Pull-through performance of carbon fibre epoxy composites

Pull-through performance of carbon fibre epoxy composites

Composite Structures 94 (2012) 3037–3042 Contents lists available at SciVerse ScienceDirect Composite Structures journal homepage: www.elsevier.com/...

1MB Sizes 0 Downloads 189 Views

Composite Structures 94 (2012) 3037–3042

Contents lists available at SciVerse ScienceDirect

Composite Structures journal homepage: www.elsevier.com/locate/compstruct

Pull-through performance of carbon fibre epoxy composites Tomasz C´wik a,⇑, Lorenzo Iannucci a, Marc Effenberger b a b

Department of Aeronautics, Imperial College London, Roderic Hill Building, South Kensington Campus, London SW7 2AZ, UK Faserinstitut Bremen e.V. (FIBRE), Am Biologischen Garten 2, 28359 Bremen, Germany

a r t i c l e

i n f o

Article history: Available online 3 April 2012 Keywords: A. Carbon fibres A. Fabrics/textiles B. Delamination B. Mechanical properties Pull-through

a b s t r a c t An investigation of the through-thickness properties of carbon fibre prepreg laminates, Non-Crimp Fabric laminates and non-crimp 3D orthogonal woven composites by pull-through testing was performed. Influence of matrix system and curing temperature on the performance of the 3D woven composites was investigated. The results showed that the prepreg specimens endured the highest pull-though loads, followed by the 3D woven composites, and the NCFs. It was found however that the prepreg laminates were characterized by the largest extent of damage after the test, whereas only very localized damage was present in the 3D woven composites. The NCF specimens presented overall the worst bolt pull-thought performance. No significant influence of the matrix system and curing temperature on the 3D woven composites’ performance was observed. Ó 2012 Elsevier Ltd. All rights reserved.

1. Introduction It is well known that laminates have very low transverse properties. Poor through-thickness (TT) properties of laminates are reflected in low interlaminar fracture toughness of a composite. In such materials delaminations are prone to appear even under small impact loads or due to manufacturing imperfections [1]. Hence most applications have been designed such as not to expose the composite to out-of-plane loads. As a result, research in TT properties of composites has been limited. For many years there has been no unified standard for pull-through testing of composites. The ASTM Standard Test Method for Measuring the Fastener PullThrough Resistance of a Fiber-Reinforced Polymer Matrix Composite appeared in 2007 [2]. A number of investigations have been conducted in order to develop composite material with enhanced transverse properties. Upgrading matrix toughness by using either a thermoplastic matrix or toughened epoxies is described in Refs. [3–5]. Improvement in laminate transverse properties was noted when z-pins or stitches running in transverse direction were inserted [6]. These two solutions usually offer improvement in the Mode I and the Mode II toughness of a composite by a factor of 2–10. Non-Crimp Fabrics (NCFs) are also characterized by improved fracture toughness with respect to prepreg laminates. It is believed that further enhancement in fracture toughness, impact behaviour and in some cases ballistic performance of a material can be obtained by using 3D woven performs as reinforcement. This group of materials can be subdivided into 3D through-thickness interlock weave, 3D ⇑ Corresponding author. Tel.: +44 (0) 20 7594 5113; fax: +44 (0) 20 7594 5078. E-mail address: [email protected] (T. C´wik). 0263-8223/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.compstruct.2012.03.027

layer-to-layer interlock weave, and 3D orthogonal weave. The 3D woven orthogonal composites could be manufactured on modified traditional 2D weaving machines or by using 3WEAVEÒ technology, commercialized by 3Tex. Fabrics made on traditional weaving looms are prone to deterioration in their in-plane mechanical properties due to the nature of the manufacturing process and due to high crimp. The amount of damage and crimp is substantially reduced if a 3D woven fabric is manufactured by 3WEAVEÒ technology. Fabrics manufactured using this technology are commonly referred to as the ‘non-crimp 3D woven orthogonal’. More detailed information regarding 3D woven fabrics and applicable manufacturing technologies can be found in Refs. [7–10]. In the present study, pull-through properties of prepreg laminates, Non-Crimp Fabrics and non-crimp 3D woven orthogonal composites are compared. Additionally, the influence of the matrix system and curing temperature on the test results are investigated. 2. Experimental work 2.1. Materials and specimen preparation Three different materials were used for the production of the 2.5 mm and 5 mm thick specimens, namely: prepreg laminates, Non-Crimp Fabric laminates and unitary non-crimp 3D woven orthogonal material. The latter two were infused, via Resin Infusion under Flexible Tooling (RIFT) technique, with two aerospace class resins – RTM-6 and MVR444. Information regarding curing temperatures and averaged fibre volume fractions of specific materials can be found in Table 1. The NCF and the 3D woven specimens had a cross-ply lay-up, whereas the prepreg laminates had a quasi-

´ wik et al. / Composite Structures 94 (2012) 3037–3042 T. C

3038 Table 1 Experimental results.

Average Initial Failure Load (kN)

Specimen type

Nominal thickness (mm)

Fibre volume fraction, Vf

Average thickness in hole area (mm)

Prepreg NCF (MVR-444@180°) 3D woven (MVR-444@180°) 3D woven (MVR-444@90°)

2.5

0.65 0.55 0.57 0.59

2.50 2.16 2.79 2.73

5.182 2.827 4.888 4.282

Prepreg NCF (MVR-444@180°) 3D woven (MVR-444@180°) 3D woven (RTM6@180°) 3D woven (MVR-444@90°)

5

0.65 0.56 0.61 0.62 0.62

4.84 3.85 4.79 4.84 5.42

10.158 4.345 8.251 7.389 7.594

Specimen type

Nominal thickness (mm)

Prepreg NCF (MVR-444@180°) 3D woven (MVR-444@180°) 3D woven (MVR-444@90°)

2.5

Prepreg NCF (MVR-444@180°) 3D woven (MVR-444@180°) 3D woven (RTM6@180°) 3D woven (MVR-444@90°)

5

Final Failure Load, Pf (kN)

Average Load Drop (kN)

Coefficient of variation (%)

5.15 5.84 3.23 2.43

1.024 0.180 0.099 0.105

18.65 79.44 45.46 58.10

11.02 20.64 3.31 2.22 5.11

2.917 1.001 1.337 1.447 0.589

36.68 71.03 16.68 15.20 62.64

Coefficient of variation (%)

Slope before the Initial Failure Load

Slope after the Initial Failure Load

Change in stiffness (%)

Area under L vs. D curveenergy (J)

Damage size (mm2)

6.847 4.526 5.898 6.668

0.964 0.807 1.156 1.113

0.536 0.650 0.829 0.997

44.37 19.38 28.20 10.39

33.547 17.152 25.936 28.679

Whole 2989 1496 1696

13.744 8.017 13.332 13.075 14.684

5.237 3.308 4.475 4.473 4.789

1.490 1.133 1.085 1.484 1.161

71.50 65.76 75.76 66.82 75.76

81.467 48.981 68.412 70.261 88.526

Whole 5525 1978 2340 2209

isotropic (included plies at ±45°), balanced and symmetric lay-up. For the prepreg laminates, the Hexcel’s M21 material with 35% resin content, a T700GC fibre and an areal weight of 268gsm was used. The non-crimp 3D woven orthogonal fabric (3WEAVEÒ), developed by 3Tex, consisted of the following fibres: warp yarns – Toho Tenax-E HTS40 F13, 12K 800tex; fill yarns – Toho Tenax-E HTA40 E13, 26K 400tex; Z-yarns – Toho Tenax-J HTA40 H15, 1K 67tex. The material was provided in two thicknesses, approx. 2.5 mm and 5 mm. In the Non-Crimp Fabric specimens Toho Tenax-E HTS40 F13, 12K 800tex fibres were used. Stitches were made from PES 48 dtex SC. Single NCF ply had an approximate thickness of 0.5 mm. The plies were stacked to obtain the required 2.5 and 5 mm thick laminates. The NCF laminates had the following lay-up for 2.5 mm and 5 mm thick specimens [90°/0°/0°/90°]s and [90°/0°/0°/90°/90°/0°/0°/90°]s, respectively. The aforementioned thicknesses were nominal thicknesses of the specimens. All test specimens were prepared according to the Airbus guidelines [11]. The specimen configuration is presented in Fig. 1. 2.2. Quality assessment of cured panels/specimens Quality assessment was performed at each stage of the manufacturing process. Both cured panels and machined specimens

were C-scanned. Some of panels were of poor quality. The potential influence of manufacturing flaws is described in Section 3. The 3D woven panels infused with the RTM-6 or MVR444 cured at 180° when removed from the oven (after cooled down to around 80 °C), on cooling to room temperature, exhibited audible microcracking. No visual damage was noticed. It appeared to the authors that the noises were a result of cracking from within the panels. This did not occur in the case of the 3D woven panels infused with MVR444 and cured at 90°. Such behaviour was previously noticed by Zhang [12]. The author stated, based on a microscopy samples, that the microcracks appear in the resin rich areas near the fibre/resin interface and around the Z-yarns. Zhang sees the reason for that unusual behaviour in a mismatch of coefficient of thermal expansion (CTE) of fibres and matrix, concluding that the ‘‘microstructure of 3D orthogonal fabric provides a very different residual stress distribution from other reinforcement structures. The large resin rich area results in large heterogeneous residual stress distribution and large variation of the residual stress in the microstructure of the composite, due to the mismatch of the coefficient of thermal expansion of resin and fibres’’ [12]. Some of the samples were analysed using cross-polarized illumination. The analysis is used to determine whether any residual strains are present within the structure. If a sample has no residual strain then it will appear black under cross polarized illumination. If there are residual strains then some of the reflected light will pass through the polarisers and the region with strain will be lighter. The brighter the region under crosspolarized illumination the greater the residual strain. Fig. 2 shows the manufactured 3D woven sample with residual stresses, and confirms the statement. The microcracking might have had a negative influence on the specimen performance. 2.3. Pull-through testing

Fig. 1. Specimen dimensions.

The tests were performed on an INSTRON 4505 machine. All tests were conducted to Airbus guidelines [11]. The authors were asked not to disclose detailed information about the Airbus testing method. Therefore, only general indications about the difference between the Airbus guidelines and the aforementioned ASTM standard will be given herein. Firstly, the loading rigs differ substan-

´ wik et al. / Composite Structures 94 (2012) 3037–3042 T. C

3039

Fig. 2. The 3D woven sample under cross-polarized illumination analysis.

Fig. 4. Normalized load vs. displacement chart for different types of 2.5 mm thick materials.

Fig. 3. Pull-through testing rig based on the Airbus guidelines.

tially. Thus, specimen dimensions are different for each of the tests. Fig. 3 presents the Airbus rig used in this study. Secondly, the ASTM standard suggests standard crosshead displacement rate of 0.50 mm/min, whereas the Airbus guidelines specifies higher displacement rate. Also, the ASTM guidelines requires ‘‘at least five specimens per test condition unless valid results can be gained through the use of fewer specimens’’ [2]. The Airbus guideline requires six specimens: five are loaded to final failure, whereas the sixth one is loaded to a certain point. The ASTM standard, states clearly that the Initial Failure Load (IFL), defined as ‘‘the first significant (greater than 10%) drop in applied force’’ is the point of structural failure. The Airbus guidelines does not state the percentage value, however notes that the IFL can be defined by either significant drop in load, change of the load–displacement curve, or substantial noise indicating internal damage. Moreover, the Airbus guideline does not define the end point of the pull-though test, whereas the ASTM standard says ‘‘the specimen is loaded until a force maximum is reached and force has dropped off about 30% from the maximum’’. The specimen was fastened at right angles to the control plate, using an aircraft class bolt (EN6115:2007), with a preload of 6 Nm. Post-mortem B- and C-scans were performed for all tested specimens. 3. Results and discussion 3.1. The load vs. displacement response The tabularized results, presented in Table 1, show raw data from the tests, however loads presented in Figs. 4 and 5 are nor-

Fig. 5. Normalized load vs. displacement chart for different types of 5 mm thick materials.

malized with respect to the average specimen thickness. This step was made in order to identify potential influence of difference in thickness between the specimens on the final results. It is noted that almost all specimens failed in a similar manner within their material group. The coefficients of variation for the Average Initial Failure Loads (IFLs) were far below 10%, typically in the range of 2–5%. The exceptions are the 5 mm thick Prepreg and the NCF infused with

3040

´ wik et al. / Composite Structures 94 (2012) 3037–3042 T. C

MVR444 and cured at 180, which have coefficients of 11.02% and 20.64%, respectively. Pre-test C-scans (not shown here) indicated small non-uniformity in the prepreg panels, which might account for the variation in the results. The NCF MVR444@180 panel was characterized by high degree of waviness on the front surface, which resulted in varying in thicknesses through specimens. This may have influenced the panel performance during the test. The highest Initial Failure Load was achieved by the Prepreg specimens in both thickness categories. The 3D woven MVR444@180 specimens showed the second highest IFL values. The supremacy of the prepreg IFL is even more pronounced when the loads are normalized with respect to specimen thickness. The normalized values of the 5 mm thick 3D woven MVR444@180 and the 3D woven MVR444@90 are similar and lower than the prepreg value. The IFLs for the 2.5 mm thick 3D woven MVR444@90 specimens were determined on the basis of significant audible cracking heard during the tests, as no visible damage was noticed. Figs. 4 and 5 show that the difference between the prepreg IFL and the second best contender is approx. 0.5 kN/mm and 0.8 kN/mm, for 2.5 and 5 mm thick specimens respectively. The lowest IFL (and the normalized IFL) were yielded by the NCF MVR444@180 specimens in both thickness categories. The highest Final Failure Load (FFL) in the 2.5 mm thick category was achieved by the Prepreg, which was just slightly higher than the FFL of the 3D woven MVR444@90 specimens. The normalized values for the two materials show similar results. In the 5 mm thickness category, the 3D woven MVR444@90 achieved the highest FFL values, almost 1 kN higher than the prepreg specimens. However the normalized results show that all 3D woven specimens have similar FFL values, which indicates that the increased thickness of the 3D woven MVR444@90 might be responsible for the high FFL value. At this point it should be noted that the 5 mm thick 3D woven MVR444@90 panel was identified before the tests as the one of the lowest quality. Therefore, it can be presumed that if no defects were present, the specimens could possibly yield higher IFL and FFL values. The lowest FFL values were obtained by the NCF MVR444@180 specimens in both thickness categories, yet the normalized values show that the difference between the NCFs and the 3D woven MVR444@180 FFL is relatively small for the 2.5 mm thick specimens. However, in the 5 mm thickness category the normalized FFL of the NCFs is much lower than the FFLs obtained by the other materials. The ASTM standard [2], states clearly that the Initial Failure Load is the point of structural failure, therefore this value should be taken into strength analysis calculations, not the Final Failure Load. This statement emphasise the importance of the Average Load Drop (ALD). The ALD values are not as uniform as the IFL values – the coefficient of variation (CV) ranges from 15% to 80%. Nevertheless, the load vs. displacement charts (not shown here) show a fairly uniform trend for all tested specimens. Therefore, it appears that the values are probably valid, or at least provide an indication of the magnitude of the ALD for a given material. The highest ALD values were noted for the Prepreg specimens, 1.024 kN and 2.917 kN, whereas the lowest average ALDs were achieved by the 3D woven MVR444@180 specimens (0.099 kN) and 3D woven MVR444@90 (0.589 kN), for the 2.5 and 5 mm thicknesses respectively. The 3D woven NCF MVR-444@180 specimens had the second lowest value of the ALD (1.001 kN) in the 5 mm thickness category, however had also a very high CV of 71.03%. The magnitude of the ALD values is related to reduction of stiffness in specimens, shown in Table 1. For the 2.5 mm thick specimens the reduction varies from 10% to 45%, whereas for the 5 mm thick specimens all values oscillated around 70% of the drop in stiffness. Due to the fact that determination of these values is highly susceptible to human readout error, they should be treated with caution.

Figs. 4 and 5 show that the 2.5 mm thick specimens have higher displacement to Initial Failure Load curve slope than the 5 mm thick specimens, approximately one and a half times higher slope. However, after the Initial Load Drop, the displacement to the Final Failure Load is higher for the 5 mm thick samples than for the 2.5 mm thick ones. The displacement to the Final Failure Load ratio is roughly the same for both types of specimens. 3.2. Damage area A post-mortem visual inspection and C-scans (see Fig. 9) indicated that the greatest amount of damage (delamination) was present in the prepreg specimens. Slightly less damage was noted in the NCFs, and the smallest amount of damage was present in the 3D woven specimens. The size of damage area in the 3D woven specimens infused with the MVR444 resin and cured at 180 °C was similar to the size of the damage in the specimens cured at 90 °C. Regarding the exact damage size values, presented in Table 1, the values are only indicative, as the digital measurement technique was largely dependent on the user selection. The difference in colours1 in Fig. 9 between the prepregs and the rest of the specimens is a result of using different type of software for the prepreg specimens. Figs. 6–9 present specimens after the pull-through tests. Taking into consideration reinforcement architecture of each type of the specimens and the results, it appears that presence of stitches in the NCFs and z-direction yarns in the 3D woven specimens reduces the damage area. The smallest amount of damage was noted in the 3D woven specimens. In Fig. 9, the presented 2.5 mm thick 3D woven MVR444@90 specimen has the bolt-caused delamination in the middle and two thin damages at the sides. Looking at the position of the specimen in the rig (see Fig. 3) one can notice that these damages occurred at the edge of the clamping block. Taking into consideration the fact that the edge was filleted (r = 3 mm), it seems hard to justify the presence of the damage by the contact of the edge with the specimen. However, Fig. 4 shows that the 3D woven MVR444@90 specimen had higher final displacement value than the NCFs and the 3D woven MVR444@180 specimens. The figure also shows that the prepreg specimens had the highest final displacement, however the authors noticed during the tests that after the FFL the bolt head was still within the prepreg specimen being hold only by a couple of outer ±45° layers (contributing to the final displacement value) – see Fig. 6b. Whereas, in case of the 3D woven specimens, once the bolt completely sheared out the adjacent material (reached the FFL), it left the specimen, with much smaller contribution of the outer plies to the final displacement value. This is reflected in the smaller damage area of the outer surface of the specimen (see Fig. 7), compared to the prepreg specimen. Therefore, it appears that specimen-clamping blocks contact during large displacements might have been responsible for the two thin damages in the 3D woven MVR444@90 specimens. No damage at all to the bolts used for the tests was noticed. 3.3. Energy For all specimens the area under the load vs. displacement curve was calculated. The area expresses the work done during the test, which is equal to energy required to cause damage to the specimens. Table 1 shows the calculated values. It should be noted that the pull-through tests were stopped when the load dropped to around 200–300 N, therefore the values presented are slightly affected by this limit.

1 For interpretation of colour in Fig. 9, the reader is referred to the web version of this article.

´ wik et al. / Composite Structures 94 (2012) 3037–3042 T. C

3041

Fig. 6. The prepreg specimens after failure: (a) 5 mm thick; (b) 2.5 mm thick in the loading rig.

Fig. 8. The 5 mm thick NCF MVR444@180 specimen after failure.

3.4. Different resin systems

Fig. 7. The 2.5 mm thick 3D woven MVR444@90° specimen after failure.

It appears that the prepreg specimens dissipated the largest amount of energy, followed by the 3D woven MVR444@90, and 3D woven MVR444@180. The NCFs dissipated the smallest amount of energy. It is difficult to establish any relationship between the energy and damage size in specimen. The results showed that in both thickness categories the NCFs had the lowest energies and almost the largest damage area. The prepreg laminates had the highest energy value among the 2.5 mm thick specimens, but the largest damage area (delaminations). Whereas, the 3D woven MVR-444@90 5 mm thick specimens have the highest energy value and relatively small damage area.

The study indicates no major influence of matrix type (cured at the same temperature, 180 °C) on specimen results. The 3D woven specimens infused with the MVR444 resin, cured at 180 °C, had higher (by almost 1 kN) average Initial Failure Load values than the specimens infused with the RTM-6 resin, however both the Average Load Drop values and the Final Failure Load values were very similar. The MVR444 matrix system has a 4% strain to failure, compared with 3–4% of the RTM-6 matrix system. This small difference may have accounted for the differences in test results. The normalized IFL values show that the 3D woven specimens cured at 90 °C had lower values compared to the woven specimens cured at 180 °C. The 3D woven specimens cured at 90 °C had higher FFL values than the 3D woven specimens cured at 180 °C in both thickness categories, however the normalized FFL values showed that only in the 2.5 mm category the 3D woven specimens cured at 90 °C performed better than the woven specimens cured at 180 °C. This means that the high FFL value of the 5 mm thick 3D woven specimens cured at 90 °C was attributed more to the increased thickness of the specimens than to the influence of curing temperature.

Fig. 9. Comparison of damage area of different specimens: (a) 2.5 mm thick; (b) 5 mm thick.

3042

´ wik et al. / Composite Structures 94 (2012) 3037–3042 T. C

The authors were unable to identify to what extent the matrix microcracking influenced the performance of the 3D woven specimens cured at 180 °C. 3.5. Fibre volume fraction Table 1 shows the fibre volume fractions for all specimens. The results show that the fibre volume fraction may have influenced the IFL and the FFL values. The prepreg samples, of the highest Vf = 0.65, had the highest IFL value in both thickness categories. The normalized values show a similar trend. Whereas, the NCF samples, of the lowest Vf = 0.55, had the lowest both the IFL and the normalized IFL. The influence of fibre volume fraction on the IFL is less pronounced for the rest of the samples. This can be potentially attributed to the very small difference in the Vf between the specimens e.g. only 0.02 difference between the 2.5 mm thick 3D woven MVR444@180 and the 3D woven MVR444@90; and to the fact that such small change in the Vf might have been overshadowed by the effects caused by variations in the specimen thicknesses. The results show that the highest FFLs were achieved by specimens of the highest fibre volume fractions. The prepregs and the 3D woven MVR444@90 achieved the highest FFL in both thickness categories. The NCFs yielded the lowest FFL values. Also, it was the specimens of the highest fibre volume fraction that absorbed the largest amount of energy during the test. The authors do not know to what extent exactly the variations in the fibre volume fraction influenced the results. 4. Conclusion In total, 54 specimens were tested in the pull-through test, according to the Airbus Specimen and Pull-out Test Guidelines [11]. The bolt pull-through performance of three differently reinforced composites (prepreg laminate, Non-Crimp Fabric, non-crimp 3D woven fabric) was evaluated. The majority of the specimens tested exhibited similar behaviour (load–displacement curve) within their material group. The normalized plots showed that the IFL values are very similar for each specimen type in both thickness categories. However, the normalized FFL values showed that only the prepreg and the NCFs had similar values in both thickness categories. The difference between FFL values of the 3D woven MVR444@180 in the two thickness categories was approx. 23%, whereas for the 3D woven MVR444@90 about 10%. The authors believe that the higher fibre volume fraction in 5 mm thick specimens is responsible for the higher FFL values. The comparison of the IFL values and the normalized IFL values showed that in general the prepreg specimens had the highest IFL value, whereas the NCF specimens had the lowest. It appears that the slight difference in a specimen thickness influenced the relative performance between the groups. This was particularly noticed in the case of the 3D woven specimens. Regarding the normalized FFL values the prepreg laminates yielded the highest values only in the 2.5 m thickness category. In the 5 mm thickness category all specimens but NCFs had similar FFL values. The NCFs yielded the lowest FFL values. Although the results indicate that the prepreg specimens had the best pull-though performance (were capable of enduring the highest loads and had one of the largest areas under L vs. D curve), the Average Load Drop values revealed that these

specimens are also characterized by the highest drop in performance once the Initial Failure Load is reached. This is reflected in the change of curve slope at the load–displacement charts and extent of damage in the specimens. The prepreg laminates exhibited the greatest degree of delamination. Slightly less damage was apparent in the NCF specimens, though delaminations spread across entire width of samples as well. The 3D woven specimens had the smallest amount of delamination, localized in the bolt area. The authors believe that fact that the 3D woven composites exhibited relatively good pull-through performance compared to the prepreg specimens and that the damage area in the 3D woven specimens was far smaller than in prepreg laminates, make them an attractive alternative to the prepreg laminates. The authors did not identify any significant influence of resin system and curing temperature on the specimen performance. However, it is appears that the difference in fibre volume fractions might have influenced their specimens’ relative performance. In a number of cases the delaminations spread across entire width of the specimens. It is hard to estimate how far the delaminations would reach if the specimen was wider. The influence of specimen dimensions on character of the delaminations is beyond the scope of this paper, and is considered as a future continuation of the presented work. Acknowledgements The authors would like to acknowledge the Composite Manufacturing Engineering Research Team from Airbus in the UK for provision of the materials. They would also like to thank Alexander E. Bogdanovich from 3TEX for valuable discussion on some aspects of this paper. The support of Alexander Fergusson is acknowledged. References [1] Jain L, Mai Y-W. Mode I delamination toughness of laminated composites with through-thickness reinforcement. Appl Compos Mater 1 1994;1(1):1–17. [2] ASTM (2007) D 733210 7332M-07. Standard test method for measuring the fastener pull-through resistance of a fiber-reinforced polymer matrix composite. [3] Kim J, Mai Y-W, Baillie C, Poh J. Fracture toughness of CFRP with modified epoxy resin matrices. Compos Sci Technol 1992;43:283–97. [4] Kim J-K, Mackay D, Mai Y-W. Drop-weight impact damage tolerance of CFRP with rubber-modified epoxy matrix. Composites 1993;24:485–94. [5] Dorey G, Bishop S, Curtis P. On the impact performance of carbon fibre laminates with epoxy and PEEK matrices. Compos Sci Technol 1985;23: 221–37. [6] Dransfield K, Baillie C, Mai Y-W. Improving the delamination resistance of CFRP by stiching – review. Compos Sci Technol 1994;50:305–17. [7] Bogdanovich AE, Mohamed MH. Three-dimensional reinforcements for composites. SAMPE J 2009;45:8. [8] Lomov SV, Bogdanovich AE, Ivanov DS, Mungalov D, Karahan M, Verpoest I. A comparative study of tensile properties of non-crimp 3D orthogonal weave and multi-layer plain weave E-glass composites. Part 1: materials, methods and principal results. Compos Part A: Appl Sci Manuf 2009;40:1134–43. [9] Ivanov DS, Lomov SV, Bogdanovich AE, Karahan M, Verpoest I. A comparative study of tensile properties of non-crimp 3D orthogonal weave and multi-layer plain weave E-glass composites. Part 2: comprehensive experimental results. Compos Part A: Appl Sci Manuf 2009;40:1144–57. [10] Bogdanovich AE. Advancements in manufacturing and applications of 3D woven preforms and composites. In: Proceeding of the 16th international conference on composites materials (ICCM-16), Kyoto, Japan; July 8–13, 2007. [11] Salisbury, Ross, Research Engineer Airbus Operations Ltd. Personal communication with Salisbury, Ross; 2009. [12] Zhang D. Damage tolerance of 3D orthogonal woven composites. MSc thesis. Imperial College London; 2007.