You are on page 1of 17

Engineering Fracture Mechanics 118 (2014) 49–65

Contents lists available at ScienceDirect

Engineering Fracture Mechanics
journal homepage: www.elsevier.com/locate/engfracmech

Determination of the mode I crack resistance curve of polymer
composites using the size-effect law
G. Catalanotti ⇑, A. Arteiro, M. Hayati, P.P. Camanho
DEMec, Faculdade de Engenharia, Universidade do Porto, Rua Dr. Roberto Frias, 4200-465 Porto, Portugal

a r t i c l e

i n f o

Article history:
Received 29 May 2013
Received in revised form 2 October 2013
Accepted 31 October 2013

Keywords:
Fibre reinforced materials
Fracture mechanics
R-curves
Size effects

a b s t r a c t
This paper presents a new method to measure the crack resistance curve associated with
the longitudinal failure of polymer composites reinforced by unidirectional fibres. Rather
than using compact tension test specimens, the identification of the size-effect law of double edge notched specimens is used to obtain the crack resistance curve. Special emphasis
is placed on the appropriate calculation of the stress intensity factor of the specimens when
using quasi-isotropic or cross-ply laminates. For this purpose, both analytical closed-form
solutions and numerical methods are investigated. Four different carbon-epoxy material
systems, T800/M21, IM7/8552, T700/AR-2527, and T700/ACE are tested and the corresponding size effect laws and R-curves are measured. A good correlation between the crack
resistance curve obtained using the size effect law and that previously measured for one of
the material systems using the compact tension test is obtained. The highest value of the
longitudinal fracture toughness was obtained for the T800/M21 material.
Ó 2013 Elsevier Ltd. All rights reserved.

1. Introduction
The most recent analysis methods that predict fracture of polymer composite materials require not only the value of the
fracture toughness, but also its relation with the increment of the crack length, i.e., the crack resistance curve. Taking the
thickness of the individual ply as the representative length scale it is possible to formulate ‘mesomodels’ that account for
both delamination (interlaminar cracking) and ply failure mechanisms (intralaminar cracking) [1–3]. The softening constitutive relation that simulates longitudinal failure, where the fracture plane is approximately perpendicular to the fibre direction, requires the fracture toughness to regularize the numerical solution [3]; however, the crack resistance curve must also
be measured to identify the different regions of the softening constitutive relation so that the failure mechanisms acting at
the crack tip and along the wake of the crack are properly accounted for [4].
Recently, Finite Fracture Mechanics models that use the laminate thickness as the representative length-scale have been
developed to predict fracture of multidirectional composite laminates in the presence of stress concentrations [5–7]. These
methods are typically used for the preliminary design and optimisation of composite structures, and are based on the simultaneous fulfilment of a stress-based criterion, which requires a stress allowable, and of an energy based criterion, which
requires the fracture toughness [5–7] or the crack resistance curve [8].

⇑ Corresponding author. Tel.: +351 22508 1716/1049; fax: +351 225081315.
E-mail address: giuseppe.catalanotti@fe.up.pt (G. Catalanotti).

http://dx.doi.org/10.1016/j.engfracmech.2013.10.021
0013-7944/Ó 2013 Elsevier Ltd. All rights reserved.

e. with respect to a. While a strong emphasis has been placed on the use of compact tension test specimens [9]. Catalanotti et al. respectively stress intensity factor half of the length of the specimen size of the element length of fracture process zone fitting parameter used in the bilogarithmic regression fit applied load peak load R-curve R-curves for the 0° ply an 90° ply. when crack propagation occurs between the plies of the laminate (i. respectively steady-state value of fracture toughness steady-state value of the fracture toughness of the 0° ply components of the compliance matrix computed in the x1  x2 coordinate system thickness of the specimen nodal displacement half of the width of the specimen preferred axes of the material nodal load shape parameter.  E f GI h h0 . a0 b. when crack propagation occurs within the individual plies of the laminate. initial value of the shape parameter parameters used in the R-curve fit crack increment error elastic parameter correction factor correction factor j for a ¼ a0 derivative.50 G. For example. N P Pu R R0 . recent results have shown that using the current geometry of the compact tension test specimen it is not possible to measure the fracture toughness of modern resin systems that result in high values of the fracture toughness [10]. of the correction factor j for a ¼ a0 matrix for the polynomial fitting of j elastic parameter shape-parameter elastic parameter remote stress ultimate nominal stress corrected value of the ultimate stress correction factor for an infinitely long specimen matrix for the polynomial fitting of / correction factor for the orthotropy of the material correction factor for the length of the specimen matrix for the polynomial fitting of w average value standard deviation Based on the above observations. c Da  f j j0 j 0 K k n q r ru r^ u / U v w W Avg. SD crack length. / Engineering Fracture Mechanics 118 (2014) 49–65 Nomenclature a. h90 KI l le lfpz M. and intralaminar. R90 Rss R0ss slm t ul w x1 . it becomes apparent that reliable test methods for the measurement of the intralaminar fracture toughness1 of composite laminates and of the corresponding crack resistance curve (R-curve) are required. C  C  A. initial value of the crack length fitting parameter used in the linear regression I fit fitting parameter used in the linear regression II fit equivalent modulus correction factor for the dimensionality of the specimen energy release rate in mode I thickness of the laminate thickness of the 0° and 90° plies. in previous attempts to measure the fracture 1 Two different types of failure mechanisms are usually considered in fibre reinforced composites: interlaminar. x2 Ym a. . a0 A. delamination).

In a two-dimensional orthotropic body. the objective of this paper is to develop a new methodology to obtain the R-curve of polymer composite laminates reinforced by unidirectional fibres. Therefore. the mode I component of the energy release rate for crack propagation in the x1 -direction. GI . The general overview of such relation is presented in Section 2. Analytical model 2. (1) is a function of q. taking x1 and x2 as the preferred axes of the material (see Fig. / Engineering Fracture Mechanics 118 (2014) 49–65 51 toughness of cross-ply Hexcel’s T800/M21 carbon-epoxy laminates using the geometry proposed in [9] the region of the specimen subjected to compressive stresses buckled [10]. The conclusions and suggestions for future work are presented in Section 4. 2. 1. The equivalent modulus reads: where KI and E ¼ E  s11 s22 1þq 2 1=2 k1=4 ð2Þ where slm are the components of the compliance matrix calculated in the x1  x2 coordinate system. The length of the specimen is 2l. the stress intensity factor of a notched bar can be written as: Fig. Double edge cracked specimen (DEC). 1. 1). which follows Bazˇant’s seminal work [11]. The main idea put forward here. Using the orthotropic rescaling technique [12.1. Catalanotti et al. and of the shape and size of the specimen. and k and q are two dimensionless elastic parameters defined as: k¼ s11 . such an elastic instability renders the test results meaningless.G. . Overview Consider the geometry shown in Fig. of the remote stress r. Section 3 describes the experiments that were performed in three different types of carbon-epoxy composite laminates and presents the corresponding R-curves. reads [12]: GI ¼ 1 2 K  I E ð1Þ  are respectively the stress intensity factor and the equivalent modulus. is to relate the size effect law with the crack resistance curve of the composite material. s22 q¼ 2 s12 þ s66 pffiffiffiffiffiffiffiffiffiffiffiffi 2 s11 s22 ð3Þ The stress intensity factor in Eq. The width of the specimen is equal to 2w and a0 is the initial crack length.13].

differentiating (8) with respect to w. lfpz . are tangent to R-curve. w) and remembering that the R-curve does not depend on the size of the specimen w (@R=@w ¼ 0). RðDaÞ.e. 2. (5) can be re-written as: GI ðDaÞ ¼  P2 Da j2 a0 þ . The correction factor for the double edge cracked specimen is given in Tada et al. [13]. i. Pun . Following [11]. Mathematically the peak load. f  w 4 w t2 E  ð6Þ where a0 is the initial value of the shape parameter. Under these circumstances it is possible to express j as: j ¼ f ðaÞ vðqÞ where ð10Þ v is the correction factor for the orthotropy of the material vðqÞ ¼ 1 þ 0:1ðq  1Þ  0:016ðq  1Þ2 þ 0:002ðq  1Þ3 ð11Þ f is the configuration correction factor for the dimensionality corresponding to the isotropic case. (4) will be a function of a and q only. and the corresponding formula for the calculation of the length of the fracture process zone. therefore.. Following Bazˇant and Planas [11]. Assuming that j is an increasing function of the crack length (the specimen has a positive geometry). wn . for a sufficiently long specimen the correction factor used in Eq. using (5) in the first of Eq. C j20 1  C  A.2. and for the fracture toughness at propagation. P u . N j20 A. are reported in Table 1 where j0 ¼ jja¼a0 and j 0 ¼ @ j=@ aja¼a0 . the crack driving-force curves GI corresponding to the peak loads. For different sizes. the effect of the parameter f is negligible when f 6 1=2. ru ¼ P u =ð2wtÞ. The size effect law should be determined by testing geometrically similar specimens. (9) can be solved for w ¼ wðDaÞ. when q ¼ 1. R. and (iii) the linear regression II. RðDaÞ ¼ 1 w  E ru ¼ ru ðwÞ. fÞ ð4Þ 1=4 where f ¼ k n. Determination of j According to Bao et al. Regressions fit Bilogarithmic Formula ln ru ¼ M ln pffiffiffiffiffiffiffiffi Nþw Linear regression I r2u ¼ A w þ C Linear regression II 1 wr2u 1  ¼A 1 w  þC Fitting parameters Rss M. Replacing w in (8) yields the R-curve. or the ultimate nominal stress. Rss . (ii) the linear regression I. [14]: Table 1 Size effect law fits [11]. The regression fits normally used. (7) yields: r2u j2 ð8Þ This equation is valid for every w. the size effect method can be used to measure the fracture toughness of the material [11]. j20 1  E  E  E M A  C lfpz 2 j0 0 2 j N j0 C 0 A 2 j j0 A 0 C  2 j . q. and this fact can be used to measure the R-curve. the following equation is obtained: @  w @w  r2u j2 ¼ 0 ð9Þ Eq. a0 ¼ a0 =w. Replacing (4) in (1) the energy release rate reads: GI ¼ 1 w  E r2 j2 ¼  2 Pj  t 4wE 1 ð5Þ where t is the thickness of the specimen and P is the applied load. q. three different kinds of fitting are generally used: (i) the bilogarithmic regression. can be obtained from the following system of equations: ( GI ðDaÞ ¼ RðDaÞ @GI ðDaÞ @ Da ð7Þ DaÞ ¼ @Rð @ Da Assuming that the size effect law is known.52 G. Eq. under the hypothesis that geometrically similar specimens are tested (a0 and n are not functions of the width. / Engineering Fracture Mechanics 118 (2014) 49–65 KI ¼ r pffiffiffiffi w jða. a ¼ a=w and n ¼ w=L are the shape-parameters. Catalanotti et al. and j is the correction factor that depends on the geometry and orthotropy of the material.

The length of the specimen of Fig. for a balanced cross ply. knowing the R-curve of the laminate. h0 ¼ h90 ¼ h=2. Since Eq. Taking advantage of the capabilities of the commercial software Abaqus [15]. R. taking values larger than 4. Taking into account j will be used in the test method to measure the R-curve of the ply in the longitudinal direction. If a cross-ply laminate is used. (10) is not applicable. is simply twice the fracture toughness of the laminate. a new expression for j should be found. / Engineering Fracture Mechanics 118 (2014) 49–65 f ðaÞ ¼ pffiffiffiffiffiffiffih 4 pa 1 þ 0:122 cos api 2 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi   ap 2 tan ap 2 53 ð12Þ For materials with 0 6 q 6 4. 2. (10) cannot be used. appropriate boundary conditions must imposed in the finite element model. for a cross-ply laminate. respectively. the crack driving force is equal to: GI ¼ Y m un =le ð15Þ where Y m and ul are the load and the displacement in the x2 direction of the nodes m and l. the fracture toughness in longitudinal direction. [9]. 2). h0 and h90 are the thickness of the laminate. Fig. respectively. The material properties used in the CPS4R elements are the laminate elastic properties. the R-curve of the ply in the longitudinal direction reads: R0 ¼ 2 R. the R-curve of the 0° plies can be obtained neglecting the fracture toughness of the 90° plies. and le is the size of the element (see Fig. q. it is possible to obtain easily the R-curve of the ply in the longitudinal direction (i. the solution of Eq. 1 corresponds to the free length of the specimen. The Virtual Crack Closure Technique [17] is used to calculate the energy release rate. In this case. 2).e. The energy balance for a self-similar crack propagation da in the cross-ply reads: R h da ¼ R0 h0 da þ R90 h90 da ð13Þ where R0 and R90 are the R-curves for the 0° and 90° plies. This means that. and n. Neglecting R90 (R90  R0 ). R. Consequently. (10) is applicable in the case of quasi-isotropic laminates (in this case v ¼ 1 and j ¼ f ). and n. j. One-quarter of the specimen was modelled using 4-node plane stress reduced integration elements (CPS4R) [15] (Fig. 1 and 2). the error obtained is negligible if j is obtained multiplying the correction factor for isotropic material of Eq. Taking into account the symmetry of the problem. which are either calculated from the ply elastic properties and lay-up using lamination theory (unidirectional reinforcements). or directly measured in experimental tests (woven fabrics). respectively. (11)) [13]. . As observed by Pinho et al. can be eliminated (because j ¼ 1 when s11 ¼ s22 ) and this parameter will only depend on a. R0 . q. Eq.G. the influence of k on the calculation of the correction factor. after simplifying. (12) for the correction for the orthotropy (Eq. The importance of determining the expression of j under this circumstances results from the fact that. Finite element model using for the calibration of j. in the fibre direction). a parametric model of the specimen was build in Python [16] to calculate the correction factor j for different values of a. therefore Eq. a displacement is applied at the end of the specimens (see Figs. the 0° and 90° plies. (13) can be rearranged as: R0 ¼ h R h0 ð14Þ If a balanced cross-ply is used. Therefore. or of multidirectional laminates that are not highly orthotropic (0 6 q 6 4). Catalanotti et al. The use of a balanced cross-ply to measure the fracture toughness of the 0° ply is proposed here. and h. therefore. the parameter q will not satisfy this inequality.

54 G. respectively. 4(a) and (b) compare the numerical results and the polynomial fitting surfaces for the two extreme cases of n ¼ 0 (infinitely long specimen) and n ¼ 1. 3 shows the numerical results and the polynomial fitting for j obtained using the previous equation. The values of the components of the matrix K are reported in Table 2. Fig. j. Numerical results and polynomial fitting for The correction function j (Eq. Observing Eq. 0 6 a 6 0:9) is less than 2%. (16)). Considering also Eqs. 0 6 q 6 20. in other words. (8) and (9) it is clear that the experimental tests should be performed with the same n or. (16) it is concluded that the correction factor j depends on n. Fig. For an infinitely long specimen (n ¼ 0). The matrix U reads: 2 1:7482487564 0:053754159533 6 0:76896688866 0:0068632911438 6 6 0:23922363475 U¼6 6 0:85633404777 6 0:25334178248 4 0:67470597429 0:18495379886 0:084067007027 0:0040142704949 0:0029984681658 0:023289123198 0:022297779266 0:0068989066533 9:8480085881e  05 3 7 0:00010108691939 7 7 0:00062358861997 7 7 7 0:00056784694513 5 0:00016783852495 ð18Þ . Catalanotti et al. and k. / Engineering Fracture Mechanics 118 (2014) 49–65 Fig. scaling the free length. 3. The average error in the calibration range (0 < n  1. j can be obtained fitting the numerical results using a polynomial function: rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pa XXX tan K ijk ai1 qj1 nk1 2 i j k j¼ ð16Þ where K ijk is the element of the matrix K of indexes i. the correction factor j reads: rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi / ¼ jjn¼0 ¼ tan pa XX 2 Uij ai1 qj1 i ð17Þ j where Uij is the element of the matrix U of indexes i and j.

55 G.12893905581246 0.62916671898641e-05 0. kÞ 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 2 2 3 3 3 3 3 3 1 1 1 2 2 2 3 3 3 4 4 4 1 1 1 2 2 2 3 3 3 4 4 4 1 1 1 2 2 2 1 2 3 1 2 3 1 2 3 1 2 3 1 2 3 1 2 3 1 2 3 1 2 3 1 2 3 1 2 3 1.00417757542138585 0.0205786990826619 9.0526303311437509 0. (17) can be used to approximate the correction factor.157514192137989 0. Eq. In the range 0:2 < n < 0:5 the error is higher but still less than 5% and therefore acceptable. 0 6 a 6 0:9) is again less than 2%.0111920357855815 0.000726535951382543 0.0278810474891446 0. As expected. Numerical results and polynomial fitting for / (Eq.815416824739527 0. when q increases (Fig.122154495971872 0.00293052383668594 0.203663008343557 0.101727935542504 0. the error  increases increasing the shape parameter n. If the specimen is long enough. 6 reports the cases with . which is defined in Eq. In the range 0 < n < 0:2 the error is exactly the same of that for n ¼ 0.111538580444948 1.2 0.231728483007845 0.00393368572531936 0.302437427062842 0.000182071222365654 0. 5 shows the average error. Eq. kÞ i j k K ði. For quasi-isotropic laminates (q ¼ 1) Eqs. 4 shows the numerical point and the polynomial fitting surface.177099439797605 0.000322325843661191 0.8 1 Fig. Fig.530314776856868 3 3 3 3 3 3 4 4 4 4 4 4 4 4 4 4 4 4 5 5 5 5 5 5 5 5 5 5 5 5 3 3 3 4 4 4 1 1 1 2 2 2 3 3 3 4 4 4 1 1 1 2 2 2 3 3 3 4 4 4 1 2 3 1 2 3 1 2 3 1 2 3 1 2 3 1 2 3 1 2 3 1 2 3 1 2 3 1 2 3 0.297280482414042 0. The corresponding error is: /  jFEM j  ¼ ð19Þ FEM Fig. Catalanotti et al.00658541595211619 0.0234077533074353 0.00425506303131702 0.000945832449681189 0.196826534461867 0.000867546012453078 9.000230184088065499 0. Fig. (11) and (17) yield the same results.78955615222354e-05 3.207433832357023 0.779004510684639 0.78897127917155 0.767208701652775 0.00561050809815955 4 3 2 1 0 20 15 10 5 0 0 0.0356324675667488 0. (17)).366847699226865 0. (16) and (17) provide an important improvement on the estimation of j when compared with Eq. In the range n > 0:5 the error is unacceptably high and Eq. as a function of n. 4.00446242346729918 0. (16) should be used.984700920323280 1.0163543233856253 0.48057443564002e-06 2. j.000834205920090498 0. (19). (11). Eqs.46770359644942 0.140766754998207 0. However.385474036008317 0.000263571025795895 0.592470330296433 0.6 0.00714960132308337 0.484668819900970 0.4 0. i j k K ði.108348678423743 0.110059901240345 0.0133183413168479 0.00822678007628292 0.572777961274980 1.32697528948074e-06 0. / Engineering Fracture Mechanics 118 (2014) 49–65 Table 2 Elements of the K matrix. (17) can still be used to estimate the correction factor j. j.0338305033576344 0. The average error in the calibration range (0 6 q 6 20.00433112725257584 0.186022036976596 0. 6 shows the correction factor j as a function of a for different values of q for an infinitely long specimen.74874169287311 0.

Catalanotti et al.56 G. As expected. q ¼ 5 and q ¼ 10). Fig. Sometimes it is not possible to scale the free length of the specimens or to respect the condition n < 0:5.8 0. nÞ ð20Þ where / is the correction factor for an infinitely long specimen (obtained using Eq.6 0. Correction factor 0. (8) and (9). when testing specimens whose dimensions lie between a large range of sizes (i. Defining a corrected value of the ultimate stress (or ultimate load) as: r^ u ¼ ru wðq. / Engineering Fracture Mechanics 118 (2014) 49–65 15 10 5 0 0 0. however. First of all it is convenient to factorize j as follows: jða.5 0 0 0.5 1 0. when n ¼ 0 (infinetely long specimens). 3 2. (20) is less accurate than Eq. a for different values of q. the difference between the predictions becomes very high and the use of Eqs.4 0. nÞ ¼ /ða.9 j vs. Average error as a function of n.e.1 0. its use is still acceptable. (17)) and w reads: w¼1 XX Wij qi1 nj i ð21Þ j Wij is the element of the matrix W of indexes i and j: 0:018136304459 0:17640537683 0:35130642410 0:16996415435 W¼ 0:0051147015355 0:038096285848 0:033267185410 0:00918499914 ð22Þ Eq.8 1 Fig.6 0. nÞ ð23Þ .5 2 1. qÞ wðq. when a coupon and a structural part are tested). 5.7 0.2 0. (16) and (17) is recommended. 6. For example. the correction factor w is equal to 1. In this case is still possible to obtain the R-curve of the material by an appropriate manipulation of Eqs.4 0. 7 shows the correction factor w as a function of q and n.3 Fig. q.5 0. because the average error between the fitted function and the numerical points is less than 5%.2 0. (16).

31 0.8 42.03 0. The cross-ply laminates also enable the calculation of the R-curve in the longitudinal direction of the ply. if the orthotropic rescaling technique applies. the correction factor for the orthotropic body.8 1 Fig. The elastic parameter of the laminates as well as the references to the corresponding previous investigations.4 0. fÞ. (16)–(25) are still valid if n is replaced by f. The nominal ply thickness is 0. The nominal ply thickness is 0. can still be obtained choosing the best regression method of those reported in Table 1.1 1 0.125 mm. but not-symmetric ply by ply. nÞ was formally derived for the case of a cross ply laminate (that is a solid with cubic symmetry being k ¼ 1). Table 3 Laminates investigated.2 0. (25) can be solved obtaining w ¼ wðDaÞ that substituted in Eq. Laminates CP-IM7/8552 CP-T800/M21 QI-T700/AR-2527 CP-T700/ACE Laminate’s elastic parameters Ref. For this reason. Catalanotti et al.04 [18] [6] [19] [20] ^u ¼ r ^ u ðwÞ. Experiments The proposed methodology is applied to measure the R-curve of the following laminates:  CP-IM7/8552: ½90=08s 16-ply cross-ply laminate with a nominal laminate thickness of 2 mm.57 G. Eqs. 2 The orthotropic rescaling technique is based on the rescaling of the x1 -axis by n ¼ k1=4 x1 and it allows to draw a parallel between an orthotropic body (k – 1) and a solid with cubic symmetry (k ¼ 1). RðDaÞ.  CP-T700/ACE: ½0=905 TORAYCAÒ woven fabric laminate with a nominal thickness of 2 mm and a nominal fabric thickness of 0.3 mm. q.7 0 5 10 15 20 0 0. (24) yields the R-curve.6 0. Eq. takes the same value of Eq.6 5. Eqs. invoking the orthotropic rescaling technique2 [12].3 32. / Engineering Fracture Mechanics 118 (2014) 49–65 1.8 0.08 mm. The size effect law. (16) substituting n by f. The nominal ply thickness is 0. It should be noted that in the case of multidirectional highly orthotropic laminates (multidirectional laminates with q > 4) a solution for the correction factor jða. Ex (GPa) Gxy (GPa) mxy (–) 90.125 mm.  CP-T800/M21: ½90=08s 16-ply cross-ply laminate with a nominal laminate thickness of 2 mm.0 16. It should be remembered that n and f are both dimensionless parameters and that in the case of a cross ply laminate n ¼ f. q.2 mm. w as a function of q and n.9 0. jða. This lay-up is symmetric by C-PLY™ ½0=  45 non-crimp fabric (NCF) bi-angle layer. r (8) and (9) can be rewritten as: 1 ^ 2u /2 wr  E  @  ^ 2u /2 ¼ 0 wr @w RðDaÞ ¼ ð24Þ ð25Þ As previously explained.6 55.03 0.7 90.3 5. fÞ is easily obtained. q. 3.  QI-T700/AR-2527: ½ð0=  45Þ=ð45=0Þ=ð90=45Þ=ð45=90Þs 16-ply quasi-isotropic laminate with a nominal laminate thickness of 1. In fact. . even if jða. are shown in Table 3. 7.2 0.

CP-IM7/8552 CP-T800/M21 QI-T700/AR-2527 CP-T700/ACE Avg. Table 4 Specimen geometry. A B C D E F – – 484 20 356 22 339 12 309 9 426 22 351 28 325 56 289 16 – – 316 24 323 11 269 11 380 35 291 24 299 24 256 10 – – 268 21 277 16 – – 299 19 253 14 271 6 .0 4. SD Avg. Specimen label 2 w (mm) a0 (mm) A B C D E F 10 15 20 25 30 35 3.58 G. SD Avg. SD X X F X X X X X ru (in MPa).5 9. Technical draw of the specimen. SD Avg.5 6. 8. CP-IM7/8552 CP-T800/M21 QI-T700/AR-2527 CP-T700/ACE A B C D E X X X X X X X X X X X X X Table 6 Average and standard deviation of the ultimate remote stress. / Engineering Fracture Mechanics 118 (2014) 49–65 Fig.5 Table 5 Test matrix. Catalanotti et al.0 10.0 7.

2w. the specimens were machined in a CNC machine equipped with a 1 mm drill bit.G. a0 . and the width. in particular concerning the crack length. The test matrix is shown in Table 5. To keep the tolerance under control. The use of a 1 mm drill bit does not lead to sharp crack tips. 2l. Table 4 shows the size investigated for each laminate and the label corresponding to each size. was controlled during the experimental tests to ensure that the condition n  0:2 was always respected. Catalanotti et al. therefore at failure the crack tip is sharp. and (ii) it has been shown that the fracture toughness measured in a centre cracked specimen with a crack machined using a drill bit and in a specimen with a crack manufactured using a thin saw is virtually the same [21]. Fig. 9. The free length. / Engineering Fracture Mechanics 118 (2014) 49–65 59 The geometry of the specimens is reported in Fig. Specimen after testing. however this is not problematic because: (i) the specimen fails when the crack has already propagated until its critical length. 8. .

Size effect laws: experiments and fitting for the investigated material. 10. / Engineering Fracture Mechanics 118 (2014) 49–65 Fig. Catalanotti et al.60 G. .

) . Fig. respectively. 9).G. 11. (For interpretation of the references to colour in this figure legend. A = 0. This is equivalent to obtain the R-curve as envelope of the crack driving force curves. the reader is referred to the web version of this article.) 400 350 300 250 200 150 100 50 0 0 0. the reader is referred to the web version of this article. A Knowing the size effect law. the R-curve is calculated solving Eqs. The average and the standard deviation of the remote stresses are shown in Table 6. Catalanotti et al. 11 and 12 show the fracture toughness for the ply in the longitudinal direction of the IM7/8552 and T800/M21 laminates. R-curve of the 0° ply for T800/M21 (in black) and the corresponding driving force curves (in blue). 10 shows the experimental data and the fitting obtained for all the laminates. N ¼ 3:719 mm for CP-T800/M21.384E5 MPa2 for QI-T700/AR-2527. The pictures of the failed specimens are shown in (Fig. For the IM7/8552 0° ply.5 2 2.5 3 Fig. pffiffiffiffiffiffiffiffiffi M ¼ 1440 MPa mm. C = 0. As explained before these curves are obtained multiplying by two the fracture toughness of the laminate. The size effect law (see Table 1) that best fit the data are the bilogarithmic (for CP-IM7/8552 and CP-T800/M21 laminates) and the linear regression I (QI-T700/AR-2527). N ¼ 8:614 mm for CP-IM7/8552. An Instron 4208 testing machine equipped with a 100kN load cell was used to perform all the tensile tests. The parameters that give the best fitting are:     pffiffiffiffiffiffiffiffiffi M ¼ 1242 MPa mm. (8) and (9). the experimental points obtained from the compact tension (CT) specimens are also reported using the FEM based data reduction method proposed in [9].656E6 MPa2 mm1. 12.5 1 1. / Engineering Fracture Mechanics 118 (2014) 49–65 61 Three specimens were tested for each size at a speed of 1 mm/min. R-curve of the 0° ply for IM7/8552 (in black) obtained as envelope of the driving force curves (in blue) and comparison with experimental results obtained using CT specimens (every marker a different specimen). C  ¼ 3:7E  007 MPa2 mm1 for CP-T700/ACE. Fig. (For interpretation of the references to colour in this figure legend. 300 250 200 150 100 50 0 0 1 2 3 4 5 6 7 Fig.  ¼ 0:672E  5 MPa2 .

R-curve of the CP-T700/ACE quasi-isotropic laminate (in black) and the corresponding driving force curves (in blue). Catalanotti et al. and on the length of fracture process zone. corresponding to an improvement over the classical CT test method. R-curve of the QI-T700/AR-2527 quasi-isotropic laminate (in black) and the corresponding driving force curves (in blue). The values of the length of the fracture process zone. are calculated and shown in Table 7. For the T700/AR-2527 ply these values are calculated using the analytical model previously developed [21] to predict the fracture toughness of the 0° ply from that of a multidirectional laminate. Using double-cantilever beam test specimens. the reader is referred to the web version of this article.650 3.527 4.4226 3.758 0.319 0. 14. [22] demonstrated that the rising part of the R-curve and the corresponding length of the fracture process zone depend on the specimen size. 13.62 G.63 1. some differences are observed on the rising part of the R-curve. it should be emphasised that the method proposed here provides a robust way to identify the steady-state value of the fracture toughness. (For interpretation of the references to colour in this figure legend. lfpz . / Engineering Fracture Mechanics 118 (2014) 49–65 120 100 80 60 40 20 0 0 0. the reader is referred to the web version of this article. IM7/8552 T800/M21 T700/AR-2527 lfpz (mm) R0ss (kJ/m2) c (mm1) b (–) 2. lfpz . 13 shows the R-curve of the QI-T700AR-2527 laminate. However. (For interpretation of the references to colour in this figure legend.92 205 283 254 0. Fig.004 350 300 250 200 150 100 50 0 0 1 2 3 4 5 6 7 8 Fig. . However.5 1 1. and of the steady state value of the fracture toughness of the 0° ply.14 1.) The steady-state value of the R-curve obtained using the double-edge notched specimens is in good agreement with that obtained using CT specimens.) Table 7 Parameter of the R-curve in the longitudinal direction of the ply.5 3 Fig. up to a critical size after which constant R-curves are obtained.5 2 2. R0ss . This indicates that further studies based on CT test specimens with different sizes are required to clarify the differences observer in the rising part of the R-curve. Foote et al.

as shown in Figs. 16. Neglecting the energy dissipated by the additional failure mechanism overestimates the fracture toughness of the ply. Catalanotti et al. Typical fracture surface observed in all the laminate tested (specimen type B). R-curve and fitting of the T700/AR-2527 quasi-isotropic laminate. 9(d) and 15(d).G. 14). and in this case only the R-curve of the laminate was calculated (see Fig. In fact. 15. Unfortunately this model cannot be used to determine the fracture toughness of the 0° ply for the CP-T700/ACE material system. . the fracture surface of this materials includes several pulled-out boundles of fibres and a damage region that is not confined to the uncracked ligament of the specimen. / Engineering Fracture Mechanics 118 (2014) 49–65 63 Fig. Fig.

respectively. if compared with the CT specimen. The authors gratefully acknowledge the support provided by AIRBUS. The parameters that best fit the R-curves of the ply in the longitudinal direction are shown in Table 7. Acknowledgements This work was funded by AIRBUS under the Project 2genComp – Second generation composites. 17. and T700/AR-2527 showed a crack resistance curve. 1. . 283 kJ/m2 for T800/M21. It is concluded that this is the preferred method to calculate the stress intensity factor for general lay-ups and geometries. The stress intensity factor used in the model can be easily obtained using a polynomial approximation of the results of the application of the Virtual Crack Closure technique in parametric Finite Element models of specimens with two edge cracks loaded in tension. Peter Linde for the useful discussions. for which the determination of the steady-state value may be ineffective. and Dr. T800/M21. The respective values of the fracture process zone are 2. R0ss .63 mm.14 mm and 1. 17. and 254 kJ/m2 for the T700/AR-2527 laminates.64 G. however. IM7/8552. The steady-state values of the longitudinal fracture toughness are 205 kJ/m2 for IM7/8552. using the method proposed here it was possible to obtain the R-curve of T800/M21. whereas the lowest value is observed for IM7/8552. both for the multiaxial laminate tested and for the 0° ply. R-curves of the ply in the longitudinal direction for the materials investigated. It is observed that the T800/M21 material has the highest value of the steady-state fracture toughness. / Engineering Fracture Mechanics 118 (2014) 49–65 Fig. The methodology proposed here circumvents both the need to perform complex post-processing analysis based on Finite Elements and the need to measure the crack length during the test. The methodology proposed here provides a robust way to measure the steady-state value of the R-curve for fibre-reinforced composites. As an example. A formula that has shown a good fitting of the R-curve is: h i R ¼ Rss 1  ð1  c DaÞb ð26Þ where c and b are the parameters that best fit the formula to the R-curve. 4. To simplify the use of R-curve in numerical and analytical models it is useful to express it analytically.92 mm. Conclusions Using the size effect law measured in composite laminates with two edge cracks it is possible to obtain the crack resistance curve. Fig. All the three carbon-epoxy material systems investigated here. 16 shows the R-curve for QIT700/AR-2527 laminate and the corresponding fitting. Catalanotti et al. It should be noted that the previous attempts to measure the R-curve of T800/M21 using compact tension test specimens were not successful due to specimen buckling [10]. The first and the second authors would like to acknowledge the support of the Portuguese Foundation for Science and Technology under the grants FCT-DFRH-SFRH-BPD-78104-2011 and FCT-DFRH-SFRH-BD-88593-2012. A comparison of the R-curves of the 0° ply in the longitudinal direction for the different materials is shown in Fig. The information generated in this paper will be used in the definition of the constitutive relations of the analysis models that aim to predict the mechanisms of crack initiation and propagation of composite structures.

A continuum damage model for fiber reinforced laminates based on ply failure mechanisms. ICASE. Paris PC.29(9):1105–16. [21] Camanho PP. [13] Bao G. April 2002. [17] Krueger R. [5] Camanho PP. Linde P. Catalanotti G. Int J Solids Struct 1992. [18] Camanho PP.158(2):211–23. Compos Struct 2006.8–3 Documentation. Irwin GR. 2013. Robinson P.76(1– 2):162–73. [6] Erçin GH. Learning python. Stress gradient effects in laminated composites. Pinto R. approach and applications. Wang TC. [8] Catalanotti G. Crack growth resistance curves in strain-softening materials. Leguillon D. On the validation of a damage mesomodel for laminated composites by means of open-hole tensile tests on quasiisotropic laminates.42(10):1515–1524. . Size effects on the tensile and compressive failure of notched composite laminates. Rose CA. Bao G. Compos Struct 2013. Compos Sci Technol 2006.67(13):2715–27.. Erçin GH. Xavier J. Daghia F. Prediction of size effects in notched laminates using continuum damage mechanics.28:235–48. Mahdi S. 2002-10. Iannucci L. Mai Y-W. Camanho PP. Del Research Corporation. Compos Sci Technol 2007. Eng Fract Mech 2011. [19] Arteiro A. Mech Mater 2013. Linde P. Planas J. Camanho PP. [20] Catalanotti G. Hampton. [11] Bazˇant ZP. 2013. Ho S. 2008. The role of material orthotropy in fracture specimens for composites. The virtual crack closure technique: history. Technical Report NASA/CR-2002-211628 ICASE Report No. Catalanotti G. Int J Solids Struct 2012.. [9] Pinho ST. Fracture and size effect in concrete and other quasibrittle materials. A procedure for superposing linear cohesive laws to represent multiple damage mechanisms in the fracture of composites. Fan B. Int J Fract 2009. A finite fracture mechanics model for the prediction of the open-hole strength of composite laminates. 1973. [7] Martin E. [2] Schuecker C. University of Porto. Notched response of non-crimp fabric thin-ply laminates.59:36–49. Stress analysis of cracks handbook. Maimí P. Inc.. [4] Dávila CG. Carrère N. Pettermann HE.76(0):69–76. CRC Press LLC.66:2069–79. Cotterell B. O’Reilly Media. Bessa MA. Catalanotti G. Fan B. Camanho PP. Technical report D-AG-20120414-T-C-4PB. Catalanotti G. On the relation between the mode I fracture toughness of a composite laminate and that of a 0° ply: analytical model and experimental validation. Compos Sci Technol 2013. Camanho PP.79(0):97–114. Mahdi S. Xavier J. 1997. [22] Foote RML. Suo Z. [3] Camanho PP.G. [14] Tada H. Catalanotti G. 2008. Vogler M. A coupled strength and toughness criterion for the prediction of the open hole tensile strength of a composite plate. Rolfes R. / Engineering Fracture Mechanics 118 (2014) 49–65 65 References [1] Abisset E. Compos Part A Appl Sci Manuf 2012.43(8):1219–25. [15] Dassault Systèmes. University of Porto.96(0):736–744. Catalanotti et al. Orthotropy rescaling and implications for fracture in composites. Abaqus Version 6. A semi-analytical method to predict net-tension failure of mechanically fastened joints in composite laminates. [10] Erçin GH. [12] Suo Z. Dávila CG.49(26):3915–22. Virginia. PhD thesis. Camanho PP. Compos Sci Technol 2013. Compos Part A Appl Sci Manuf 2011. Fracture toughness of the tensile and compressive fibre failure modes in laminated composites. Ladevèze P. Modeling the inelastic deformation and fracture of polymer composites – part II: smeared crack model.34:593–607. 3rd ed.78(13):2535–47. Int J Solids Struct 1990. J Mech Phys Solids 1986. [16] Lutz M.