnotnecessarilyhelpful.SubstrateLayerApproachStraightforwardApproachSubstrateLayerApproachStraightforwardApproachSubstrateLayerApproachStraightforwardApproachEpoxyStressEpoxyStress