DAMAGE INITIATION AND DEVELOPMENT IN TEXTILE COMPOSITES: A GALLERY S.V. LOMOV, D.S. IVANOV, V. KOISSIN, J. KUSTERMANS, K.VALLONS, J.XU, I. VERPOEST, Departement of Metallurgy and Materials Engineering Katholieke Universiteit, Leuven Kasteelpark Arenberg 44, B-3001, Leuven, Belgium Stepan.Lomov@mtm.kuleuven.be, tel +32-16-32-12-10, fax +32-16-32-19-90

V. CARVELLI, V.N. TOMASELLI, Department of Structural Engineering, Politecnico di Milano, Milan, Italy

B. VAN DEN BROUCKE, EADS Innovation Works, Munich, Germany

V. WITZEL, IFB - Institut für Flugzeugbau, Universität Stuttgart, Germany

SUMMARY The paper presents results of experimental observations of damage processes during tensile quasi-static and fatigue loading for carbon/epoxy textile composites: (1) woven fabrics; (2) non-crimp fabrics (NCF); (3) structurally stitched NCF. The test methodology includes, apart from the “normal” tensile testing, registration of acoustic emission and full-field measurements of strains on the surface of the sample. The damage in the samples loaded to a certain characteristic level of strain or certain number of fatigue cycles, is studied using X-ray inspection and micrography of the cross-sections. Analogies between damage development in quasi-static tensile test and tension-tension fatigue are analyzed and links between the damage initiation thresholds in quasi-static tests and fatigue life are established. Meso-level finite element modelling is performed for quasi-static tension of NCF and woven fabric composites, using damage mechanic approach to model the progressive damage.

1 INTRODUCTION The paper continues the topic, started by presentation on experimental methodology of studying damage in textile composites (SEICO-2006), later published in [1]. Since then, progressive damage in different textile composites has been studied, accompanied with finite element (FE) modelling on meso-scale level [2]. The paper reports the results of these studies of quasi-static and fatigue tensile behavior of various carbon/epoxy textile composites. The results for each group of materials are organised as follows: (1) material description; (2) results of static tensile tests: stress-strain diagrams, acoustic emission (AE), damage identification; (3) tension/tension fatigue tests; (4) FE modelling; Intended as a “gallery”, the paper presents the results in brief, graphical/tabular way. More detailed data and in-depth discussion will be given in future publications.

2 WOVEN COMPOSITE 2.1 Material Table 1 shows parameters of woven carbon/epoxy composite and geometrical model built with WiseTex software [3]. Table 1 Parameters of woven carbon/epoxy composite

Reinforcement Homogenised yarns, GPa Weave Twill 2/2 E11 141 Areal density, g/m2 289 E22 = E33 9.5 Ends/picks, 1/cm 3.52 G12 = G13 3.69 Yarn linear density, tex 400 G23 3.26 Fibre count 6K ν12 0.27 Fibre type (Tenax) HTA 5131 ν23 0.46 VF inside yarns, % 59 Strength of HY, MPa Composite VF, % 32 Tens. L 2200 Matrix Tens. T 35 Young module, GPa 2.73 Comp. L 2060 Poisson coefficient 0.4 Comp. T 146 Tensile strength, MPa 76 Shear 73

WiseTex model of the fabric

2.2 Tension Figure 1 shows stress-strain diagram and energy of AE events for tensile test of woven carbon/epoxy composite in warp direction. Experimental methodology [1] was followed; digital image correlation system was used as optical extensometer. Damage starts at applied strain of about 0.15%, and very fast develops in a system of transverse cracks in weft, which multiply till the very end of the test. However, stiffness of the material is not affected by the lost of the load-.carrying capability of weft yarns. This may be explained by the intrinsic increase of stiffness of carbon fibres which compensates the loss of about 5 GPa of the stiffness due to damaged transverse yarns.

0

100

200

300

400

500

0.0 0.2 0.4 0.6 0.8 1.0

strain, %

stre

ss, M

Pa

1.E+001.E+011.E+021.E+031.E+041.E+051.E+061.E+071.E+081.E+09

AE

ener

gy [-

]

stress-strainAE

a

0

200

400

0.0 0.5 1.0strain, %

stre

ss, M

Pa

experimentf ine meshrough meshrough mesh stif feningfine mesh stif fening

b Figure 1 Tension of woven carbon/epoxy composite: (a) stress-strain diagram and AE energy; (b) comparison of the stress-strain diagram and ABAQUS FE modelling

2.3 Finite Element Simulations FE model of the unit cell of composite was build using the WiseTex geometrical model. The model was corrected to exclude interpenetrations of the yarn volumes, meshed and direction-dependent material properties assigned, using MeshTex software [2]. The mesh and the material properties and coordinate systems was

transferred to ABAQUS. UMAT subroutine, realising stiffness degradation damage algorithm with Tsai-Wu damage initiation criterion was employed to calculate progressive damage in the yarns and matrix, as described in [2]. Figure 1b shows the calculated stress-strain diagrams in comparison with the experimental data. Two meshes with different size of the elements were used: fine mesh with 26639 elements (Figure 2), and coarse mesh with 5375 elements. Comparison of the diagrams shows that influence of the element size on the resulting diagram is not very strong. Influence of the element size on the numerical damage initiation threshold is more pronounced: it changes from 0.29% to 0.35% with transition from coarse to fine mesh. These values are not far from the experimentally observed range of damage initiation threshold strain of 0.15…0.25%. The stress-strain diagrams, calculated with constant longitudinal stiffness of the yarns show decrease of stiffness of composite with progression of damage. This contradicts the observed constant stiffness of the composite up to final failure of the sample. When the intrinsic increase of stiffness of carbon fibres [5] is implemented in the model (almost linear increase of the stiffness by 20% up to strain of the fibres of 1%), the calculated diagram is close to the measured (Figure 1b). Figure 2 illustrates damage modelling more in detail, showing progress of transverse damage from initiation to the pronounced pattern.

a b c

Figure 2 FE modelling of tension (horisontal) of woven carbon-epoxy composite: (a) transverse stress in yarns, applied strain 0.29%; (b,c) damaged elements (transverse damage), applied strain 0.29% and 0.52%. Damaged elements: red.

3 NON-CRIMP FABRIC COMPOSITES 3.1 Material Table 2 shows parameters of two types of non-crimp fabric (NCF) carbon/epoxy composites (Figure 3). Table 2 NCF reinforcement and carbon/epoxy composite samples

A B A B Fibre orientation +45°/-45° Number of the plies 4 4 Carbon tows 24K 12K Plate thickness, mm 2.1 3.0 Areal density, g/m2 540 322 VF, % 56 25 Stitching spacing, mm 5.0x3.5 4.9x1.7 Stitching pattern Chain Tricot

3.2 Tension Figure 4 shows stress-strain diagram and typical AE registration for tensile loading on the NCF carbon/epoxy in direction along the fibres. One notes a distinct damage initiation threshold of 0.3% strain (stress 200 MPa). The damage is transversal matrix cracking (Figure 5); the density of these cracks is saturated at about 1% of the applied strain. The damage does not affect sample stiffness till the very late stage of deformation before the final failure.

1000

Figure 3 NCF: ply distortions by the stitching

Figure 4 NCF type A: Tension diagram and AE, loading in bias direction. AE sensors removed at strain 1%. Circles: AE events, dashed line: cumulative AE energy →

0.5% 0.9% 1.3%

Figure 5 X-ray images of cracks in NCF composites type A after loading in bias direction (vertical) up to three different strain levels.

3.3 Tension-tension Fatigue Tensile-tensile fatigue testing of NCF composite (R=0.1) reveals interesting features of fatigue behaviour, which are depicted in Figure 6.

− Fatigue life limit of about 600 MPa (no sample failure up to 10 mln cycles) is considerably higher then the damage initiation level (DIL) of about 200 MPa determined in static tests. Fatigue loading at DIL and somehow above DIL leads to transverse cracking, which do not affect the load carrying ability of the sample.

− The fatigue crack pattern shows full similarity with the crack pattern in static loading (compare Figure 6c with Figure 5). The crack system is saturated quite late, at about 2 mln fatigue cycles. 3.4 Finite Element Simulations Finite element simulations on meso-level were performed for NCF composites type B, loaded at 45° direction relative to the fibres [4] – Figure 7. The model accounts for change of fibre directions near the resin-rich zones, created by stitching. Puck criterion was used for damage assessment together with the stiffness degradation scheme described in [2] and applied to the whole damaged ply. Strength of the UD fibre ply in transverse and shear direction was measured to be 28 MPa and 74 MPa correspondingly. Experimentally measured DIL is 1.3% strain. Figure 7 shows that FE modelling predicts DIL close to the experimentally observed and well predicts stiffness degradation of the composite for off-fibre direction (matrix dominated) loading.

0

200

400

600

800

0.0 0.5 1.0 1.5 2.0stress- strain, %

stress, MPa

rain

1.E+05

1.E+06AE energy [-]

1.E+04

1.E+03

1.E+02

1.E+01

1.E+00st

AE

a

5⋅103

N b

5⋅10

curve; (b) sat in static test; (c) transversal cracks, X-ray registration (figures indicate number of cycles)

5

c

Figure 6 Tension-tension fatigue loading (R=0.1) of NCF composites type A in fibre direction: (a) S-N uration of crack density, load level corresponds to the damage initiation limit

a

b

c

d

factor (SEF) of Puck, applied strain 1.5%

Figure 7 NCF composite type B: FE modelling: (a) mesh in the plies; (b) change of the fibre orientations near a resin rich zone; (c) degradation of Young module of composite(d) stress intensity

4 STRUCTURALLY STITCHED COMPOSITE 4.1 Material The layers of structurally stitched (tufted) carbon/epoxy composite (Figure 8) are two different NCF fabrics (0°/90° and 45°/-45°) with areal density 556 and 540 g/m2. The laminate is tufted with 1K carbon yarn 67 tex, twist 15 1/m with the tufting step 5x5 mm. The composite is impregnated by RTM6 resin. Plate thickness is 3.2 mm, fibre volume fraction is 55%. The tufted composites are compared with non-tufted.

Figure 8 Structurally tufted NCF composite: face and back side and lay-up scheme

4.2 Tension Tension test (with strain mapping and AE registration, performed in 0° and 90° direction) has shown (Table 3) that, as expected, the elastic properties of the composite are not affected by tufting. Strength and ultimate elongation are increased by about 70 MPa and 200 µε, which is a trend opposite to normally observed in literature. However, the damage initiation thresholds (determined using AE) are lowered as result of the stitching. For the stitched material damage initiates at the stitching sites, which are focuses of stress concentrations (Figure 9). The damage is initiated in the sites on non-structural stitching of the NCF layers in non-tufted composites; in the tufted material important damage clusters are observed at the tufting sites. Comparison of AE cumulative energy curves gives insight into combination of lower DIL and higher strength of the tufted materials: whilst damage, indeed, is initiated earlier, high AE energy damage events are somehow hindered by the stitching (Figure 10). Table 3 Change of mechanical properties of tufted NCF composites in comparison with not tufted

Young module = Strength + Damage initiation limit – Poisson coefficient = Elongation + Intensive damage threshold –

a b cFigure 9 Strain map on the surface of tufted composite (a); X-ray image of damage patterns in non-stitched (b) and tufted (c) composites, strain 0.3%

a

b

Figure 10 AE diagrams for (a) unstitched and (b) stitched NCF composites, loading in 0° direction

0

200

400

600

1.E+00 1.E+01 1.E+02 1.E+03 1.E+04 1.E+05 1.E+06 1.E+07

Nstre

ss, M

Pa

tuftednon-stitchedPower (non-stitched)Power (tufted)

Figure 11 S-N curve for tension-tension fatigure (R = 0.1) of non-stitched nad tufted NCF composites, loading in 0° direction and X-ray image of the stitched specimen after 100,000 cycles, 240 MPa

4.3 Tension/Tension Fatigue DIL for non-stitched and stitched materials are about 150 and 100 MPa. S-N curve for tension-tension fatigue (Figure 11) show that the fatigue life limits for the both material types are well above the DIL (but the difference is not that large as in previously studied NCF composites, Figure 6). When comparing the tests in the same direction, it seems that tufted materials have lower fatigue life (failure 10 times earlier for the same stress level). More fatigue statistics will give more understanding for this matter.

5 CONCLUSIONS The methodology of studying damage processes in textile composites, presented in [1], has been applied to different materials in conjunction with fatigue testing and FE modelling. Clear relation is identified between the damage initiation limit in static tests and fatigue life limit, as well as analogy between the progressive damage patterns for static tests (increasing strain) and fatigue (loading cycles). Meso-FE modelling proved to be able to produce adequate description of damage processes in textile composites.

6 ACKNOWLEGEMENTS The work reported here was supported by ITOOL project, FP7 – EC, grant from FWO – Flanders (JX), PhD grant from K.U. Leuven (DI), Erasmus program (VT)

7 REFERENCES 1. S. V. Lomov, D. S. Ivanov, T. Truong Chi, I. Verpoest, F. Baudry, K. Vanden Bosche and H. Xie, Composites Science and Technology, 68 (2008) 2340. 2. S.V. Lomov, D.S. Ivanov, I. Verpoest, M. Zako, T. Kurashiki, H. Nakai, and S. Hirosawa, Composites Science and Technology, 67 (2006) 1870. 3. Verpoest, I. and S.V. Lomov Comp Sci Techn, 65 (2005), 2563 4. Truong Chi, T., D.S. Ivanov, D.V. Klimshin, S.V. Lomov, and I. Verpoest Composites part A, 39 (2008) 1380 5. Toyama, N. and J. Takatsubo Comp Sci Techn, 64 (2004), 2509