This paper compares stiffness degradation models of cross-ply glass fiber/epoxy laminates based on four of the most commonly used approaches to micromechanical modelling: shear-lag, variational, McCartney, and synergistic damage mechanics (SDM). All of these include the process of defining 0/90s laminate unit cell, from which governing differential equations and corresponding boundary conditions are stated. Afterwards, these boundary value problems (BVP) are solved in order to obtain a stress function which couples the initial and perturbation stresses, the latter being in function of crack density, thus related to material stiffness reduction. When compared against experimental results, shear-lag model presented accurate results however, additional differentiation and integration steps were required in order to obtain the final stress field. Hashin’s variational method predicts correctly the boundary conditions at crack surfaces and gives out the complete stress field. McCartney’s approach shows further improvement over the previous two models, taking into account thermal strains and stresses. Finally, SDM, which is designed for numerical experimentation, implying a more economical alternative in comparison to traditional physical experimentation, also presented very good agreement with experimental results and can be extended to arbitrary laminate stackings, going beyond the classical cross-ply.
1. Introduction
Composite laminates have found increased use in mechanical applications over the last years, especially in areas related to aerospace, civil, and mechanical systems. Efforts have been particularly intense on damage assessment of large composite structures. There are many models such as the ones developed by Shokrieh [1] that focus on the degradation of mechanical properties, relying on experimental curve fits which do not necessarily have a direct physical interpretation. This can represent a large obstacle for understanding the mechanisms of onset and propagation of damage over a component or structure.
Of these damage modes, matrix cracking has received the most attention, followed by delamination. After the material reaches certain level of crack density, there is progressively less space for transverse crack formation. This phenomenon is known as crack saturation [2] and the corresponding density is known as Characteristic Damage State (CDS), but before this happens, delamination may occur. Delamination [3] is the mechanism of separation between plies caused by an interfacial weakening partially due to previously formed intralaminar cracks. Delamination proves to be more catastrophic than matrix cracking, because it could render the structure useless in a short amount of time driving it to failure [4].
Matrix cracking in 90-degree plies was first studied by Reifsneider [5], who experimentally investigated this phenomenon on [0/±45/90]s laminates subjected to uniaxial quasistatic and cyclic tests. Although practical, physical experiments can become relatively expensive when less assumption is considered, especially when compared with ever-increasing computer capacity [6].
The first micromechanical model was shear-lag, introduced by Garrett and Bailey [7–9] and significantly improved by Cox [10]. Shear-lag model relies on the basic principle that relaxed stresses arising from the formation of intralaminar cracks transfer to ply interface [11] in form of shear stress, increasing in turn the occurrence of delamination. In the present review, Berthelot complete parabolic shear-lag model is used [12], which means a parabolic variation of axial displacement is introduced for both 0- and 90-degree layers (see (1a)-(1b)). Then, corresponding stress functions are obtained, which ultimately translate into stress accumulation and degradation mechanical properties, particularly stiffness [13].
Another family of micromechanical approaches is proposed by Hashin [14] who proposes a variational stress-perturbation function, departing from crack-free stress state. In order to achieve this, Hashin applied the minimum complementary energy principle and optimized it. Subsequently, the total stress field (crack-free plus perturbation) was recovered for cross-ply (0/90m)s laminates. Similar to the first approach, the stresses formed cracks, which in turn translated in material degradation, chiefly stiffness reduction. Hashin’s variational analysis has become a seminal work in the field of micromechanical damage analysis on composite materials [14]. For example, it has been used and adapted for predicting fatigue [15], random crack growth [16], and delamination [17]. Thereafter, McCartney [18] introduced his model based on Hashin’s variational approach and classical elasticity, presenting three novelties: inclusion of the thermal expansion strains, explicit expressions for displacements, and a further expansion to three-dimensional stress evaluation that will not be reviewed in this work.
Continuum damage mechanics (CDM) is another family that has become more commonplace in literature over the last years due to increased computational capacity. These models are based on updating stiffness matrices in function of crack opening displacement (COD) or density. Their scheme may be iterative, such as the so-called self-consistent methods [19, 20] and the proposed by Barbero [21–23], or departing from an initial set of real-life [24] or virtual finite element simulations such as the Synergistic Damage Mechanics (SDM) model presented by Singh and Talreja [25–27], which expanded the scope of micromechanical modelling to laminates other than cross-ply.
Driven by aerospace and wind energy applications in which other laminate configurations such as the [0/±45/90]s are preferred, recent models such as the ones proposed by Hajikazemi [28] and Adumitroaie [29] have also expanded to laminates other than cross-ply. However, as these models are usually extensions of previous models (Hashin’s variational for the case of Hajikazemi’s model and SDM for Adumitroaie’s model), the main goal of this work is drawing a comparison between these four approaches (shear-lag, Hashin’s variational, McCartney, and SDM) and highlighting the advantages and setbacks for each one. This comparison will be helpful for selecting the most appropriate matrix cracking approach for future works related to damage propagation in fiber-reinforced polymers for different applications. The nomenclature used throughout the article is summarized in the Nomenclature.
2. Models2.1. Shear-Lag Model
As mentioned above, the shear-lag models were the first approaches on intralaminar cracking. The basic governing principle consists in stress redistribution along the loading direction causing an increase in interfacial shear-load due to crack onset. The basic unit cell is shown in Figure 1, representing a symmetric (0/90)s laminate element composed of two 0°-degree outer plies of thickness t0 and a 90°-degree inner ply of thickness t90. Total laminate thickness is thus h=2(t0+t90). In addition, x is loading direction, 2l is crack spacing, and ply interface is located along coordinate z=t90.
Unit cell for shear-load and variational methods. (a) Geometry. (b) Equilibrium. The crack is supposed to span all the laminate width, adapted from [12, 14].
Shear-lag model is considerably simple compared to the other approaches presented below because of the following assumptions: (a) plane stress and (b) longitudinal stresses σxx are assumed to be uniform across the thickness of 0° and 90° plies and (c) the tensile load applied to 0° plies is considered to be transmitted into the 90° ply by shear of the transverse ply [12].
However, these assumptions may compromise the accuracy of the model. For this reason, Berthelot [12] introduced parabolic profiles for x-direction displacements in 0- and 90-degree plies.(1a)u0x,z=A0xz2+B0xz+C0x(1b)u90x,z=A90xz2+B90xz+C90x,where A, B, and C are integration constants. These profiles are then introduced within the framework of classic equations for elasticity: first kinematics (strain-displacement), then compliance, and finally equilibrium within the unit cell shown in Figure 1(b) [12]. (2)∂σxx90∂x+∂σxz90∂z=0.This produces the boundary value problem (BVP) with governing equation in terms of average longitudinal stress in the 90-degree layer (σ-xx90)(3a)d2σ-xx90dx2-λ2σ-xx90=-λ2E22Ex0σc,where λ2=3Gxy0t0+t90Ex0/t902t0E11E22, and the laminate stiffness constants for shear Gxy0=G23/(1+t0G23/t90G12) and Young’s modulus Ex0=(t0E11+t90E22)/(t0+t90), subject to a global load σc.
This governing equation is subject to the boundary conditions (BCs) shown in Table 1.
Comparison of boundary conditions of the four models analyzed in the present article.
Boundary conditions governing shear lag model [12]
(i) Traction free crack surfaces
σ-xx90l=0
(ii) Interlaminar shear stress symmetry condition
τ0=0
Boundary conditions governing variational model [14]
(i) Symmetry
σxz90x,0=0
(ii) Traction continuity
σxz90x,t90=σxz0x,t90
σzz90x,t90=σzz0x,t90
(iii) Traction-free end
σxz0x,h=0
σzz0x,h=0
(iv) Traction-free crack surfaces
σxx90±a,z=-σ90
σxz90±a,z=0
Boundary conditions governing McCartney model [18]
(i) Perfect interlaminar bonding
σxxm=σxxf,σxym=σxyf@x=a,y≤L
um=uf,vm=vf@x=a,y≤L
(ii) Stress-free external surface of outer plies
σxxf=0,σxyf=0@x=h,y≤L
(iii) Symmetry about y-axis
σxym=0,um=0@x=0,y≤L
(iv) Symmetry about x-axis
σxym=0,vm=0@y=0,0≤x≤a
σxyf=0,vf=0@y=0,a≤x≤h
(v) Stress-free crack surface
σyym=0,σxym=0@y=L,0≤x≤a
(vi) Bridging outer plies
σxyf=0@y=L,a≤x≤h
vf=Lϵx@y=L,a≤x≤h
Boundary conditions governing SDM model [25]
(i) Symmetry about midplane
w=0@z=0
(ii) Displacement condition at right face
u=u0@x=2l=sα
(iii) Fixture at left face
u=0@x=0
Thus, the complete solution for x-direction stress field in 0-degree and 90-degree plies is given out by(4a)σ-xx90x=σcE22Ex01-coshηax/lcoshηa(4b)σ-xx0x=σcE11Ex01+t90t0E22E11coshηax/lcoshηa(4c)σxx0x,z=σ-xx0x-σcE112G12E22Ex0t90t0η2zt902-21+αzt90+23α2+2α+1coshηax/lcoshηa(4d)σxx90x,z=σ-xx90x+σcE2222G23Ex0η2zt902-13coshηax/lcoshηa(4e)a=lt90(4f)η2=λ2t902=3Gxy0t0+t90Ex0t0E11E22(4g)α=t0t90,where (4e) expresses the aspect ratio a. This parameter is of great importance as it reflects the relationship of crack spacing to ply thickness, having a profound impact on the degradation model. Meanwhile, (4g) shows the stacking parameter α. Shear-lag method calculates in a straightforward fashion the interfacial shear-load that may be applied later to delamination models. The complete model for interlaminar shear stress is also obtained as follows: (5)τx=σcE22E11ηsinhηax/lcoshηa.The shear stress field is obtained introducing (4c)-(4d) into (2) and solving for σxz. Subsequently the z-direction axial stress field is obtained introducing this solution into equilibrium equation (6) and solving for σzz. This procedure is repeated for the 0-degree ply.(6)∂σxz90∂x+∂σzz90∂z=0.Finally, these stresses have an effect on the material, which is reflected in degradation rules such as the stiffness reduction formula obtained by Ogin [13]:(7)ExEx0=11+t90/t0E22/E111/ηatanhηa.
2.2. Variational Method
In 1985, Hashin proposed another approach to solve the crack problem in a cross-ply laminate, already shown in Figure 1. The variational method is an energetic approach that consists in posing an internal energy formulation for a perturbation stress, which is considered a variation from the intact material stress, caused by crack onset. Because this is an equilibrium situation, total energy change is considered near zero, with this method being therefore soundly based on virtual work statics.
The assumptions taken for this approach are as follows:
The normal stress in external load direction is constant over ply thickness.
Shear stresses develop only within a boundary layer of unknown thickness in between plies.
Cracks remain sufficiently far apart so that their mutual interaction can be neglected.
Said this, the stresses in the cracked laminate are(8)σijm=σij,0m+σij,pm,
where i,j subscripts stand for x,z direction of the stresses. The index m represents the ply, whether 0- or 90-degree. Finally, the subscript 0 represents the stress of the intact matrix and p, the perturbation stress generated by the crack. After applying the equilibrium equations obtained from basic elasticity over the region illustrated in Figure 1 and the boundary conditions listed in Table 1, the perturbation stress field in terms of intact 90-degree ply stress (σ90) is found out.(9a)σxx,p0x,z=-σ90ϕx(9b)σxz,p0x,z=σ90ϕ′xz(9c)σzz,p0x,z=σ90ϕ′′x12ht90-z2(9d)σxx,p90x,z=σ90t90t0ϕx(9e)σxz,p90x,z=σ90t90t0ϕ′xh-z(9f)σzz,p90x,z=σ90t90t0ϕ′′x12h-z2,where the perturbation function ϕ(x) is evaluated by the following formulation, provided that in most epoxy/carbon materials, 4q>p2:(10a)ϕ=A1coshα1ξcosα2ξ+A2sinhα1ξsinα2ξ(10b)ξ=xt90(10c)A1=2α1coshα1asinα2a+α2sinhα1acosα2aα1sin2α2a+α2sinh2α1a(10d)A2=2α2coshα1asinα2a-α1sinhα1acosα2aα1sin2α2a+α2sinh2α1a,where α1=q1/4cos(1/2)Θ, α2=q1/4sin(1/2)Θ, and tanΘ=4q/p2-1. The material constants p and q may be evaluated as follows:(11a)p=C02-C11C22;q=C00C22(11b)C00=1E22+1αE11;C02=ν23E22α+23-ν12α3E11(11c)C22=α+13α2+12α+860E22;C11=131G23+αG12.For the equations governing case 4q<p2, the reader is referred to Talreja [11]. As in the previous method, a stress field is recovered and then applied in a degradation rule that has an effect on material’s elasticity. (12a)1Ex=1Ex0+1E22k12μαaχa,where k1=σ90/σc and is obtained from traditional laminate analysis. The other functions are evaluated as follows:(12b)χa=-ϕ′′′ξ=a(12c)μα=3α2+12α+860.Finally, it is important to take into account the fact that this formulation is only applicable to equally spaced cracks. If the separation is arbitrary, a probabilistic distribution such as those used by Vinogradov [16] must be used.
2.3. McCartney Method
McCartney proposed another analytical method in 1992 [18]. Similar to variational approach, the method models the same phenomenon by optimizing an energetic expression. As in the previous two models, it only considers cross-ply laminates. Nevertheless, there are three noteworthy novelties. First of all, components of thermal expansion are taken into account for stain and stress modelling. Then, even though precision is similar to Hashin, McCartney model includes explicit expressions for displacement field. Finally, in contrast to shear-lag, the McCartney model satisfies exactly a more comprehensive set of boundary conditions. The basic unit cell of the laminate between two cracks, subjected to uniaxial stress, is shown in Figure 2(a).
(a) Unit cell used to derive the equations of the McCartney model. (b) Coordinates for boundary conditions shown in Table 1. Adapted from [18].
Then, from basic elasticity and including thermal strains, the field equations for generalized plane strain conditions are expressed by (13)∂σxx∂x+∂σxz∂z=0,∂σxz∂x+∂σzz∂z=0ϵxx=-ν12′E11′σzz+1E11′σxx+αA′ΔT-ν12E22E11ϵc∗=∂u∂xϵzz=1E22′σzz-ν12′E11′σxx+αT′ΔT-ν23ϵc∗=∂w∂zϵxz=σxz2G12=12∂u∂z+∂w∂x.It is important to keep in mind that (13) actually represents two sets of equations, one for outer 0-degree plies and another one for inner 90-degree plies. Engineering elastic constants that feature an apostrophe are the modified moduli, which are shown in more details in [18]. The original moduli appear with no apostrophe. Finally, the development for the uniform transverse strain in the cracked laminate ϵc∗ is also shown in the referred paper. Assuming perfect interlaminar bonding, McCartney model is governed by the boundary conditions shown in Table 1 (cf. Figure 2(b)), where ϵc is the longitudinal strain experienced by the cracked composite and is also explained in further details in [18]. Now, assume that longitudinal stress components σxx0 and σxx90 are independent of z and are of the form(14)σxx0=Cx+E11′ϵ+v12E22E11ϵ∗-αA′ΔT=Cx+σ0σxx90=-t0t90Cy+E22′ϵ+v12ϵ∗-αT′ΔT=-t0t90Cx+σ90,where ϵ and ϵ∗ are the corresponding longitudinal and transverse strains for the undamaged laminate. We now proceed to introduce (14) in field (1a) and (1b) with boundary conditions (Table 1) in order to obtain the following stress field:(15)σxz0=C′xh-zσzz0=12C′′xh-z2σxz90=t0t90C′xzσzz90=12t0t90C′′xht90-z2.It is interesting to note the similarities between (10a), (10b), (10c), and (10d) from Hashin and (14) and (15) from McCartney. Now, the corresponding displacement field (16a), (16b), (16c), and (16d) is obtained(16a)w90=t0t90v23′E22′Cxz-16t0t901E22′C′′xzz2-3t90h-v23′E22′σ90-αT′ΔT+v12ϵ∗z(16b)w0=v23′E22′t0-v12′E11′z-t90Cx+16E22′t03-h-z3+13E22′t0t90h+12t0C′′x-v12′E11′σ0-αT′ΔT+v23ϵ∗z-t90-v23′E22′σ90-αT′ΔT+v12ϵ∗t90(16c)u0=v12′2E11′z-t902-v23′E22′t0z-t90-12G12z-h2-t02C′x-124E22′z-h4+4t03z-t90-t04+13E22′t0t90h+12t0z-t90C′′′x+Ax(16d)u90=t02t901G23-v23′E22′z2-t902C′x+124E22′t0t90z4-t904-6t90hz2-t902C′′′x+Ax,where A(x) is the x-displacement at the interface of inner and outer plies(17)Ax=1E11′C-x-v12′3E11′-v23′2E22′+13G12t02C′x+120E22′+112E22′t90t02t90t0+3t04C′′′x+ϵx.And perturbation function C(x), similar to Hashin’s, is defined for most composites as(18a)Cx=Pcoshp1xt0cosq1xt0+Qsinhp1xt0sinq1xt0(18b)p1=12G2F+HF,q1=12G2F-HF(18c)P=Λq1sinhp1lt0cosq1lt0+p1coshp1lt0sinq1lt0t90t0σ90,Q=Λq1coshp1lt90sinq1lt90-p1sinhp1lt90cosq1lt90t90t0σ90(18d)1Λ=q1sinhp1lt0coshp1lt0+p1sinq1lt0cosq1lt0.Here F, G, and H are constants dependent on geometry and material and are defined thoroughly in [18]. Then, the function C-x is computed from the following integral:(19)C-x=∫0xCx′dx′.Once the perturbation function is obtained it is used to calculate stiffness reduction in the cracked laminate.(20a)1Ex=1Ex0+1Ξ1-νxy0v12E22′Ex0t902hlHΦ(20b)Φ=2Λp1q1p12+q12cosh2p1lt0-cos2q1lt0(20c)Ξ=1-v12-v12E22E112t02hlΦHEy0,where νxy0, Ex0, and Ey0 are the longitudinal Poisson ratio and longitudinal and transverse Young’s moduli of the undamaged laminate, respectively. Finally, in the same work, McCartney introduces a novel 3D version of the formerly explained model, which is not included in this analysis for the sake of comparison, and because the plane strain assumption under which the 2D model operates is enough for adequately modelling the mechanics of cracked composite plies.
2.4. Synergistic Damage Mechanics (SDM)
Up to now, micromechanical models shown in Sections 2.1 to 2.3 refer to cracked composites limited to the cross-ply case, which has very limited applications. For most aerospace and wind energy applications, a [0/±45/90]s configuration is more common. For this reason, Singh and Talreja developed the Synergistic Damage Mechanics (SDM) model [25], which is explained as follows.
The basic cell under which the SDM model operates is shown in Figure 3(a). The picture clearly shows that the model is flexible for stacking more arbitrary ply orientations than those presented previously. For the sake of comparison against previous shown models, the basic cell used for simulations shown in Section 3 is also cross-ply [24], Figure 3(b); nevertheless, in this section the general case is described in order to discuss all the innovations the SDM model poses.
(a) Representative unit cell for a laminate under SDM micromechanics analysis, adapted from [25]. (b) Special case for cross-ply analysis, adapted from [24].
Of these innovations, the most important over the previous micromechanical models is that SDM can initiate and multiply cracks in any of the layers of the laminate, even if the probability of this occurring at 0° is relatively low. Each of these crack orientations is defined in SDM as damage modes, denoted by α. As it will be described, each of these modes will have their particular stress fields, stiffness changes, and damage tensors as each one of them contributes individually to the total damage state and stiffness reduction in the laminate.
The model starts with a damage state tensor Dijα, with a given spacing (sα), thickness (tα), restriction parameter (κα), and orientation ninj for the damage mode α. Finally, the total laminate thickness is represented by t.(21)Dijα=καtα2sαtninj.The normal orientation vector is expressed by ni=(sinθ,cosθ,0). Meanwhile, the restriction parameter κα accounts for the constraining effect on ply cracks caused by adjacent plies in the laminate. (22)κα=u~nαϵ22tα,where ϵ22 is the transformed applied strain component normal to the crack surface and u~nα is the normal average crack opening displacement (COD) for a given crack spacing. These averaged CODs are computed previously from a series of FE simulations using software such as ANSYS [30]. The micromechanical unit cell (Figure 3) is modelled by replicating the material stacking and elastic constants, defined by the size of crack spacing, and three boundary conditions [25] shown in Table 1.
Another important quantity resulting from the model is the stiffness tensor and its reduction caused by the presence of the microcracks. For the case being [25], the damage stiffness tensors for each mode Cpqα are expressed by(23a)Cpq1+Cpq2=κθtθsθt2a11,2a41,20a41,22a21,20002a31,2=Dθai1,2(23b)Cpq3=κ90t90s90t2a13a430a432a230002a33=D90ai3(23c)Cpq4=κ0t0s0t2a14a440a442a240002a34=D0ai4.
This stiffness reduction scheme needs the determination of material constants aijα. In order to obtain these constants, stiffness reduction curves are taken from previously described ANSYS simulations over [0/90]s laminates where the values of E11, E22, G12, and v12 are plotted against the value of crack spacing sα for each ply. Finally, a curve fitting procedure is conducted in order to satisfy the following set of equations: (24a)Cpq=Ex01-vxy0vyx0vxy0Ey01-vxy0vyx00vxy0Ey01-vxy0vyx0Ey01-vxy0vyx0000Gxy0-∑αaαDθaiα,where the 0-superscripted engineering constants correspond to those of the virgin laminate. For a [0/±θ/90]s laminate, the total damage tensor is given by(24b)∑αaαDθaiα=∑αCpqα=2Cpq1+Cpq2+Cpq3+2Cpq4.Once the model is ready, (24a) and (24b) is applied again, but now for the current crack separation parameter present in the laminate. The final results are the degraded elasticity engineering constants of the material. (25)Ex=C11C22-C122C22,Ey=C11C22-C122C22,vxy=C12C22,Gxy=G66,vyx=vxyEyEx.Subsequently, the resulting stress tensor is recalculated(26)σijα=CijklDijαϵkl.As stated previously, even if Synergistic Damage Mechanics (SDM) model has the capability of analyzing stresses and stiffness degradation in laminates such as [0/±45/90]s, it is first necessary to perform simulations on cross-ply laminates. As in all other stiffness degradation models, SDM stiffness reduction for [0/90]s glass/epoxy laminates is plotted against crack density, which is also in terms of the reciprocal of crack spacing as per (27). This means that for each case of crack density (sθ = 16, 8, 4, 2, 1, 0.75, 0.5, 0.2, and 0.1 mm) FE models were prepared and analyzed using ANSYS by applying loading, geometry, and boundary conditions as shown in Figure 3(b). An intact case simulated with highly spaced cracking (sθ=32) was also prepared.(27)c=1sθ=12at90.Boundary conditions for each crack spacing case remained the same, with applied displacement at right face of ux=2l = 5 μm. Left face is fixed and there is a symmetry condition in the bottom face, along z-axis. Cracks were supposed to have grown throughout the width of the cell. Subsequently, each model was meshed using 10,000–25,000 3D tetrahedral elements. Then, stress field was recovered and averaged along the volume. At last, longitudinal Young’s modulus is then obtained, according to [31], by applying the following equation:(28)Ex=σxxϵxx=1/VRVE∫VσxxdVu0/2l,where VRVE is the volume of the whole unit cell, ∫VσxxdV~∑i(σxx,iVi), where σxx,i and Vi are the normal axial stress and volume of element i; u0 is the applied displacement in the right face, as well. Finally, these curves can later be used for obtaining through fitting, materials constants aijα, which can later be used in (24a) and (24b) in order to calculate stiffness degradation in stacking cases such as [0/45/90]s.
3. Results and Discussion
From the formulation explained in Section 2, stress distribution is drawn for longitudinal stress σxx in 90-degree layer for an E-glass fiber/epoxy laminate with material constants (Table 2) obtained from Highsmith and Reifsneider [5], with 90-ply thickness t90 = 0.203 mm, stacking parameter α=1, and aspect ratio of cracking a = 2.5 and k1 = 0.464. Cracking aspect ratio and 90-ply thickness will be kept constant through results analysis; for other values their variation will be presented as needed.
Elastic engineering constants for glass-fiber/epoxy and carbon fiber/epoxy composites [5].
Property
GFRP
CFRP
Longitudinal Young’s modulus (E11), GPa
41.7
208.3
Transverse Young’s modulus (E22), GPa
13.0
6.5
Longitudinal shear modulus (G12=G13), GPa
3.4
1.65
Transverse shear modulus (G23), GPa
4.58
2.3
Longitudinal Poisson’s ratio (v12=v13)
0.3
0.255
Transverse Poisson’s ratio (v23)
0.45
0.413
It is seen from Figure 4 that minimum normal stress is zero and is found where the cracks are located; meanwhile the maximum longitudinal stress is found midway between consecutive cracks. Furthermore, it is observed from Figure 5 that crack onset has increased interlaminar shear.
Normal longitudinal stress distribution σxx in 90-degree layer for a glass fiber/epoxy cross-ply 0/90s laminate using Berthelot-modified shear-lag model [12].
Variation of interlaminar shear in 0/90 interface for a glass fiber/epoxy cross-ply 0/90s composite using Berthelot-modified shear-lag model [12].
Nevertheless, shear-lag method is at disadvantage when compared to variational micromechanics because the shear-lag assumption forces the model not to follow the zero-traction boundary condition, in stark comparison to Hashin formulation. This brings considerable error for the shear-lag method as it approaches the crack surface (Figure 6).
Comparison of longitudinal stress obtained by shear-lag and variational methods for 90-degree ply in a glass fiber/epoxy cross-ply 0/90s composite under longitudinal stress.
Taking the shear stress variation for the interface located between 0- and 90-degree plies from Hashin variational method (Figure 7(a)) [14], it is observed from Figure 7 that shear stress redistributes to somewhere in the middle between crack surface and the z-axis of symmetry because of the restriction imposed on crack surface. Again, this contrasts with Berthelot shear-lag model, which puts the maximum location of shear redistribution just above the cracking surface (Figure 5). Finally, a complete stress distribution field for σxz90 by Hashin method is illustrated in Figure 7(b), from where it is observed that σxz90 becomes progressively higher as it approaches the interlaminar interface.
(a) Interfacial shear stress variation along the interface between 0 and 90 plies of a glass fiber/epoxy cross-ply 0/90s laminate under longitudinal stress and (b) shear stress distribution for a glass fiber/epoxy cross-ply laminate under longitudinal stress, both using variational method [14].
Afterwards, degradation of longitudinal Young’s modulus D=Ex/Ex0 against crack density c, (27) for [0/90]s glass/epoxy laminates (α=1,k1=0.464), is plotted for all four stiffness reduction models (Figure 8). For shear-lag model used by Berthelot [12], Ogin model stiffness reduction curve is applied [13]. Meanwhile, for the Hashin variational method [14], Reifsneider formulation was employed instead [5]. McCartney [18] and SDM [25] incorporate their own stiffness reduction methods. As a first verification step, convergence for all four curves is verified against (29) proposed by Reifsneider [5]; there is very good agreement between all model results.(29)1Ex=1Ex0+k12α+11E22+1αE11.Regarding experimental results, a comparison was drawn between all four models and Smith and Wood experimentation [32]. Again, from Figure 8 it is seen that only shear-lag and SDM follow good agreement at initial cracking stages. Smith and Wood do not provide experimental results for advanced (c>2 cracks/mm), nonetheless convergence has been proved with (29).
Curve comparing the four methods covered in this paper for longitudinal Young’s modulus reduction on a 0/90s glass fiber/epoxy laminate.
4. Conclusions
A comparison between four micromechanical models has been presented. From degradation rules, it is noticed that all four methods covered in this paper predicted convergence value quite accurately. From Hashin’s variational method it is important to remark that it not only predicts correctly the boundary conditions at the surfaces of the crack but also readily gives out the complete stress field. Although limited by model assumptions, there are novel methods capable of analyzing arbitrary stacking [28] based on Hashin’s approach. In contrast, shear-lag degradation model is more accurate when compared against experimental results [32] but it takes additional differentiation and integration steps in order to obtain the final stress field. However, the assumptions under which it is governed limit its application to cross-ply laminates.
McCartney’s approach shows further improvement over the last two methods because it takes the generalized plane strain assumption, which soundly resonates with laminate mechanics. This model is also unique because it takes into account thermal strains and stresses. Furthermore, this model is applied to multiaxial loading conditions. By last, McCartney provides explicit expressions for computing displacements throughout the laminate cell. However, McCartney model is computationally expensive because of higher order differentiation involved in the computation of stiffness reduction.
For the case of SDM, the model not only converges but also replicates experimental results with relative accuracy. Furthermore, SDM is designed for numerical experimentation, implying a more economical alternative in comparison to traditional physical experimentation. Damage mechanics models present the advantage of being flexible because they may be applied to a wider range of laminate stacking, going beyond the classical cross-ply case and thus proving more useful in practical cases that involve laminates of arbitrary stacking [23, 25], such as the widely used [0/±45/90]s laminates. However, these modelling approaches may be computationally expensive and heavily dependent on mesh size for the case of virtual simulations.
NomenclatureA(x):
Interlaminar region displacement function (only McCartney model)
A1,A2:
Integration constants (variational method)
A0,B0,C0:
Integration constants for parabolic displacement profile in 0-degree plies (shear-lag)
A90,B90,C90:
Integration constants for parabolic displacement profile in 90-degree plies (shear-lag)
C(x):
McCartney perturbation function
C00,C22,C02,C11:
Variational method material constants
Cpq:
Damaged stiffness tensor
Cpq0:
Undamaged stiffness tensor
Dijα:
SDM damage tensor
E11:
Ply longitudinal Young’s modulus
E22:
Ply transverse Young’s modulus
Ex0:
Intact laminate longitudinal Young’s modulus
Ey0:
Intact laminate transverse Young’s modulus
Ex:
Cracked laminate longitudinal Young’s modulus
F,G,H:
McCartney material constants
G12:
Ply in-plane shear modulus
G23:
Ply out-of-plane shear modulus
P,Q:
McCartney integration constants
VRVE:
Volume of unit cell
a:
Crack aspect ratio
aijα:
SDM material constants for α ply
c:
Crack density
h:
Half-laminate thickness
k1:
Stress ratio in cross-ply laminates, obtained from classical lamination theory (CLT)
l:
Half crack spacing
ni,nj:
Normal orientation vectors
p,q:
Variational method material constants
p1,q1:
McCartney integration constants
sα:
Crack spacing, that is, sα=2l
t0,t90,t:
Ply thickness
u,v,w:
Displacements
u0:
Applied displacement
u~n:
Normal average crack opening displacement
x,y,z:
Positions
ΔT:
Temperature increase
Θ:
Variational method integration constant
Λ:
McCartney integration constant
Ξ,Φ:
McCartney material degradation constants
α:
Laminate stacking parameter
α1,α2:
Variational method integration constants
αA,αT:
Longitudinal and transverse thermal expansion coefficients
ϵ,ϵ∗:
Longitudinal and transverse strains in undamaged laminates
ϵc,ϵc∗:
Longitudinal and transverse strains in cracked laminates
ϵii,ϵij:
Normal and shear strain
η:
Shear-lag integration constant
θ:
Ply orientation respect loading axis
κ:
SDM interlaminar restriction parameter
λ:
Shear-lag integration constant
v12:
In-plane ply Poisson’s ratio
v23:
Out-of-plane ply Poisson’s ratio
vxy0:
In-plane intact laminate Poisson’s ration
ξ:
Normalized x-position coordinate
σ0,σ90:
Stress in intact 0- and 90-degree plies
σc:
Longitudinal load applied on laminate
σii:
Normal stress
σij:
Shear stress
σ-xx:
Average longitudinal stress
τ(x):
Interlaminar shear stress function
ϕ:
Variation method perturbation function.
Conflicts of Interest
The authors declare that there are no conflicts of interest regarding the publication of this paper.
ShokriehM. M.EsmkhaniM.Fatigue life prediction of nanoparticle/fibrous polymeric composites based on the micromechanical and normalized stiffness degradation approachesReifsniderK. L.Some fundamental aspects of the fatigue and fracture response of composite materialsProceedings of the Fourteenth Annual Meeting in Recent Advances in Engineering ScienceNovember 1977Bethlehem, Pa, USA373384KimT. W.KimH. J.ImS.Delamination crack originating from transverse cracking in cross-ply composite laminates under extensionBrunnerA. J.StelzerS.PinterG.TerrasiG. P.Cyclic fatigue delamination of carbon fiber-reinforced polymer-matrix composites: data analysis and design considerationsHighsmithA. L.ReifsniderK. L.Stiffness-reduction mechanisms in composite laminatesDaiG.MishnaevskyL.Damage evolution in nanoclay-reinforced polymers: a three-dimensional computational studyGarrettK. W.BaileyJ. E.Multiple transverse fracture in 90° cross-ply laminates of a glass fibre-reinforced polyesterParviziA.GarrettK. W.BaileyJ. E.Constrained cracking in glass fibre-reinforced epoxy cross-ply laminatesBaderM. G.BaileyJ. E.ParviziA.CurtisP. T.The mechanisms of initiation and development of damage in multi-axial fibre-reinforced plastics laminatesCoxH. L.The elasticity and strength of paper and other fibrous materialsTalrejaR.SinghC. V.BerthelotJ.-M.LeblondP.El MahiA.Le CorreJ.-F.Transverse cracking of cross-ply laminates: Part 1. AnalysisOginS. L.SmithP. A.BeaumontP. W. R.Matrix cracking and stiffness reduction during the fatigue of a (0/90)s GFRP laminateHashinZ.Analysis of cracked laminates: a variational approachNairnJ. A.The strain energy release rate of composite microcracking: a variational approachVinogradovV.HashinZ.Probabilistic energy based model for prediction of transverse cracking in cross-ply laminatesAkshantalaN. V.TalrejaR.A micromechanics based model for predicting fatigue life of composite laminatesMcCartneyL. N.Theory of stress transfer in a 0°-90°-0° cross-ply laminate containing a parallel array of transverse cracksQuJ.HoisethK.Evolution of transverse matrix cracking in cross-ply laminatesKanaunS. K.LevinV.BarberoE. J.DamianiT. M.TrovillionJ.Micromechanics of fabric reinforced composites with periodic microstructureBarberoE. J.CortesD. H.A mechanistic model for transverse damage initiation, evolution, and stiffness reduction in laminated compositesGhayourM.Hosseini-ToudeshkyH.JalalvandM.BarberoE. J.Micro/macro approach for prediction of matrix cracking evolution in laminated compositesJoffeR.KrasnikovsA.VarnaJ.COD-based simulation of transverse cracking and stiffness reduction in [S/90n]s laminatesSinghC. V.TalrejaR.Analysis of multiple off-axis ply cracks in composite laminatesSinghC. V.TalrejaR.A synergistic damage mechanics approach to mechanical response of composite laminates with ply cracksMontesanoJ.SinghC. V.A synergistic damage mechanics based multiscale model for composite laminates subjected to multiaxial strainsHajikazemiM.SadrM. H.VarnaJ.Analysis of cracked general cross-ply laminates under general bending loads: a variational approachAdumitroaieA.BarberoE. J.SchagerlM.Matrix cracking in non-symmetric laminates under combined membrane and flexural loadingANSYS® Workbench, Release 17.1, http://www.ansys.com/SinghC. V.SmithP. A.WoodJ. R.Poisson's ratio as a damage parameter in the static tensile loading of simple crossply laminates