• Aucun résultat trouvé

Conserving and decaying energy for finite-strain three-dimensional solids with rotational degrees of freedom in nonlinear dynamics

N/A
N/A
Protected

Academic year: 2021

Partager "Conserving and decaying energy for finite-strain three-dimensional solids with rotational degrees of freedom in nonlinear dynamics"

Copied!
11
0
0

Texte intégral

(1)

HAL Id: hal-01996646 https://hal.utc.fr/hal-01996646

Submitted on 5 Feb 2019

HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

Conserving and decaying energy for finite-strain three-dimensional solids with rotational degrees of

freedom in nonlinear dynamics

Abir Boujelben, Adnan Ibrahimbegovic

To cite this version:

Abir Boujelben, Adnan Ibrahimbegovic. Conserving and decaying energy for finite-strain three-

dimensional solids with rotational degrees of freedom in nonlinear dynamics. Comptes Rendus Mé-

canique, Elsevier, 2018, 346 (7), pp.571-580. �10.1016/j.crme.2018.04.006�. �hal-01996646�

(2)

C. R. Mecanique 346 (2018) 571–580

Contents lists available atScienceDirect

Comptes Rendus Mecanique

www.sciencedirect.com

Model reduction, data-based and advanced discretization in computational mechanics

Conserving and decaying energy for finite-strain

three-dimensional solids with rotational degrees of freedom in nonlinear dynamics

Conservation et dissipation d’énergie dans les solides tridimensionnels en grandes transformations avec des degrés de liberté rotationnels en

dynamique non linéaire

Abir Boujelbena, Adnan Ibrahimbegovicb

aSorbonneUniversités,UniversitédetechnologiedeCompiègne,LaboratoireRobervaldemécanique,CentrederecherchesdeRoyallieu, CS 60319,60200Compiègnecedex,France

bSorbonneUniversités,UniversitédetechnologiedeCompiègne,LaboratoireRobervaldemécanique,ChairofComputationalMechanics, CentrederecherchesdeRoyallieu,CS 60319,60200Compiègnecedex,France

a r t i c l e i n f o a b s t r a c t

Articlehistory:

Received4May2017 Accepted23January2018 Availableonline9May2018

Keywords:

3Dsolids Nonlineardynamics

Largedisplacementsandrotations Conserving/decayingalgorithms

Mots-clés : Solide3D

Dynamiquenonlinéaire Grandestransformations

Algorithmedeconservation/dissipation

Inthiswork,wedesignatime-steppingscheme,whichcanensureeitherconservationof energyordissipationofenergyofhigh(unresolved)modesfornonlineardynamicanalysis.

The latter is neededto improve the performance instress computation and long-term numericalstability.Finiteelementimplementationdetailsaregivenforfinite-strainthree- dimensional solid model with independent rotational degrees of freedom. The addition ofarotationfieldrequires aparticular choiceoflarge strain measures,allowingoneto separatelargerotationandlarge displacement.Severalnumericalsimulationsillustratea verysatisfyingperformanceoftheproposedtime-steppingscheme.

©2018PublishedbyElsevierMassonSASonbehalfofAcadémiedessciences.Thisisan openaccessarticleundertheCCBY-NC-NDlicense (http://creativecommons.org/licenses/by-nc-nd/4.0/).

r é s um é

Danscetarticle,nousdévelopponsunschémaimplicited’intégrationtemporelle capable d’assurer, soitla conservationde l’énergie, soit ladissipation de l’énergied’un système dynamique non linéaire. Ce dernier est particulièrement intéressant pour améliorer la stabilité numérique pour le calcul sur un grand intervalle du temps. L’implémentation de la méthode des éléments finis est présentée en détail pour un modèle de solide tridimensionnelen grandes transformations qui fait intervenir des degrésde libertéde rotationindépendants.Lapriseencomptedelarotationimpliqueunchoixjudicieuxde la mesuredesgrandes déformations pour séparerlesgrands déplacements desgrandes

E-mailaddresses:abir.boujelben@utc.fr(A. Boujelben),adnan.ibrahimbegovic@utc.fr(A. Ibrahimbegovic).

https://doi.org/10.1016/j.crme.2018.04.006

1631-0721/©2018PublishedbyElsevierMassonSASonbehalfofAcadémiedessciences.ThisisanopenaccessarticleundertheCCBY-NC-NDlicense (http://creativecommons.org/licenses/by-nc-nd/4.0/).

(3)

rotations.Laperformanceduschémaproposéestillustréeàtraversdenombreuxexemples desimulations.

©2018PublishedbyElsevierMassonSASonbehalfofAcadémiedessciences.Thisisan openaccessarticleundertheCCBY-NC-NDlicense (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Versionfrançaiseabrégée

L’analysedynamique nonlinéaire s’effectue généralementpardesschémas d’intégrationtemporelle,enprivilégiantles schémasimplicitestelsqueceluideNewmark,carilsn’imposentpasderestrictionssévèrespourlechoixdupasdetemps.

Cependant, dans lecas des systèmes raides laréponse est dominée par unegrande différence des modesà basse et à haute fréquences,ces schémasnesontpassuffisants pourgarantir lastabilitédu calculnumérique. Pour surmonterce problème,nousdévelopponsunschémad’intégrationquigarantitlaconservationdel’énergieàchaquepasdetempsetqui permet,encasdebesoin,deforcerladissipationnumériquedelacontributiondeshautesfréquences.

Dans cetravail,nousnousintéressonsplusparticulièrementàl’implémentationde laméthodedesélémentsfinispour leschémaproposédemodèledesolidetridimensionnelengrandestransformationsquifaitintervenirdesdegrésdeliberté rotationnels indépendants.Lapriseencomptedesrotationsimpliqueunchoixspécifiquedelamesuredeladéformation : la déformation de Biot. Lesapproximations discrètes des champsde déplacement et de rotation sont obtenues par une interpolationlinéairestandard.Afind’homogénéiserl’approximationdesmesuresdedéformation,onfaitappelàlaméthode desmodesincompatibles,permettantd’augmenterl’ordredel’approximationduchampdedéplacementparl’additiond’un champdedéplacementamélioré(quadratique).

Leproblèmeestprésentésousuneformulationvariationnellerégulariséeengrandesdéformations,avecuneformequa- dratiquesimpledutermed’inertievoirl’Éq. (3).Cedernierestexpriméentermesdedéplacements,tandisquelechamp derotationcontribueuniquementauxmesuresdeladéformation.Ceciimpliquelaprésenced’unblocdetermesnulsasso- ciésauxdegrésde libertéderotationdanslamatrice masse,provoquant desmodesenfréquencesinfinies.Pourcontrôler l’effet indésirable de ces modes,nous proposonsd’abord d’introduireuneforme régulariséedestermes d’inertie associés aux degrés deliberté de rotation,afind’éviterla présencedesfréquences infinies.Ensuite,nousexploitons leschémade conservationd’énergieafind’assurerlastabilitéducalcul,mêmeencasdedominationde modesàhautesfréquences,qui apparaissentcommelerésultatdelarégularisation.

Le schémaproposéest basésurleschémade point milieu,enyapportant quelques modificationsafinde satisfairele caractèreconservatif.D’unepart,cesmodificationsconcernentuneexpressionalternativedelarelationdecomportementen tn+1

2 sousformealgorithmique,detellesortequeletermede déformationdanslaformulationvariationnellecorresponde à l’incrément de l’énergie potentielle voirles Éqs. (12) et (13). D’autre part,une approximation spécifique du vecteur de vitesselinéairepermetd’égaliser letermed’inertieetl’incrémentde l’énergiecinétiquevoirlesÉqs. (10) et(11).En supposant queleseffortsextérieurssontconservatifs,laconservationd’énergie estbienassurée. Unautreenjeus’impose lors de la construction du schémad’intégration : c’est l’approximation du champ de rotation,qui est extrêmement non linéaire.Dansunedémarchedesimplification,onretientuneapproximationlinéaire.

Une autre alternative proposéepour contrôler les problèmes d’instabilité causés par des systèmes d’équations raides consiste à introduireunedissipationnumérique deshautesfréquences àchaque pasdetemps.L’originalité de ceschéma réside dansl’additiondestermesdissipatifsvoirlesÉqs. (18) et(20) – sanstoucheràl’algorithmedéfinidanslecasde la conservationd’énergie,de manièreà ceque lesdeux schémaspuissentpartager lemêmeespace de stockagepour les variablesinternes.Autrementdit,lesmodificationsconcernentseulementl’actualisationduvecteurdevitesseetlarelation algorithmiqueducomportement.

Enconclusion,laconservationouladissipationdel’énergietotaleestunepropriétéimportantedesschémasd’intégration temporelle,particulièrementpourdescalculssurungrandintervalledetemps,l’onsouhaiteatténuerl’effetindésirable deshautesfréquencessurledéclenchementd’instabilités,sansoublierlacontributiondelarégularisationdemassedansle casdusolide3Davecdesdegrésdelibertéderotation.D’ailleurs,cemodèleserévèlefavorablepourl’analysedestructures combinées d’éléments solides et d’éléments structures, d’autant plus que plusieurs travaux antérieurs ont été réalisés à proposdelaconservationd’énergied’élémentspoutres.

1. Introduction

Thedynamic analysisofgeometricallynonlinearproblemsdominatedbyhigh-frequencymodesisextremelydifficultto solvebyusingstandardimplicittime-steppingschemes,suchasNewmark’smethod.Toresolvethisimportantissue,avery active researchhavebeenelaboratedonconservingor/anddecayingenergytime-steppingschemesfornonlineardynamics ofrods[1–3] andshell-likestructures[4,5] undergoinglargeoverallmotion.Themainmotivationoftheseapproachesisto ensurethesatisfyingcomputationalperformanceinthecaseofstiffproblems,mainlyforlong-termresponse.

In this paper, we are interested in the regularized variational formulation for the nonlinear dynamics of three- dimensional solid with independent rotational degrees of freedom. The formulation is set in a fixed frame, featuring a

(4)

A. Boujelben, A. Ibrahimbegovic / C. R. Mecanique 346 (2018) 571–580 573

simplequadraticformofthekineticenergy.Thelatterisexpressedonlyintermsofthedisplacementfield,sincetherota- tionfieldisinvolvedjusttocomputethelargestrainmeasures.Thisleadstothepresenceofzerotermsinthemassmatrix associated withtherotational degreesoffreedom requiring infinitefrequencies anddisrupting thecomputation stability.

Therefore, we seek to constructan energy conserving time-steppingscheme to overcome the deficiency of (unresolved) high-frequency participation.Thisscheme isbased onthe mid-pointapproximation.The pertinentmodifications that are neededto enforce theenergy conservationare algorithmic constitutive equationsand appropriate choice ofrotation and velocityupdates.

In seeking to improve the robustness of the scheme stability performance, we propose to redesign the energy con- serving scheme by introducing desirableproperties ofcontrollable dissipation in highermodes. The main idea is to add regularizedtermstoalgorithmicstressresultantsandvelocityupdates,andtoassurean artificialenergydissipation ofthe high-frequencymodes.

Theoutlineofthepaperisasfollows.Inthenextsection,wepresenttheregularizedvariationalformulationindynamics oftheproposedenhanced3Dsolidundergoinglargeoverallmotion.InSection 3,weexamineindetailtheformulationand theimplementationofatime-steppingenergyconserving/decayingschemesfortheproposedmodel.A setofillustrativenu- mericalsimulationsispresentedinSection4,showinganexcellentperformanceoftheproposedscheme.Someconcluding remarksaregiveninSection5.

2. Variationalformulationsandfiniteelementimplementation

2.1. Hamiltonprincipleindynamics

We consider a three-dimensional solid model with rotational degreesof freedom undergoing overall large motionin geometricallynonlinear dynamic problems.TheBiot strain His chosen todefine strain measures:H=R(I+D)I.The latterintroduces explicitlythe rotationtensorR definedthroughthe polardecompositionofdeformation gradient [6].In orderto improvethe performanceof thediscrete approximation,the enhanced displacementgradient isdefinedthrough thedisplacementgradientD,givenasa sumofthestandarddisplacementgradientuandofan enhanceddisplacement gradientdprovidedbythemethodofincompatiblemodes.

TheequationsofmotionareobtainedbyappealingtotheclassicalHamiltonprinciple[7]:

(i) T(u)=

V

1

2u˙·ρu˙dV (ii) (u,R,d)=

V

{1

2symm[R(I+D)I] ·Csymm[R(I+D)I] +1

2skew[R(I+D)I] ·γskew[R(I+D)I] −P·d}dV−

V

u·fdV

(1)

where P is the first Piola–Kirchoffstress tensor, conjugated tothe enhanced displacement gradient d,whereas γ isthe regularization parameter. We emphasize that the kinetic energy is expressed onlyin terms ofdisplacement field, which impliesthepresenceofzeromassesassociatedwiththerotationdegreesoffreedom.Thisleadstoinfinitefrequenciesand the extremestiffproblemwithall the difficultiesthat itwill impose.These difficultiesare morepronounced inthe case ofa 3D solidelement capable ofrepresentingmorevibration modescomparedto structureelements. Forthat reason, in additionofaenergyconserving time-steppingscheme,itisparticularlynecessaryto addthemassmatrixcontributionof theangularaccelerationinordertoreduce theinfluenceofinfinitefrequencies.Thiscontributionispresentedintheform ofaregularizedexpressionforkineticenergy,equaltoasimilarquadraticformbutmultipliedbyacoefficient αthatvaries from0 to 1. Therefore,one can attenuatethe high-frequency contentinthe structuralmotionsimply bychoosing larger valuesforthemassregularizationparameterα.

Forcomputationalefficiency,wefurtherswitchtothecompactformmatrixnotation,detailedinAppendixA.Wedenote by andL thestressresultants,energyconjugatestothesymmetricparte andtheskew-symmetricpart ωofthestrain measureswritteninamatrixform,respectively,

(u,R,d):=C e(u,R,d);L(u,R,d):=γ ω(u,R,d) (2)

Then,thecorrespondingvariationalequationscanbewrittenas

(5)

δu δw

·r(u,R,d,p):=

V

δu·ρu¨dV+

V

{δe·Ce(u,R,d)+δω·γω(u,R,d)} −

V

δu.fdV=0

δd·h(u,R,d,p):=

V

d·(R)Ce(u,R,d)+δd·(Rω(u,R,d)δdp}dV=0 δp·g(u,R,d,p):=

V

pd}dV=0

(3)

2.2. Discreteapproximation

The discreteapproximationsforthestandarddisplacementfieldandtherotationfield areconstructedaccordingtothe standardisoparametricinterpolationforan8-node3Dsolidelementwith

u(x,t)= 8 I=1

NI(x)uI(t)Nue(t);w(x,t)= 8

I=1

NI(x)wI(t)Nwe(t) (4)

where NI arethestandardshapefunctions,whileuI andwI arethecorrespondingnodalvalues.

Theenhanceddisplacementgradientareapproximatedbyusingderivativesoftheincompatibledisplacement α

d(x,t)=

Nim

J=1

GˆJ(x)αJ(t)≡ ˆG(x)αe(t),Gˆ=

Gˆ1,Gˆ2,Gˆ3 , GˆI=

⎢⎢

⎢⎢

⎢⎣

MI

x1I3

MI

x2I3

MI

x3I3

⎥⎥

⎥⎥

⎥⎦

(5)

where M1=(1ξ2), M2=(1η2) and M3=(1ζ2) are chosen as quadratic polynomials. In order to ensure the convergenceofincompatible modemethodinthespiritofpatchtest [8],weimpose that

VGˆdV=0,eliminatingP from thevariationalequations.

3. Energyconserving/decayingschemes 3.1. Energyconservingscheme

Inthissection,weseektodesignatime-steppingschemethatensuresenergyconservation.Atatypicaltimetn,theval- uesofunandRn areknown.Thecorrespondingvaluesattimetn+1arecomputedineachstepindependentlyforsingle-step schemes:forthedisplacementvector,wehavesampleadditiveupdates

un(i++11)=u(ni+)1+u(ni+)1 (6)

where u(ni+)1 is the incremental displacement at each iteration (i). The rotation update is somewhat more involved in that we havetochoose betweenvariouspossibilities ofparameters forrotationrepresentation[9].Ingeometrically exact beam theory,SimoandTarnow proposed to usethe materialrotation vector withCayleytransform mappinginorder to assure rotationupdates [10].However, we optedfora quaternion representationthat reducestherotation tensorto four parameters (see, e.g.,[11]),providingamore efficientcomputational procedureby replacingthe matrixmultiplicationby thequaternionalgebra.Then,therotationalupdatescanbewrittenbyusingtheaxialvectorofincrementalrotationwn(i+)1 as

Rn(i++11)=(2q((in++11))201)I3+2q(0i(+n+1)1)[q(ni++11)×I3] +2qn(i++11)q(ni++11)

q(0i(+n+1)1)=q(w0i)(n+1)q(0i()n+1)q(wi)(n+1).q(ni+)1;q(ni++11)=q(w0i)(n+1)qn(i+)1+q(0i()n+1).q(wi)(n+1)+q(wi)(n+1)×q(ni+)1 {q(w0i)(n+1),q(wi)(n+1)} = {cos(w(ni+)1

2 ),sin(w

(i) n+1

2 )

w(ni+)1

w(ni+)1};w(ni+)1=(w(ni+)1.w(ni+)1)12

(7)

Weassumethatthedisplacementandrotationfieldareexpressedattn+1

2 byalinearapproximation.Thisplaysacrucial roleinthedesignoftheproposedschemenamelyfortherotationfield,whichishighlynonlinear.

un+1

2 =1

2(un+1+un);Rn+1

2 =1

2(Rn+1+Rn) (8)

(6)

A. Boujelben, A. Ibrahimbegovic / C. R. Mecanique 346 (2018) 571–580 575

Thetimediscretizationoftheweakformoftheequationsofmotionisestablished,usingthemid-pointruleapproxima- tion.

(i)

ϑ

δu·ρu¨n+1 2 dϑ+

ϑ

e(un+1 2,Rn+1

2,dn+1 2)·n+1

2 +δω(un+1 2,Rn+1

2,dn+1 2)·Ln+1

2}

ϑ

δu.fn+1 2 dϑ=0 (ii)

ϑ

{δd·(Rn+1 2)n+1

2 +δd·(Rn+1 2)Ln+1

2}dϑ=0

(9)

wherethelinearaccelerationu¨n+1

2 iscomputedusingthemid-pointapproximationas

¨ un+1

2=u˙n+1− ˙un

t ; ˙un+1= −˙un+ 2

tun+1 (10)

Byusingtheresultin(10),thefirsttermin(9)(1)canbewrittenasthecorrespondingincrementofkineticenergy

T=

V

δu·ρu¨n+1

2 dV=

V

t

2 (u˙n+1+ ˙un)·ρ 1

t(u˙n+1− ˙un)dV=Tn+1Tn (11) Toensurethattherestof(9)(1)givestheincrementofinternalenergy,thestressresultantsattn+1

2 mustbecomputedwith aparticularchoiceofalgorithmicconstitutiveequations

alg

n+12=1

2C (en+1+en); Lalgn+1

2 =1

2γ (ωn+1+ωn) (12)

whichleadsto

int=

V

e(un+1 2,Rn+1

2,dn+1 2)·alg

n+12 +δω(un+1 2,Rn+1

2,dn+1 2)·Lalg

n+12}dV

=

V

{1

2(en+1en)·C(en+1+en)+1

2(ωn+1ωn)·γ(ωn+1+ωn)}dV

=int|n+1int|n

(13)

Withoutlossofgeneralityforourpurpose,weassumethattheexternalloadingisobtainedfromapotential,assuringthat ext=ext|n+1ext|n.Thus,withtheresultsin(11) and(13),wegettheform(9)(1)

δu δw

·r(un+1,Rn+1,dn+1)=0 n+1 ext|n+1+int|n+1

+Tn+1=n+Tn (14)

Byusingthevariationofstrainmeasures,thevariationalequationsin(9) arerewrittenas

(i)

V

δu·ρu¨n+1

2 dV+

V

{yu)+ [Y(un+1

2)+Dn+1

2w} · {(Rn+1 2)alg

n+12 +(Rn+1 2)·Lnalg+1

2

}

V

δu.fn+1 2 dV=0 (ii)

V

d·(Rn+1 2)alg

n+12 +δd·(Rn+1 2)Lalg

n+12}dV=0

(15)

whichassuresenergyconservation.

3.2. Energydecayingscheme

Anotheralgorithmicmodificationisproposedheretocontroltheeffectofthehigh-frequencycontributionbydeveloping a time-stepping energy-decaying scheme. In particular, the latter serves as a filter to remove the high-frequency noise, whichcannotberesolvedforaparticularfiniteelementmeshoneach timestep.First,adissipationtermisaddedtothe kineticenergythroughthedisplacementfieldapproximation

Références

Documents relatifs

KEY WORDS : dynamic fracture mechanics; numerical stability; energy balance; extended finite element method; dynamic stress intensity

Armero, An objective finite element approximation of the kinematics of geometrically exact rods and its use in the formulation of an energy-momentum conserving scheme in

In this paper, we have discussed the regularized variational formulation for 3D solid element with the incompatible mode method to analyze geometrically nonlinear problems in

We propose a space semi-discrete and a fully discrete finite element scheme for the modified phase field crystal equation (MPFC).. The space discretiza- tion is based on a splitting

There exists several possibilities to synthesise symmetric functions and partially symmetric functions. For instance, Arnold and Harrisson [AH63] has proposed a representation and

Table 4.3 compares the proposed accelerated alternating minimization with a standard implementa- tion of Beck and Teboulle’s algorithm [9] (FISTA) applied to the dual problem (4.5)..

1D-ballistic diode test case: Newton iterations for the primal and dual schemes, number of time steps and adaptative time steps for the Chen model.. Two-dimensional