08/05/2017 ModellingcrackpropagationusingXFEM

+31(0)418644699 [email protected]

SimuleonFEABlog
Modellingcrackpropagation
usingXFEM
PostedbyDolfBroekaartonMar7,201710:00:00AM

Findmeon:

Tweet 1

InapreviousblogIshowedhowtomodelastationarycrack
andcalculatetheJintegraltodeterminewhetherthecrack
propagates.Abaqusoffersdifferenttechniquestosimulate
crackpropagation,includingsurfaceandelementbased
cohesivebehaviourandthevirtualcrackclosuretechnique.
WhenusingoneofthesemethodswithconventionalFEM,the
locationofthecrackneedstobeprescribedbeforehand.When
theeXtendedFiniteElementMethod(XFEM)isused,thisis
notnecessary.Inthiscase,enrichmenttermsareaddedtothe
normaldisplacementinterpolation,soacrackwithinan
elementcanbedescribed.InthisblogIwillexplainhowto
modelcrackpropagationusingthesurfacebasedcohesive
behaviourapproachandXFEM.

Asanexample,loadingwillbeappliedtoagear,similartothe
stationarycrackexample.Onlyinthiscase,nocrackwillbe
presentinitiallyanditwilldevelopbasedontheloading.

Meshing
Meshingismuchsimpler,becauseitisnotnecessaryto
partitionaninitialcrackorcircularregionsaroundthecracktip.
Partitioningisonlyusedtoallowafinermeshintheregion
wherethecrackwilldevelopcomparedtotherestofthegear.

http://info.simuleon.com/blog/modellingcrackpropagationusingxfem 1/11
08/05/2017 ModellingcrackpropagationusingXFEM

Linearbrickelementsareused(quadraticelementsarenot
availableforXFEM).Halfofthegearismodelled,forreasons
ofsymmetry.

Definingdamagedamage
initiation
TomodelcrackpropagationinAbaqusbothdamageinitiation
anddamageevolutionneedtobetakenintoaccount.Different
criteriaareavailablefordamageinitiation,inthiscasethe
maximumprincipalstresscriterionwillbeused.Comparedto
criteriabasedonnominalstressorstrain,usingtheprincipal
stressorstrainhastheadvantagethatthecrackplanecanbe
perpendiculartothedirectionofthemaximumprincipalstress,
makingitsolutiondependent.Damageinitiationisdefinedas
partofthematerialproperties,usingdamageforTraction
Separationlaws=>MaxpsDamage.Withthisoption,
damagewillinitiatewhenthemaximalprincipalstressexceeds
thevaluegiven.

Definingdamagedamage
evolution
Byspecifyingdamageinitiation,youdonotyetdefinehowthe
materialchangesduetodamage.WithinAbaqus,damageis
modelledusingascalardamageparameter,D.Thiscanrange
between0(nodamage)and1(completefailure).Thestress
thatwouldhavebeentherewithoutdamageismultipliedby(1
D)tocalculatethestressincludingdamage.Withoutdamage
(D=0)thisleadstotheundamagedresponse,withcomplete
failure(D=1)thestressis0andinbetweenafractionofthe
stresswillremain.

http://info.simuleon.com/blog/modellingcrackpropagationusingxfem 2/11
08/05/2017 ModellingcrackpropagationusingXFEM

Inthisexample,damageisdefinedusingatractionseparation
lawanddamageevolutionisasuboptionofMaxpsDamage.

Eitherthemaximaldisplacementorthefractureenergy,which
istheareaunderthecurveinagraphoftractionversus
separation,mustbespecified.Differentoptionsarepossibleto
specifythesofteningbehaviour:howthetractionseparation
graphgoesfromthepointattheonsetofdamagetothe
completelyfailedstate.Inthiscaselinearsofteningisused,
correspondingtoastraightlineinthetractionseparation

http://info.simuleon.com/blog/modellingcrackpropagationusingxfem 3/11
08/05/2017 ModellingcrackpropagationusingXFEM

graph.Itispossibletotakeintoaccountmodemixing,withthe
BKlaw,powerlawortabularlyspecifieddata.Alternatively,
modeindependentbehaviourcanbespecified.

Definingdamagedamage
stabilization
Simulationsincludingdamageevolutionoftenleadto
convergencedifficultiesbecauseofthesofteninginthe
materialmodel.Abaqusallowstheuseofviscous
regularizationtostabilizetheresponseduringdamage.For
sufficientlysmalltimesteps,thetangentstiffnessmatrixwill
thenbepositivedefinite.Aviscositycoefficientcanbespecified
asasuboptionofMaxpsdamage.Itshouldbechoseninsuch
awaythattheinfluenceofthestabilizationonthefinalresults
issmall.Tocheckthis,theoutputALLVD(viscousdissipation)
canbecomparedtoALLSE(strainenergy).IfALLVDisnot
smallcomparedtoALLSE,theviscousstabilizationislikely
influencingresults.Playingaroundwiththeviscositycoefficient
canhelpobtainareasonableresultwithinareasonable
amountoftime.

Definecrack
Apartfromdefiningwhenthematerialwilldamageandhowit
willbehaveafterdamageisinitiated,theregionwhereacrack
canoccurmustbespecified.Thisistheregionwherethe
enrichmenttermswillbeadded.Choosing
Special=>Crack=>Createintheinteractionmoduleand
selectingthetypeXFEMintheCreateCrackdialogboxthat
appears,allowsyoutoselectacrackdomain.Inthiscase,the
cellwherethecrackwilldevelopischosen.
http://info.simuleon.com/blog/modellingcrackpropagationusingxfem 4/11
08/05/2017 ModellingcrackpropagationusingXFEM

Theallowcrackgrowthboxshouldbechecked,toallowthe
cracktopropagate.XFEMcanalsobeusedforstationary
cracks,allowingcontourintegralssuchastheJintegraltobe
calculatedwithlessmeshingeffort.Thisisnotdiscussedhere.
Inthisexample,noinitialcrackisdefined:itwilldevelopbased
onthedamageinitiationcriterion.Ifaninitialcrackiswanted,it
isveryeasytodefinewithXFEM.Aseparatepartrepresenting
thecrack(withoutpropertiesormesh)canbeinstancedinto
theassemblyandmovedtothecorrectposition.Bychoosing
thispartascracklocation,thecrackisdefined.Thecrackdoes
notneedtobealongtheelementedges.Infact,theXFEM
methodworksbestifthecrackcrossesthroughtheelement.
Frictionlesssmallslidingcontactcanbedefinedforthecontact
betweenbothsidesofthecrack.

Solutioncontrols
Toaidinobtainingaconvergedsolution,thesolutioncontrols
canbemodified.Fromthestepmodule,select
other=>generalsolutioncontrols=>edit=>stepinwhich
solutioncontrolsaretobeedited.Presscontinuewhena
warningmessageisdisplayed.Inthetimeincrementationtab,
discontinuousanalysiscanbechecked.ThisallowsAbaqus
todomoreiterationsbeforecheckingwhetherthesolutionis
goinganywhere.InthefirstMoretab,theparameterIAcanbe
increasedfromthedefault5,toallowAbaqusmoreattempts
beforeabortingthesimulation.Iflargecutbacksarerequired,
increasingthenumberofattemptsisuseful.

http://info.simuleon.com/blog/modellingcrackpropagationusingxfem 5/11
08/05/2017 ModellingcrackpropagationusingXFEM

Requestoutput
http://info.simuleon.com/blog/modellingcrackpropagationusingxfem 6/11
08/05/2017 ModellingcrackpropagationusingXFEM

ForasimulationusingXFEM,itisimportanttorequestPHILSM
asoutput.Thisisthesigneddistancefunctiontodescribethe
cracksurfaceusingthelevelsetmethod.IfPHILSMis
requested,Abaqusautomaticallycreatesanisosurfaceview
cutbasedonthisoutput,whichshowsthelocationofthecrack.
Ifitisnotrequested,thecrackwillnotbevisibleandresults
shownwillbecounterintuitive.STATUSXFEMisalsospecificfor
XFEM.Itgivesthestatusoftheenrichedelements,itis0.0if
theelementisundamaged,1.0iftheelementiscompletelycut
through(notractionforcesremaining)andavalueinbetween
iftheelementisdamagedbutsometractionforcesremain.Of
coursethenormaloutputssuchasstressandstrainare
availableaswell.

Results
WiththeisosurfaceviewcutbasedonPHILSMon(default)the
resultsofthissimulationnicelyshowthecrackdevelopingin
time.Byapplyingtranslucency,theinternalpartofthecrackis
alsovisible.

SIMULIA Abaqus - Fracture on gear using XF...

Areyouinterestedtoinvestigateifcrackswillcauseapotetial
failure?

TalktooneofourFEAspecialists
today!

Topics:Modellingcrackpropagation

http://info.simuleon.com/blog/modellingcrackpropagationusingxfem 7/11
08/05/2017 ModellingcrackpropagationusingXFEM

hamidafekirini
24/03/201700:14:01

Manythanksforthisinformation
Ihaveaquestion
whatisthebest,sketchacrackinthemodelor
consideracrackasanotherpartandinmode
assemblywegroupthemodelwithacrack

Replytohamidafekirini

ChristineObbinkHuizer
24/03/201714:47:05

Itisbesttocreateaseparatepart,instanceitintothe
assemblyattheappropriatepositionandrefertoitinthe
crackdefinition.Apartfromeaseofmeshing,thishasthe
advantagethatthecrackwillprobablynotbelocatedat
elementedges,whichisbeneficialfortheXFEMmethod.

ReplytoChristineObbinkHuizer

RicardoDias
07/04/201718:15:53

Thanksforsharingthisinformation!Ihavesome
questionsregardingmyworkandIwouldbegratefullif
youcouldgivemesomefeedback:
I'mdoingacrackpropagationonathreepointbending
testtoextractthevaluesofJIntegralandFactor
IntensiveStressusingXFEM.Mygoaliscomparedthe
results(JIntegralandFIS)betweenXFEMand
ContourIntegralAnalysis.ThesimulationwithContour
Integralanalysiswascarriedoutwithoutanywarning
andIgetmyresults.However,withtheXFEMmethod
IgetwarningsthatIdon'tknowhowtosolve,the
warningsare:

"Displacementincrementforcontactistoobig."
"ThereiszeroFORCEeverywhereinthemodelbased
onthedefaultcriterion.pleasecheckthevalueofthe
averageFORCEduringthecurrentiterationtoverify
thattheFORCEissmallenoughtobetreatedaszero.
ifnot,pleaseusethesolutioncontrolstoresetthe
criterionforzeroFORCE."

http://info.simuleon.com/blog/modellingcrackpropagationusingxfem 8/11
08/05/2017 ModellingcrackpropagationusingXFEM

"Theloadingconditionmightmakemaincrackunder
compressionatnset1.Thecrackpropagation
directioncannotbepredicted.Avalueofzerois
providedforit."

Ihavebeenlookingintheinternetandwasn'tableto
findanyreliableinformationaboutthis.
Bestregards,RicardoDias

ReplytoRicardoDias

ChristineObbinkHuizer
10/04/201708:44:13

Itisimpossibletoreallyknowwhatisgoingonwithout
seeingthemodel,butmyguesswouldbethatthereis
somethingwronginthemodeldefinition.Forexample,if
thewrongsideofasurfaceisdefinedincontact,thismay
resultinstrangebehaviour.Irecommendyoutogradually
buildupthecomplexityofthemodel,sothatitbecomes
clearwhatiscausingtheseissues.

ReplytoChristineObbinkHuizer

RicardoDias
17/04/201712:34:23

Thanksforthisinformation!Iwillfollowyouradvice.
Canyouexplainwhyyouchoose"discontinuous
analysis"ingeneralsolutioncontrolsStep?For
defaultthisoptionisOFF.ShouldIturnON?

Bestregards,RicardoDias

ReplytoRicardoDias

ChristineObbinkHuizer
18/04/201708:57:31

DearRicardo,
AbaqususesNewtonsmethodtosolvenon
linearequations.Thediscontinuousanalysis
optionincreasesI0(thenumberofiterationsafter
whichthecheckismadethattheresidualsare
notincreasingintwoconsecutiveiterations)and
IR(thenumberofequilibriumiterationsafter
http://info.simuleon.com/blog/modellingcrackpropagationusingxfem 9/11
08/05/2017 ModellingcrackpropagationusingXFEM

whichthelogarithmicrateofconvergencecheck
begins).So,withthissettingmoreiterationsare
allowedbeforethesolutionneedstogetcloser
eachiteration.Thiscanbebeneficialfor
extremelynonlinearproblems,suchasthose
involvingXFEM.Thesimulationmaytakelonger,
becauseitwillcontinuetoworkonanincrement
thatistoobiglonger.Apartfromthat,Idonot
expectnegativeeffects,soitmakessensetotry
itforseverelynonlinearproblems.

BensariAhmed
21/04/201722:45:41

Hello,
pleaseineedhelp.
idoasimulationofafatiguecrackpropagation,my
questionis:howdoigetthecuvelenghtofcrackvs
numberofcyclefromodbdatainabaqus.
thanksinadvance.

ReplytoBensariAhmed

FirstName*

LastName*

Email*

Comment*

Subscribetofollowupcommentsforthispost

SubmitYourComment
http://info.simuleon.com/blog/modellingcrackpropagationusingxfem 10/11
08/05/2017 ModellingcrackpropagationusingXFEM

AboutourFEABlog:
Simuleonpostsupdatesonaregularlybasisaboutallthings
relatedtoFiniteElementAnalysislikestructuralanalysisand
CFDanalysis.

SubscribetoourBlog.

YourEmail(tosubscribe)

Subscribe

RecentPosts
HowAbaqustreatsinitialoverclosuresofcontactingsurfaces
SIMULIACommunityNewsMagazineMarch2017
Scheduleatasktoautomaticallyrunall.inpfilesina
selectedfolder
ModelfibrewindingprocesswithAbaqus
ModellingcrackpropagationusingXFEM
ModellingacrackusingAbaqus
TopologyOptimizationofanAbaqusmodelinTosca
WhatsnewSIMULIAAbaqus2017
Abaqus2017WhatsNewJoinourWebinar
HowtouseAbaqusCELtomodelairpressure?

UpcomingEvents
AbaqusTrainingScheduleMay
May10,20179:00:00AMCEST

HandsonWorkshop:IntroductiontoFiniteElementAnalysis
forRubberswithAbaqus
May19,20178:45:00AMCEST

OnlineWebinarTraining:AbaqusGeometryImport&Repair

http://info.simuleon.com/blog/modellingcrackpropagationusingxfem 11/11