IIIEuropeanConferenceonComputationalMechanics Solids,StructuresandCoupledProblemsinEngineering C.A.MotaSoareset.al.(eds.) Lisbon,Portugal,5–8June2006 A MICROMECHANICS-BASED DAMAGE MODEL FOR THE STRENGTH PREDICTION OF COMPOSITE LAMINATES PedroP.Camanho1,JoanA.Mayugo2,PereMaim´ı2 andCarlosG.Da´vila3 1DEMEGI,FaculdadedeEngenharia,UniversidadedoPorto RuaDr. RobertoFrias,4200-465,Porto,Portugal e-mail: [email protected] 2 AMADE,EscolaPolite`cnicaSuperior,UniversitatdeGirona CampusMontilivi,17071Girona,Spain e-mail: {pere.maimi,ja.mayugo}@udg.es 3NASALangleyResearchCenter Hampton,VA,U.S.A. e-mail: [email protected] Keywords: ContinuumDamageMechanics,Fracture,CompositeMaterials. Abstract. A new damage model based on a micromechanical analysis of cracked [±θ◦/90◦] n s laminatessubjectedtomultiaxialloadsisproposed. Themodelpredictstheonsetandaccumu- lation of transverse matrix cracks in uniformly stressed laminates, the effect of matrix cracks on the stiffness of the laminate, as well as the ultimate failure of the laminate. The model also accounts for the effect of the ply thickness on the ply strength. Predictions relating the elastic propertiesofseverallaminatesandmultiaxialloadsarepresented. 1 PedroP.Camanho,JoanA.Mayugo,PereMaim´ıandCarlosG.Da´vila 1 INTRODUCTION Theaerospace industry is committed to improvethe performance of aircraft whilst reducing emissionsandweight. Suchagoalcanbeachievedbytheuseofadvancedpolymer-basedcom- posite materials, that have excellent properties for aerospace applications, such as low density, andfatigueandcorrosionresistance. The design procedure used for advanced composite structures relies on a ’building-block’ approach[1],wherealargenumberofexperimentaltestsareperformedthroughouttheproduct development process. The use of improved analytical or numerical models in the prediction of the mechanical behavior of composite structures can significantly reduce the cost of such structures. Such models should predict the onset of material degradation, the effect of the non- criticaldamagemechanismsonthestiffnessofthelaminate,andultimatestructuralfailure. Strength-based failure criteria are commonly used to predict failure in composite materi- als [2]-[7]. Failure criteria predict the onset of the several damage mechanisms occurring in compositesand,dependingonthelaminate,geometryandloadingconditions,mayalsopredict structuralcollapse. In multidirectional composite laminates, damage accumulates during the loading process. Final failure occurs as a result of damage accumulation and stress re-distribution. The ulti- mate failure load is higher than the initial damage predicted by strength-based failure criteria. Furthermore, stress- or strain-based failure criteria cannot represent size effects that occur in quasi-brittlematerials[8]. Simplified models, such as the ply discount method where some scalar components of the stiffnesstensorarereducedtoapproximatelyzerowhendamageispredicted,areoftenusedby theindustrytopredictultimatelaminatefailure. However,thesemethodscannotrepresentwith satisfactory accuracy the progressive reduction of the stiffness of a laminate as a result of the accumulationofmatrixcracks. Constitutive laws based on Continuum Damage Mechanics (CDM) have been proposed to predict the material response, from the onset of damage up to final collapse [9]-[14]. Although the existing CDM models can predict accurately the evolution of damage, they usually rely on fitting parameters that need to be measured at laminate level, such as the critical values of thermodynamicforces[14]. Three-dimensionalmodelsbasedonCDMusematerialpropertiesmeasuredattheplylevel. Damage mechanisms such as transverse matrix cracks are represented by strain-softening con- stitutive models and bands of localized deformation. The implementation of strain-softening models in finite element models causes convergence difficulties during the iterative solution procedure. Furthermore, strain-softening models require regularization techniques to provide mesh-independentsolutions[15,16]. Alternative methods based on the combination of elastic analysis of cracked plies and finite FractureMechanicsprovidethebasisforanaccuraterepresentationoftheresponseofcompos- ite materials [17]-[28]. Micromechanical models have been developed to predict the initiation andevolutionoftransversematrixcracksundereitherin-planeshearortransversetension. Gen- eralizationsofthesemodelsarerequiredforthemoreusualcaseofmultiaxialloading. Inordertopredictultimatefailure,amicromechanicalmodelneedstobeusedincombination with a fiber failure criterion. Furthermore, a general methodology to predict laminate strength should be able to predict matrix cracking under high values of transverse compression. These damagemechanismsusuallycorrespondtothefinalfailureoflaminatesunderuniformstatesof stress[6]. 2 PedroP.Camanho,JoanA.Mayugo,PereMaim´ıandCarlosG.Da´vila The objective of this work is to define a new damage model based on micromechanical modelsoftransversematrixcrackstopredicttheonsetandevolutionoftransversematrixcracks under multiaxial loading. A new constitutive model is derived based on the thermodynamics of irreversible processes. The model proposed herein represents transverse matrix cracks as distributeddamage. Themodelisabletopredicttheonsetandpropagationofmatrixtransverse cracksundermultiaxialloadingaswellasthefinalfailureofuniformlystressedlaminateswhere aperiodicdistributionoftransversematrixcrackscanbeassumed. 2 MICROMECHANICALMODEL The proposed continuum damage model is based on two major components: a set of stress basedfailurecriteriaandamicromechanicalmodeloftransversematrixcrackinginmultidirec- tionallaminates. The failure criteria define the onset of transverse matrix cracking, i.e. the activation of the damage variables. A micromechanical model of transverse matrix cracks is required to define theevolutionofthedamagevariables. Several micromechanical models of transverse matrix cracks that have been proposed in the literature [19, 28] can be used within the framework developed here. The micromechanical model proposed by Tan and Nuismer [17, 18] accounts for the effects of the adjoining plies on the homogenized elastic properties of a cracked ply. This micromechanical model is the basis ofthedevelopmentspresentedinthiswork. Using the assumption of generalized plane strain, Tan and Nuismer [17, 18] developed a model able to relate the density of transverse matrix cracks in a central 90o ply to the homoge- nized elastic properties of that ply. The model developed by Tan and Nuismer was used for the predictionoftheevolutionoftransversematrixcracksundereitherin-planeshearortransverse tensilestresses. The laminates under investigation are symmetric and balanced with a [±θ/90o] layup con- n s taining a periodic distribution of transverse matrix cracks, as shown in Figure 1. The micro- mechanical analysis of a balanced symmetrical laminate requires the division of the laminate in two sub-laminates: the 90o layers in the middle layer (sublaminate 1), and the outer plies (sublaminate2). x L n 0 L q,2 e 9 ,1 et t a a n nim im albus albus z h(2) h(1) h Figure1: One-quarterofthelaminate. Sublaminate 1 is taken as transversely isotropic. The outer layers, sublaminate 2, may con- tainseverallayerswithdifferentfiberorientationsbutmustbeorthotropicinthelaminateglobal system(x,yandzcoordinates). 3 PedroP.Camanho,JoanA.Mayugo,PereMaim´ıandCarlosG.Da´vila 2.1 CONSTITUTIVETENSOR ThestiffnesstensorofthebalancedandsymmetriclaminateshowninFigure1canbewritten asafunctionofthedensityoftransversematrixcracks(1/L)inthesublaminate1: ¯ ¯ A(L) A(L) 0 11 12 A¯(L) = A¯(L) A¯(L) 0 (1) 21 22 ¯ 0 0 A(L) 66 ¯ wheretheterms A (L),i,j = 1,2,6areobtainedfromtheTanandNuismermodel[18]. ij The undamaged stiffness matrix Q¯ of the laminate is obtained from lamination theory using the undamaged stiffness matrix of sublaminate 1, Q¯(1), and the stiffness matrix of sublaminate 2,Q¯(2): (cid:179) (cid:180) 1 Q(1) = h¯Q¯ −h(2)Q(2) (2) ij h(1) ij ij withh¯ = h(1) +h(2). Assumingthatthedegradationduetothetransversematrixcracksonlyoccurinsublaminate 1,thedamagedstiffnesstensoroflaminate1isgivenas: (cid:179) (cid:180) 1 A¯(1)(L) = h¯A¯ (L)−h(2)Q(2) (3) ij h(1) ij ij whereA¯(1)(L)(i,j = 1,2,6)arethescalarcomponentsofthestiffnesstensorofsublaminate1. ij These components are a function of the distance between transverse matrix cracks (L, defined inFigure1). Having defined A¯(1)(L), it is possible to calculate the effective transverse modulus, Poisson ij ratio,andshearmodulusofthe90◦ ply: (cid:104) (cid:105) 2 A¯(1)(L) 12 E¯(1)(L) = E¯(1)(L) = A¯(1)(L)− (4) 2 x 11 A¯(1)(L) 22 A¯(1)(L) υ¯(1)(L) = υ¯(1)(L) = 12 (5) 21 xy A¯(1)(L) 22 G¯(1)(L) = G¯(1)(L) = A¯(1)(L) (6) 12 xy 66 The quotient υ21 is not a function of damage. This observation is in agreement with several E2 othermodels,suchastheonesproposedbyLawsetal. [23,24],andNguyen[25]. The plane stress compliance tensor of the damaged sublaminate 1, H(1), only contains two componentsthatdependonthedensityoftransversematrixcracks: H(1)(L)andH(1)(L). The 22 66 tensorH(1),isestablishedasafunctionofthedensityoftransversematrixcracks, L,as: 1 −υ21 0 E1 E2 H(1)= −υ12 H(1)(L) 0 (7) E1 22 0 0 H(1)(L) 66 4 PedroP.Camanho,JoanA.Mayugo,PereMaim´ıandCarlosG.Da´vila with: 1 A¯(1)(L) H(1)(L) = = 22 (cid:104) (cid:105) (8) 22 E¯2(1)(L) A¯(1)(L)A¯(1)(L)− A¯(1)(L) 2 11 22 12 1 1 H(1)(L) = = (9) 66 G¯(1)(L) A¯(1)(L) 12 66 The evolution of the homogenized elastic properties of the cracked ply (sublaminate 1) as a function of β = 1/(2L) is shown in Figure 2 for a [±25◦/90◦] laminate with the material 3 s propertiesshowninTable1. E (GPa) E (GPa) G (GPa) υ 1 2 12 12 163.4 11.9 6.2 0.30 Table1: Elasticpropertiesoftypicalcarbon/epoxycomposite. Figure2: Effectiveelasticpropertiesofthe90◦plyasafunctionofthedensityofmatrixcracks. Figure2showsthattheYoungmodulusofsublaminate1,E ,isnotaffectedbythepresence 1 of transverse matrix cracks. Furthermore, the curves E (β)/E and υ (β)/υ are coincident 2 2 21 21 (Figure2). 2.2 ONSETOFMATRIXTRANSVERSECRACKS Theonsetoftransversematrixcrackinginaplyunderthecombinedeffectofin-planeshear stresses and transverse tensile stresses needs to be predicted using an appropriate failure crite- rion. Thefailurecriterionshouldbeestablishedintermsoftheactualstrengthsofaplywhenit isembeddedinamultidirectionallaminate(in-situstrengths). 5 PedroP.Camanho,JoanA.Mayugo,PereMaim´ıandCarlosG.Da´vila 2.2.1 In-situstrengths A methodology to predict the onset of matrix transverse cracking must be able to predict the in-situ strengths of a ply. Both the transverse tensile and in-plane shear strengths of a ply embedded in a multidirectional laminate are higher than the corresponding values measured in unidirectionallaminates[29]. The thick models described in [30] are used. The tensile transverse in-situ strengths of sublaminate(1)is[30]: (cid:115) 8G Forathinembeddedply: Y = Ic (10) T πtΛo 22 √ Forathickply: Y = 1.12 2Yud (11) T T where Yud is the tensile transverse strength measured in a unidirectional test specimen, t is the T thicknessofsublaminate1,G isthemodeIintralaminarfracturetoughnessandΛ◦ isdefined Ic 22 as: (cid:181) (cid:182) 1 υ2 Λ◦ = 2 − 21 (12) 22 E E 2 1 Thein-situshearstrengthsareobtainedas[29]: (cid:115) (1+χφG2 )1/2 −1 S = 12 (13) L 3χG 12 whereχistheshearresponsefactordefinedin[29],andtheparameterφiscalculatedaccording totheconfigurationoftheply: (cid:161) (cid:162) 12 Sud 2 (cid:161) (cid:162) Forathickply: φ = L +18χ Sud 4 G L 12 48G Forathinply: φ = IIc (14) πt where Sud is the shear strength measured using an unidirectional test specimen and G is the L IIc modeIIfracturetoughness. 2.2.2 Failurecriterionforthepredictionoftransversecrackingundermultiaxialloading In general, a ply represented by sublaminate 1 in Figure 1 is subjected to transverse tensile stresses and in-plane shear stresses. Under pure in-plane shear or pure transverse tension, the onset of transverse matrix cracking is predicted by comparing the components of the stress tensorwiththerespectivein-situstrengths(definedintheprevioussection). Under multiaxial loading, it is necessary to use a failure criterion to predict the onset of matrix cracking. The LaRC04 [4, 5] failure criteria are a function of the components of the stresstensorandin-situstrengths. Fortransversetension,thecriterionusedis: σ σ σ (1−g) 22 +g( 22)2 +( 12)2 −1 ≤ 0withσ ≥ 0 (15) 22 Y Y S T T L 6 PedroP.Camanho,JoanA.Mayugo,PereMaim´ıandCarlosG.Da´vila where g = GIc. G is the mode II component of the fracture toughness associated with matrix GIIc IIc transversecracking. Under moderate values of transverse compression, the cracks propagate along the laminate thicknessdirection,andtheLaRC04failurecriterionis: 1 (cid:173) (cid:174) |σ |+ηLσ −1 ≤ 0withσ < 0 (16) 12 22 22 S L whereηL isthecoefficientoflongitudinalinfluencedefinedin[4,5]. 2.3 Evolutionofmatrixtransversecracksundermultiaxialloading 2.3.1 Transversetension Tan and Nuismer [18] obtained a closed-form expression that defines the evolution of trans- versematrixcracksunderuniaxialstressstates(eithertransversetensionorin-planeshear). To predict failure under multiaxial loading, i.e. when the lamina is simultaneously under tensile and in-plane shear strains, ε and γ respectively, it is necessary to derive a relation xx xy between the density of transverse matrix cracks and the applied multiaxial strain state. It is assumedthattherelationbetweenthetensileandshearstrainsisconstantthroughouttheloading history: γ xy κ = withε > 0 (17) xx ε xx whereκisthemultiaxialstrainratio. Consideraplywithcrackdensity L,asshowninFigure 3a). 2L L L a) Crack density 2L b) Crack density L Figure3: Progressionoftransversematrixcracks. Thestrainenergyinalaminatecelloflength2Lsubjectedtotransversetensionandin-plane shear just prior to fracture, U , can be established as a function of the strain tensor and of the 2L crackdensityas: (cid:163) (cid:164) U = 2hLb E (L)ε2 +A (L)γ2 (18) 2L x xx 66 xy where E (L) is the axial stiffness of the laminate with a crack density of 1/(2L) and b is the x specimenwidth;A (L)isthelaminateeffectiveshearstiffnesscorrespondingtoacrackdensity 66 of 1/(2L), 2h and 2L are the laminate thickness and the distance between two consecutive transversematrixcracks,respectively. Afterthepropagationoftransversematrixcracks,Figure3b),thestrainenergyintheoriginal unitcelloflength 2Lis: (cid:183) (cid:181) (cid:182) (cid:181) (cid:182) (cid:184) L L U = 2hLb E ε2 +A γ2 (19) L x 2 xx 66 2 xy 7 PedroP.Camanho,JoanA.Mayugo,PereMaim´ıandCarlosG.Da´vila where E (L/2) and A (L/2) are respectively the axial stiffness and the laminate effective x 66 shearstiffnesscorrespondingtoacrackdensitydefinedby L. The energy required to generate a new matrix crack in a ply equals the loss of strain energy ofthelaminate[18]. Therefore,thedifferencebetweenequation(18)andequation(19)isequal totheenergyreleasedbythesublaminate1: ∆U = U −U = 2L (cid:189)(cid:183)L (cid:181) (cid:182)(cid:184) (cid:183) (cid:181) (cid:182)(cid:184) (cid:190) L L = 2hLb E (L)−E ε2 + A (L)−A γ2 (20) x x 2 xx 66 66 2 xy ∆U = 2h(1)bG = 2h(1)bG +2h(1)bG (21) c I II where G is the mixed-mode fracture toughness of sublaminate 1 under tensile (mode I) and c shear(modeII)loading. From(20)and(21): (cid:183) (cid:181) (cid:182)(cid:184) (cid:183) (cid:181) (cid:182)(cid:184) L L h(1)G E (L)−E ε2 + A (L)−A γ2 = c (22) x x 2 xx 66 66 2 xy hL Using the definition of κ given in (17), equation (22) can be re-written as a function of the strains: (cid:189)(cid:181) (cid:181) (cid:182)(cid:182) (cid:183) (cid:181) (cid:182)(cid:184)(cid:190) L L h(1)G E (L)−E +κ2 A (L)−A ε2 = c (23) x x 2 66 66 2 xx hL or: (cid:189) (cid:183) (cid:181) (cid:182)(cid:184) (cid:183) (cid:181) (cid:182)(cid:184)(cid:190) 1 L L h(1)G E (L)−E + A (L)−A γ2 = c (24) κ2 x x 2 66 66 2 xy hL Therelationsbetweenthenormalandshearstrainsandthecrackdensityareobtainedas: (cid:115) h(1)G 1 ε = c(cid:163) (cid:161) (cid:162)(cid:164) (cid:163) (cid:161) (cid:162)(cid:164) (25) xx hL E (L)−E L +κ2 A (L)−A L x x 2 66 66 2 (cid:115) h(1)G κ2 γ = c(cid:163) (cid:161) (cid:162)(cid:164) (cid:163) (cid:161) (cid:162)(cid:164) = κε (26) xy hL E (L)−E L +κ2 A (L)−A L xx x x 2 66 66 2 Equations(25)and(26),areestablishedasafunctionofthemixed-modefracturetoughness G thatneedstobedefined. c The mixed-mode fracture toughness is defined in terms of the mode I and mode II compo- nentsas: G = G +G (27) c I II From(27)and(22): (cid:183) (cid:181) (cid:182)(cid:184) h(1)G L I = E (L)−E ε2 (28) hL x x 2 xx (cid:183) (cid:181) (cid:182)(cid:184) h(1)G L II = A (L)−A γ2 (29) hL 66 66 2 xy 8 PedroP.Camanho,JoanA.Mayugo,PereMaim´ıandCarlosG.Da´vila Dividingequations(28)and(29)by(23)and(24),respectively: (cid:163) (cid:161) (cid:162)(cid:164) G E (L)−E L A = I = (cid:163) (cid:161) x(cid:162)(cid:164) (cid:163)x 2 (cid:161) (cid:162)(cid:164) (30) I G E (L)−E L +κ2 A (L)−A L c x x 2 66 66 2 (cid:163) (cid:161) (cid:162)(cid:164) G κ2 A (L)−A L A = II = (cid:163) (cid:161) 6(cid:162)6(cid:164) (cid:163) 66 2 (cid:161) (cid:162)(cid:164) (31) II G E (L)−E L +κ2 A (L)−A L c x x 2 66 66 2 A andA aretheratiosbetweenthemodeIandmodeIIcomponentsoftheenergyrelease I II rateandthemixed-modefracturetoughness. The criterion proposed by Hahn [31] for the prediction of transverse matrix cracking under transversetensileandin-planeshearloadsisused: (cid:114) G G G (1−g) IIc +g I + II = 1 (32) G G G Ic Ic IIc Substitutingequations(30)and(31)into(32)gives: (cid:114) A G A G (1−A )G I c I c I c (1−g) +g + = 1 (33) G G G Ic Ic IIc Thepositiverealsolutionof(33)is: (cid:195) (cid:115) (cid:33) A (G −G )2 4 G G G = G + I Ic IIc 1− 1+ Ic IIc (34) c IIc 2 G A (G −G )2 Ic I Ic IIc where A depends on the density of transverse matrix cracks, β, and on the multiaxial strain I ratio,κ(equation(30)). Figure 4 shows the relation between the crack density β and ε for different multiaxial xx strainratios. A[±25◦/90◦] laminatewiththeelasticpropertiesshowninTable 1isusedinthe 3 s predictions. Figure4: Relationbetweenappliedstrain(ε andκ)anddensityofmatrixcracks. xx 9 PedroP.Camanho,JoanA.Mayugo,PereMaim´ıandCarlosG.Da´vila The effects of multiaxial strain states on the crack density are clearly shown in Figure 4. It can be observed in Figure 4 that the density of transverse matrix cracks increases with the multiaxialstrainratioforafixedvalueof ε . xx 2.3.2 Transversecompression Transverse matrix cracks created by a combination of transverse compression and in-plane shearcloseunder theeffectofthe compressivetransversestress. Whena crackcloses, itsfaces can transmit normal tractions but shear tractions may cause slippage between the crack faces. Therefore, it can be assumed that transverse matrix cracks only affect the shear stiffness of a laminate. Following the procedure described in the previous section, the energy released by the sub- laminate1is: (cid:183) (cid:181) (cid:182)(cid:184) L ∆U = 2hLbγ2 A (L)−A = 2h(1)bG (35) xy 66 66 2 IIc Therelationbetweentheshearstrainandthecrackdensityisobtainedas: (cid:115) h(1)G γ = (cid:163) IIc (cid:161) (cid:162)(cid:164) (36) xy hL A (L)−A L 66 66 2 3 MICROMECHANICS-BASEDDAMAGEMODEL 3.1 Constitutivemodel A damage model able to represent the onset and accumulation of a periodic distribution of transverse matrix cracks should yield a compliance tensor similar to the one obtained from the micromechanical model, equation (7). An appropriate damage model can be developed by definingtheGibbsfreeenergyperunitvolume, Ψ ,as: G (cid:183) (cid:181) (cid:182) (cid:184) 1 σ2 σ2 σ2 υ υ Ψ = 11 + 22 + 12 − 21 + 12 σ σ + G 11 22 2 E (1−d )E (1−d )G E E 1 2 2 6 12 2 1 +(α σ +α σ )∆T +(β σ +β σ )∆M (37) 11 11 22 22 11 11 22 22 whereE ,E ,υ ,υ andG arethein-planeelasticorthotropicpropertiesofaunidirectional 1 2 12 21 12 lamina. The subscript 1 denotes the longitudinal (fiber) direction and 2 denotes the transverse (matrix) direction. d and d are damage variables associated with transverse matrix crack- 2 6 ing. α and α are the coefficients of thermal expansion in the longitudinal and transverse 11 22 directions, respectively. β and β are the coefficients of hygroscopic expansion in the lon- 11 22 gitudinal and transverse directions, respectively. ∆T and ∆M are respectively the changes in temperatureandmoisturecontentfromthestress-freestate. Theproposedmodeldistinguishesbetweenactive(d )andpassivedamage(d )variables, 2+ 2− correspondingtotheopeningorclosureoftransversematrixcracksunderloadreversalrespec- tively. Toaccountfortheactivedamageundereitheropeningorclosure,thefollowingequation isproposed: (cid:104)σ (cid:105) (cid:104)−σ (cid:105) 22 22 d = d +d (38) 2 2+ 2− |σ | |σ | 22 22 10