Composites: Part A 39 (2008) 1572–1581

Contents lists available at ScienceDirect

Composites: Part A

journal homepage: www.elsevier .com/locate /composi tesa

Mechanical performance of carbon fibre-reinforced composites basedon preforms stitched with innovative low-melting temperatureand matrix soluble thermoplastic yarns

Uwe Beier a, Felipe Wolff-Fabris a, Frank Fischer a, Jan K.W. Sandler a, Volker Altstädt a,*, Gerrit Hülder b,Ernst Schmachtenberg b, Hermann Spanner c, Christian Weimer c, Tim Roser c, Wolfgang Buchs c

a Polymer Engineering, University of Bayreuth, Universitätsstraße 30, D-95447 Bayreuth, Bavaria, Germanyb Institute of Polymer Technology, University of Erlangen-Nürnberg, Am Weichselgarten 9, D-91058 Erlangen, Germanyc Eurocopter Deutschland GmbH, D-81663 Munich, Germany

a r t i c l e i n f o

Article history:Received 28 March 2008Received in revised form 9 June 2008Accepted 11 June 2008

Keywords:A. Polymer-matrix composites (PMCs)PreformLow-melting temperatureSoluble thermoplastic yarnB. Mechanical propertiesE. Resin transfer moulding (RTM)

1359-835X/$ - see front matter � 2008 Elsevier Ltd.doi:10.1016/j.compositesa.2008.06.003

* Corresponding author. Tel.: +49 (0)921 55 7471;E-mail address: altstaedt@uni-bayreuth.de (V. Alts

a b s t r a c t

In order to achieve a superior overall mechanical performance of composites based on stitched preformsfor demanding aircraft applications, innovative thermoplastic stitching yarns are comparatively evalu-ated in carbon fibre-reinforced epoxy composites. Low-melting temperature yarns based on polyamideand phenoxy in comparison to a standard polyester yarn allow prestabilisation of the dry preforms bythermobonding and lead to significantly reduced laminate disturbances following liquid compositemoulding; thereby minimising the degradation of the resulting composite strength properties. Whilethe softening polyamide yarns allow partial rearrangement of the carbon fibres during the resin injectionprocess, the dissolution and subsequent phase-separation of the phenoxy can induce a further localtoughening of the epoxy matrix. The improvements in overall composite performance when using stitch-ing yarns are partly due to the particular yarn material but also depend on variations in linear yarn den-sity. Last but not least, it is demonstrated that stitching seams close to a bolted joint have only little effecton the bolt bearing strength of the stitched composite whereas seams running directly through the holeand oriented in the load direction induce small degradation of the bolt bearing strength.

� 2008 Elsevier Ltd. All rights reserved.

1. Introduction

Modern stitching technologies are considered to be one of thekey approaches towards the cost-reduced manufacture of complextextile preforms used for liquid composite moulding (LCM) of high-performance fibre-reinforced polymer composites [1]. However,depending on various parameters including the used yarn materi-als, the stitching parameters as well as the particular type of com-posite and the loading conditions, the use of such stitchedpreforms has been shown to both improve or seriously degradethe in-plane and out-of-plane composite performance [2–8].

A key parameter influencing the overall in-plane mechanicalperformance of composites based on stitched preforms has beenidentified as the induced undulations of the reinforcement fibres[3]; an effect that is strongly related to the stitching yarn thicknessand yarn tension. Although some predictive models have been pro-posed [9–13], it still is rather difficult to accurately predict theresulting mechanical properties of stitched composites by clearcalculable formulas considering all relevant parameters.

All rights reserved.

fax: +49 (0)921 55 7473.tädt).

A feasible approach to resolve these challenges is the evasionof the main origin of the performance reduction, namely the dis-turbing and undulating presence of a solid, tensile loaded stitch-ing yarn. This can be realised using yarns that either melt orsoften during the LCM process or even dissolve in the matrix dur-ing the resin injection process. In order to accomplish the soften-ing/melting event, a low-melting temperature thermoplastic yarnreplacing the usual carbon, aramid, glass, polyester etc. yarns isrequired. For complete dissolution, a good miscibility of the yarnmaterial in the matrix for the used LCM-parameters is necessary.Once the thermoplastic yarn builds up a more or less well con-nected phase or dissolves in the matrix, additional toughening ef-fects can be anticipated as a result of different tougheningmechanisms such as crack pinning, micro cracking, localised shearyielding or banding, particle bridging, crack path deflection etc.[14].

The general toughening benefits of incorporating thermoplas-tics in epoxy resins have already been demonstrated [15–23]. Re-cent studies [7] showed no decrease of the compositecompression, interlaminar shear strength and compression afterimpact properties when using low-melting temperature polyam-ide-stitched (2*23 dtex) laminates. This kind of low-melting tem-

U. Beier et al. / Composites: Part A 39 (2008) 1572–1581 1573

perature thermoplastic also offers the additional advantage ofallowing a prestabilisation of the dry preforms by thermobonding.Exploiting this potential, the handability of the preforms should besignificantly enhanced without inducing negative effects on themechanical stability and the preforming time due to high stitchdensities; a critical issue for the desired use of automated manu-facturing approaches.

In this study, a detailed comparison of the mechanical perfor-mance of non-crimped carbon fibre-epoxy composites stitchedwith various thermoplastic yarns is provided. In addition to a stan-dard polyester yarn (2*53 dtex), two low-melting temperaturepolyamide yarns with differing linear yarn density (2*23 and2*75 dtex) as well as a thick phenoxy yarn (300 dtex) are evalu-ated. The resulting variations in composite performance are ex-plained based on morphological differences induced by thedifferent yarn materials.

Fig. 1. Schematic representation of the assembly seam stitching pattern used inthis study.

2. Materials and experimental details

2.1. Materials

The fibre reinforcement of the epoxy laminates was built up oflayers of commercial non-crimped carbon fibre (NCF) fabrics. Theparticular NCF is based on either (0�/90�), (90�/0�) or (�45�/+45�), (+45�/�45�) biaxial layers of carbon fibres (Tenax HTS12 k) with a polyester 40 dtex knitting thread as a binding yarn,manufactured by Hexcel; with a carbon fibre weight per unit areaof approximately 250 g/m2 per biaxial layer.

For stitching of the non-crimped fabrics, polyamide yarns(Grilon K 140) by EMS-Griltech, Switzerland, twined from twoyarns with either 23 or 75 dtex, respectively, or a polyester(standard) yarn (Serafil 200/2) by Amann, Germany, twined fromtwo yarns with a linear density of 53 dtex, were used. Additionally,a phenoxy yarn (Grilon MS300) with a linear density of 300 dtexwas employed. All yarns were not washed separately and wereused as delivered.

The epoxy matrix used for all composites was HEXFLOW RTM6aerospace grade, supplied by Hexcel, and was used as-received.

For determination of the composite bearing strength, anadditional prepreg laminate common for aerospace applicationswas prepared in order to allow the determination of a baselinevalue for this property. This particular laminate was built up of acommercial prepreg (HexPly M18/1) manufactured by Hexcelconsisting of a bidirectional (0�/90�) 4 harness satin weave fabric(Tenax HTA 3K) with an area weight of 220 g/m2 and an impactmodified epoxy resin.

2.2. Specimen manufacturing

The preforms for the compression, tensile, tensile fatigue andinterlaminar shear strength (ILSS) tests were stitched at ±45�(a = ±45�) with a modified lock-stitch arranged in the geometryschematically shown in Fig. 1, using a 2D-CNC sewing machinemanufactured by KSL, Germany. The stitch row spacing r was setat 40 mm. In order to ensure a good handling of the dry preformsand to obtain a high dimensional (bending) stability in either direc-tion (parallel and perpendicular to the preform outline), two cross-ing seams were employed. The layer stacking sequence used forthe intended compression, tensile, tensile fatigue and ILSS testswas [(0/90)]4s. In contrast, the layer sequence for the preforms in-tended for the compression after impact (CAI) tests was [(+45/�45)/(90/0)]4s (quasi-isotropic). In this case, the preforms werealso stitched with a modified lock-stitch with a nominal stitchpitch of 3 mm but the stitch row spacing was set at 20 mm andthe orientation of the seams was a = ±22.5�.

All composites (except the prepreg reference for the bearingstrength described in the following paragraph) were fabricatedusing vacuum-assisted resin transfer moulding (VARTM). Thesquare mould was preconditioned with release agent (Loctite770-NC Frekote), evacuated and heated up to 120 �C prior toimpregnation of the dry preforms with the epoxy resin at 120 �C,applying a pressure of 0.1 MPa. After complete impregnation,which took about 10 min, a dwell pressure of 0.2 MPa was appliedfor 10 min. The final curing temperature of 180 �C was kept con-stant for 2 h, followed by a slow cooling to room temperature.The final RTM panels were 420 � 420 mm2 in size, with a nominalthickness of 2.0 ± 0.1 and 3.9 ± 0.1 mm for the [0]4s lay-up and theCAI laminates, respectively.

For the bearing strength specimens, a quasi-isotropic [(0/90)/(+45/�45)]4S lay-up was chosen. The seams had a nominal pitchof 3 mm and were either orientated in the subsequent load direc-tion (a = 0�) or perpendicular to it (a = 90�). For stitching, only thepolyamide 2 � 75 dtex yarn was employed. Composite manufac-turing of plates with a nominal thickness of 3.9 mm was carriedout as described above.

The reference prepreg composite for the bearing strength testalso had a quasi-isotropic [(0/90)/(+45/�45)]4S lay-up comparableto the RTM bearing strength specimens introduced above and wasconsolidated using a hot-press. Curing was realised in a steppedheating cycle with a first heating step at 80 �C for 1 h followedby a second step at 180 �C for an additional 2 h. The final size ofthe panel was 450 � 450 mm2 with a thickness of about 3.9 mm.

In order to investigate the effect of phenoxy in the RTM6 epoxymatrix in more detail, compact plates of both the neat resin as wellas of modified RTM6 were prepared according to the followingprocedure:

The epoxy matrix was heated to 120 �C for mixing. The hot mix-ture was placed in a vacuum oven at the same temperature for 45–60 min for degassing. The modified resin was subsequently curedin a convection oven as follows: the curing cycle started by a2 K/min ramp from room temperature to 180 �C. This temperaturewas held for 2 h and curing was completed with a �2 K/min rampfrom 180 �C to room temperature. Thus, cured 110 � 110 � 3–4 mm3 epoxy resin sheets were obtained and were subsequentlymachined to the desired specimen size for further testing.

2.3. Panel quality assessment

The quality of all manufactured composite panels with regard tothe homogeneity, delaminations and porosity was verified by

1574 U. Beier et al. / Composites: Part A 39 (2008) 1572–1581

detailed impulse echo ultrasonic scans in water (Hilgus HFUS2400 Air Tech) as well as by light microscopy (Polyvar MET byReichert-Jung) of polished cross-sections. In addition, the resultingmisalignment of the reinforcing carbon fibres due to the stitchingwas further analysed in detail using micro-computer tomography(l-CT). Here, a Skyscan X-ray Microtomograph 1072 l-CT with amaximum spot resolution of 5 lm operating at a voltage of 79 kVand a current of 94 lA was used. Specimens of 5 � 5 � 2 mm3 wereexamined with the cone-beam technology, 2-dimensional (2D)cross-sectional images were taken in steps of 0.23� up to a com-plete rotation of 360� 3-dimensional (3D) volumetric images weredigitally reconstructed from the set of 1600 2D X-ray images byusing a proprietary software ‘‘NRecon” from Skyscan.

2.4. Mechanical testing

Dynamic mechanical analysis (DMA) of the cured matrix mate-rials without the reinforcement fibres was performed using aRDAIII from Rheometric Scientific at a heating rate of 4 K/min be-tween 25 and 300 �C. A rectangular specimen of 50 � 10 �3 mm3 cut out of the plates was used in the torsion rectangulargeometry mode, at a frequency of 1 Hz and 0.1% of deformation.The glass transition temperature (Tg) of the matrix was taken asthe temperature at which tan d showed a maximum.

The fracture toughness (KIc) and the fracture mechanical modu-lus (E) of the cured matrix materials were obtained from the open-ing mode test according to the ISO 13586, performed on compacttension (CT) specimens. The size of the specimens used in thisstudy was 41 � 40 � 3–4 mm3. A sharp notch was machined intothe specimens and then a sharp pre-crack was generated by tap-ping a razor blade into the notch. Fracture tests were carried outusing a universal testing machine model Zwick Z 2.5 operating atroom temperature and a cross-head speed of 10 mm/min. A mini-mum of five specimens for both the neat and the modified RTM6formulation was tested.

Composite specimens with the required dimensions for the var-ious tests were prepared by cutting the panels with a water-cooleddiamond saw (Mutronic DIADISC 6200) in such a way (except forthe bearing strength specimens) that the direction of the stitchingseam was ±45� to the loading direction which is identical with the0� fibre orientation. All specimen edges were subsequently pol-ished in order to eliminate surface defects.

In-plane compression tests were performed according to theDIN EN 2850 (Celanese type A3) specification. The composite spec-imens with the [(0/90)]4s lay-up had a thickness of approximately2.0 mm, a width of 6.35 mm, and a length of 110 mm. The gaugelength was 10 mm. The seam geometry in the specimens is demon-strated in Fig. 2. The specimens were equipped with sheet mould-ing compound (SMC) end taps. The specimens were tested using a

Fig. 2. Schematic representation of the assembly seam stitching pattern within thetested compression, tensile, tensile fatigue and ILSS specimens.

Zwick Z050 at a constant cross-head displacement speed of 1 mm/min at room temperature. For each composite type, at least fivespecimens were tested.

Tensile tests were carried out according to ASTM D 3039. Thespecimens with the [(0/90)]4s lay-up had a thickness of approxi-mately 2.0 mm, a width of 25 mm, and a length of 250 mm. Thegauge length was 130 mm. The seam geometry in these specimensis also shown in Fig. 2. The specimens were equipped with glass fi-bre-reinforced epoxy end taps. The specimens were tested using aZwick Z1485 at a constant cross-head displacement speed of2 mm/min at room temperature. For each composite type, at leastfive specimens were tested.

The corresponding tensile–tensile fatigue tests were carried outaccording to ASTM D 3479 at room temperature using a servo-hydraulic test machine (Schenk, IST 8800/ Software Dynmat byBASF AG, Ludwigshafen) at a constant stress ratio of 0.1 by apply-ing a sinusoidal waveform at a frequency of 5 Hz. The specimengeometry employed was that chosen for the quasi-static tensiletests.

ILSS tests (short beam) were performed according to the DIN EN2563 specification, using composite specimens with the [(0/90)]4s

lay-up. Rectangular specimens with 10 mm in width and 20 mmin length were again tested using a Zwick Z050 operating at roomtemperature. The span between the support cylinders was 10 mm,resulting in a span-to-thickness ratio of about five. The loading cyl-inder was forced against the specimen at a constant cross-headdisplacement speed of 1 mm/min. The maximum apparent shearstrength, s, in the mid-plane of the composite was approximatedby the Euler–Bernoulli beam theory

s ¼ 3P=4 wt; ð1Þ

where P is the force at failure, w is the specimen width, and t is thespecimen thickness.

In order to obtain specimens for the compression after impacttests, impacted specimens with the [(+45/�45)/(90/0)]4s lay-upwere prepared by clamping a specimen on a plate according toAITM 1.0010. The impactor had a weight of 3 kg and a sphericaldiameter of 16 mm. The impact energy was fixed at 25 J for allcomposite types under investigation. An anti-rebound device con-trolled the impactor. Following the impact, the CAI tests measure-ment at room temperature were carried out according to AITM1.0010. The specimens had a thickness of 3.9 mm, a width of100 mm, and a length of 150 mm. For the compression tests, aZwick Z1485 was used at a constant cross-head displacementspeed of 1 mm/min. The CAI measurements were repeated sixtimes for each composite type.

The bearing strength of the composites was determined accord-ing to AITM 1.0009 for the pin-loading configuration, i.e. betweenthe specimen and the load transferring steel plates a gap of 0.5 mmwas realised so that no additional clamping force was applied. Theused specimens had a thickness of 3.9 mm, a width of 75 mm, anda length of 175 mm; the clearance hole had a nominal diameter of6.35 mm and was placed 25 mm from the specimen’s free edge. Inorder to characterise the seam influence, the seams were eitherorientated in the load direction through the centre of the clearancehole or perpendicular to the load direction in a distance of 1 mmfrom the loaded edge of the clearance hole, as outlined in Fig. 3.The tests were performed at room temperature using a Zwick1485 testing machine operating at a constant cross-head displace-ment speed of 1 mm/min. For each seam orientation six specimenswere tested.

The bearing stress rb is calculated as:

rb ¼ P=wt; ð2Þ

where P is the applied load, w is the specimen width, and t is thethickness of the specimen.

Fig. 3. Schematic representation of the assembly seam stitching pattern within the tested bearing strength specimens.

U. Beier et al. / Composites: Part A 39 (2008) 1572–1581 1575

3. Results and discussion

The overall aim of using this specific seam geometry and theparticular polyester, polyamide, or phenoxy yarns as assemblyseams to ensure a good handling of the dry NCF preforms wasachieved. In addition, all assembly seams appeared to be of goodquality, independent of the yarn type. Following initial thermob-onding trials, the polyamide- and the phenoxy-stitched preformsespecially revealed an enhanced stiffness and handability; an effectthat can be exploited to a greater extent in the future.

All moulded composites also appeared to be of a homogeneousgood quality (verified using the ultrasonic analysis technique) andsurface finish.

However, the cross-sectional computer tomography micro-graphs (Fig. 4, top row) as well as the cross-sectional light micros-copy images (Fig. 4, bottom row) show the reinforcement fibresdisplaced from their original alignment in the vicinity of the stitch-ing yarn. Not surprisingly, this fibre undulation appears to be in-creased when replacing the polyamide 2*23 dtex with the thicker2*75 dtex yarn. Comparison of the polyamide 2*75 dtex and phen-oxy 300 dtex yarns however revealed similar undulations althoughthe phenoxy yarn has about double the linear density as the poly-amide yarn. In contrast, the undulations induced by the polyester2*53 dtex yarn are between those induced by the two polyamideyarns, matching expectations. For both polyamide stitches (Fig. 4,top row), some sort of ‘‘inclusions” are clearly visible in the mi-cro-CT images. In these areas, the X-ray absorption is very low,indicating the presence of gas pockets.

Fig. 4. Representative cross-sectional computer micro-tomography (first row) and light mthe stitching yarn penetrating the laminate.

The mechanism of this void formation is not known but is cur-rently under investigation. In contrast to the polyester-stitchedlaminate, both polyamide-stitched composites (Fig. 4, bottomrow) showed no remaining individual filaments of the yarn, verify-ing that the yarns melt and agglomerate during the VARTM-pro-cessing [7]. Interestingly, the phenoxy yarn is not visible in thecross-sectional micrograph of the vicinity of a stitch anymore(Fig. 4, bottom row). This observation strongly implies that eitherthe yarn has dissolved completely during processing and hasphase-separated below the resolution limit of the light microscopeor that the yarn has simply been washed away from the stitch-holeby the injected resin.

3.1. Fracture toughness

The fracture toughness of the 10 wt% phenoxy-modified epoxyresin (0.77 ± 0.02 MPa

pm) is increased by approximately 20% as

compared to the neat RTM6, while the modulus deduced fromthe KIc measurements appears to be rather similar with3077 ± 250 MPa and 3320 ± 215 MPa for the neat and the modifiedresin, respectively. The Tg data obtained from the DMA measure-ments (not included here) showed a slight reduction from 230 �Cfor the pure RTM6 to 223 �C for the phenoxy-modified system. Thiseffect is a consequence of a small reduction of the crosslinking den-sity of the thermoset network as given by the storage modulus val-ues of the systems above Tg. Additional detailed evaluation of themodified RTM6 matrix by means of transmission electron micros-copy verified phase-separation of the phenoxy on the nanoscale

icroscopy micrograph images (second row) of the stitched NCF composites showing

Fig. 6. Comparative plot of the ultimate compression strength of the variouscomposites.

1576 U. Beier et al. / Composites: Part A 39 (2008) 1572–1581

within the epoxy matrix; explaining both the observed variationsin toughness as well thermal properties.

These experimental results clearly show that there is a tough-ness enhancement of RTM6 when adding 10 wt% of phenoxy with-out inducing significant detrimental effects on either matrixstiffness or Tg. Nevertheless, it is quite difficult to predict the actuallocal phenoxy-content in stitched composites when incorporatingthe phenoxy by means of the stitching yarn as the total amountof yarn material is low and as the soluble yarn material is likelywashed away from the stitch hole during the resin infiltration.Although FT-IR (Fourier-Transformations-Infrared-Spectroscopy)measurements at the stitch holes could not highlight an increasedphenoxy content at these positions within experimental accuracy,the presence of phase-separated nanoscale particles cannot be ru-led out.

3.2. Compression performance

The carbon fibre content of all composites investigated was be-tween 50 and 61 vol%. It must be noted that all the composite com-pression stiffness and strength values reported in this study arenormalised to a fibre volume content of 56 vol% (mean value ofall specimens under investigation) instead of the widely used60 vol%. This normalisation procedure ensures better comparabil-ity of all results for the different laminates; a detailed descriptionof the normalisation approach is given in the literature [24].

Fig. 5 shows a comparison of the normalised in-plane compres-sion modulus of all composite systems under investigation. As canbe seen, none of the assembly seams significantly affected themodulus of the non-crimped fabric composites.

In contrast, the data summarised in Fig. 6 shows that only theuse of the thin polyamide 2*23 dtex yarn allowed maintainingthe normalised compression strength; the 2*75 dtex polyamideassembly seam for example reduced the normalised compressionstrength of the laminates by about 16% as compared to the non-stitched NCF reference. Surprisingly, the use of the comparativelythick phenoxy yarn (300 dtex) reduced the normalised compres-sion strength by 10% only. Yet, it has to be emphasised that allthe thermobondable yarns led to superior composite propertieswhen compared to the standard polyester reference system, inde-pendent of the variations in linear yarn density.

These experimental observations regarding the effect of assem-bly seams on the compression behaviour of carbon fibre-epoxyRTM laminates are in good agreement with previously publisheddata [6,7], where compression strength reductions up to 19% werereported whereas the compression modulus was not affected by

Fig. 5. Comparative plot of the compression stiffness of the various composites.

stitching. It should be noted, that in [6] polyester yarns of2 � 53 dtex were used in a modified zigzag lock-stitch geometry.

In order to correlate the compression performance of the lami-nates with the resulting fibre undulations in the stitched NCF com-posites, these undulations were evaluated in more detail usingboth micro-computer tomography as well as light microscopy.The l-CT images of the stitched NCF composites in Fig. 4 (toprow) and the cross-sectional microscopic images in Fig. 4 (bottomrow) highlight representative cross-sections. As evidenced by theseimages, there are severe local fibre undulations as well as corre-sponding resin-rich regions in the immediate vicinity of stitchholes for all yarn types. However, the overall average fibre undula-tion appears low for all types. It is therefore not surprising that themodulus of the NCF composites is hardly influenced by the assem-bly seams, as this parameter depends on the average fibre undula-tion [3]. In contrast, the more pronounced maximum fibreundulation for the thick polyamide 2*75 dtex yarn compared tothe thin 2*23 dtex yarn can be used to explain the observed reduc-tion in compression strength [3]. This observed reduction mightalso be enhanced by the more severe inclusions observed for thepolyamide 2*75 dtex yarn, resulting in cracks starting at thesepoints. As the disturbance of both the polyamide 2*75 dtex andthe phenoxy 300 dtex yarn is about the same, it appears likely thatan additional toughening modification due to the phenoxy (sup-pressing cracks starting from the undulated regions in the vicinityof the stitch) prevents more severe reductions of the compressionstrength with the increasing yarn thickness. Similarly, the low-modulus (macroscopic) second phase evolved from the polyamide2*75 dtex yarn combined with the observed gas inclusions mightlead to an insufficient support of the locally undulated reinforce-ment fibres in the vicinity of the stitch; thus allowing local fibrebuckling and a further reduction of the compression properties.Such a reduced compression strength performance is generallyfound for low-modulus matrix systems [25].

The impact of the employed yarn materials on the compositecompression properties is further confirmed when comparingthem to the inferior properties of the polyester-stitched compos-ites. The observed severe degradation of the compression perfor-mance – although a thin yarn is used – can be attributed to aninsufficient interphase between the yarn filaments and the matrixand/or the lack of yarn tension relaxation during the VARTM pro-cess maintaining the initially induced severe undulations.

3.4. Quasi-static and dynamic tensile performance

In contrast to the compression performance, it should be notedthat all the tensile stiffness and strength values reported in this

Fig. 8. Comparative plot of the ultimate tensile strength of the various composites.

U. Beier et al. / Composites: Part A 39 (2008) 1572–1581 1577

study are normalised to the fibre volume content of 60 vol% widelyused in literature as tensile properties are considered to be lesssensible to such normalisation routines then the compressionproperties. A detailed description of the normalisation approachis given in the literature [24].

Fig. 7 shows a comparative plot of the normalised in-plane ten-sile modulus of all composite systems under investigation. As canbe seen, none of the assembly seams significantly affected themodulus of the various composites.

In contrast, the use of the standard polyester yarn induced areduction of the normalised tensile strength of the laminates by14% as compared to the non-stitched NCF reference, shown inFig. 8. Considering the polyamide-stitched laminates, the thinpolyamide 2*23 dtex yarn did not reduce the tensile strength sig-nificantly, while the thicker 2*75 dtex polyamide yarn led to aninferior tensile strength (�8%). The phenoxy yarn (300 dtex) ledto a reduction of the normalised tensile strength by about 13%. Thislimited degradation of the tensile strength for the phenoxy-stitched laminates again appears surprising, given the initial linearyarn density of about three times that of the standard polyesteryarn.

These experimental observations regarding the effect of assem-bly seams on the tensile behaviour of carbon fibre-epoxy VARTMlaminates are again in good agreement with previously publisheddata [6], where significant tensile strength reductions were re-ported whereas the tensile modulus was not affected by stitching.

The mechanical performance data can again be explained con-sidering the undulations of the reinforcing carbon fibres in thevicinity of the stitches (Fig. 4) in more detail. As in the case ofthe compression modulus, the tensile modulus of the NCF compos-ites is hardly influenced by these assembly seams as this parame-ter depends on the average fibre undulation [3].

The variations in maximum fibre undulation as well as stitchinginduced fibre damage occurring locally at critical areas of misalign-ment can be used to explain the observed reductions in tensilestrength [3] for the various yarn types. However there is reasonto believe that fibre damage is on the same level for all stitchedcomposites since the stitching-head as well as the employedstitching needle were the same for all yarns under investigation.Never the less further investigations on this issue is conducted toobtain detailed evidence.

Regarding the stitching induced misalignment of reinforcing fi-bres, the thicker standard polyester yarn (2*53 dtex) compared tothe thin 2 � 23 dtex polyamide yarn induced a minimised maxi-mum fibre undulation contributing to a superior tensile strengthon the level of the unstitched reference. Considering both polyam-ide-stitched composites, the thicker 2*75 dtex polyamide yarn de-

Fig. 7. Comparative plot of the tensile stiffness of the various composites.

graded the tensile strength more severely reflecting the morepronounced undulations. Furthermore, the gas inclusions detectedin the vicinity of the stitches might have acted as crack initiationsites and might have offset a better tensile strength performance.Although the employed phenoxy yarn has about double the lineardensity of the thick polyamide yarn, the composite tensile strengthwas only marginally lower. In view of the initial yarn thickness, thedissolution of the phenoxy yarn during processing and followingrearrangement of the carbon fibres should be responsible for thegood performance of these composites.

In contrast to the compression properties discussed previously,the tensile performance of the phenoxy-stitched laminates wasnevertheless inferior to those of the composites based on the poly-amide 2*75 dtex yarn. This effect is likely due to the differentmechanisms of tensile and compression failure which are stronglyrelated to the matrix stiffness. As stated before, the low-modulus(macroscopic) second phase build up of the polyamide 2*75 dtexyarn combined with the observed gas inclusions might lead to aninsufficient support of the locally undulated reinforcement fibresin the vicinity of the stitch under compressive loads, thus reducingthe compression properties. In contrast, this soft second phasewith an elevated failure strain may reduce the local stresses ofthe undulated reinforcement fibres in the vicinity of stitch holesunder tensile loads, thereby contributing to superior tensile prop-erties. Again, such a behaviour of improved tensile strength perfor-mance is commonly found for low-modulus matrix systems [26].

The measured composite tensile fatigue properties are dis-played in double logarithmic graphs (Fig. 9), showing the norma-lised maximum tensile stress plotted as a function of the cyclesto failure. As can be seen, all stitched laminates under investigationshowed a significantly shorter fatigue life at high stress levels. Thisdrawback due to stitching tends to lessen as the fatigue stress is re-duced. Depending on the specific yarn employed, the composite fa-tigue life reached the same level as the non-stitched reference atdifferent stress levels. The phenoxy 300 dtex stitched compositesretained the level of the non-stitched laminate at the highest stresslevel.

(It must be noted that the 2*23 dtex polyamide-stitched lami-nate is not included in this evaluation due to an insufficient num-ber of valid data points.)

The observed tensile fatigue performance of the various com-posites once again reflects the two main parameters – namelythe linear yarn density as well as the employed yarn material. Con-sidering the superior level of the composite with the comparablythick phenoxy 300 dtex yarn at low stress levels in comparisonto the polyamide 2*75 dtex-stitched composite, there is reason tobelieve that the toughening modification due to the phenoxy or

Fig. 9. Comparative plot of tensile fatigue life of the various composites.

Fig. 10. Comparative plot of the apparent interlaminar shear strength of the variouscomposites.

1578 U. Beier et al. / Composites: Part A 39 (2008) 1572–1581

at least the lack of a second (macroscopic) phase may outweigh thedetrimental effects of the induced comparable undulations. Fur-thermore, a stiff matrix is known to enhance the tensile fatigue lifeat low maximum stress levels, whereas low-modulus matrix sys-tems show an increased fatigue life at high stress levels [27]. Thus,a soft matrix as in the case of the polyamide 2*75 dtex yarn com-bined with the observed gas inclusions could be responsible for thedetrimental effects on the tensile fatigue life of such stitched com-posites at low stress levels.

The impact of the employed yarn material on the tensile fatiguelife of the stitched composites is further confirmed by the inferiorproperties of the polyester-stitched composites containing thecomparatively thin polyester yarn.

These experimental observations regarding the interaction ofstitching and the tensile fatigue performance of stitched compos-ites are consistent with data from the literature [8]. Kevlar-stitchedfibre-dominated graphite/epoxy laminates showed a similarlyinferior fatigue life at high stress levels and a comparable level atlow stress levels. Here, this increased behaviour at low stress levelswas attributed to a delayed propagation of delaminations due tothe presence of through-the-thickness yarns annihilating the dis-turbance of the reinforcing fibres exerted by the stitching holes.

3.5. Apparent interlaminar shear strength

The apparent interlaminar shear strength of the various com-posites measured using the short beam testing method is summa-rised in Fig. 10. As can be seen, stitching with either polyamide orphenoxy yarn types did not influence the ILSS property withinexperimental accuracy, whereas the standard polyester yarn-stitched composites revealed a significantly reduced shear strengthperformance.

This reduced ILSS value for the polyester yarn-stitched compos-ites reflects stitching-induced flaws such as the weak interface be-tween the polyester yarn and the epoxy matrix. It appears likelythat both the polyamide and phenoxy yarn types help to suppresscracks starting at the undulated regions in the vicinity of the stitch,although the particular mechanisms are different.

3.6. Compression after impact

The ultrasonic C-scan images displayed in Fig. 11 show that thegeneral appearance and size of the delaminated areas induced bythe 25 J impact was quite similar for the non-stitched (A) and thepolyester-stitched (B) composites, whereas the appearance of bothpolyamide-stitched (C, D) and the phenoxy-stitched composites (E)

were different. In agreement with this optical evaluation, the del-aminated areas were determined as 19 ± 8 cm2 and 17 ± 7 cm2

for the non-stitched and the polyester-stitched NCF composites,respectively. In contrast, the delaminated areas for both polyam-ide-stitched composites as well as for the phenoxy-stitched lami-nates were about 11 ± 2 cm2.

Although these values indicate an enhanced damage resistancefor the polyamide and the phenoxy-stitched composites, there isno measurable evidence for increased compression after impactproperties as shown in Fig. 12; all four CAI strength values arewithin the same range. The anticipated increase in CAI strengthdue to toughening effects generated by either the polyamide orthe phenoxy yarn did not occur. However, this experimental obser-vation is consistent with data from the literature [28], where car-bon/epoxy NCFs laminates stitched with Kevlar 29r showed apronounced decrease in delaminated area after impacting com-pared to corresponding polyester-stitched NCFs but revealed no in-creases in residual compression strength.

The observed behaviour of residual compression strength maybe attributed to specific delamination arrangements and intensi-ties after impact due to various stitching yarns as well as to theempolyed binder-yarn connected biaxial reinforcement layers ofthe NCF substructure itself. The potential suppression of sublami-nate ply buckling, which may occur using this CAI test methodmay promote failure due to microbuckling and kinking of theload-bearing fibres in the damaged region independent fromdelamination size.

3.7. Bearing strength

Bolted joints are still the most important joining technique inaircraft production although it does not appear to be very suitablefor endless fibre-reinforced composites [29]. Several factors thatcan affect the bearing behaviour of such joints e.g. the stacking se-quence, the clamping force, and the hole clearance have beeninvestigated thoroughly in the past [30–32]. As stitching technolo-gies become an interesting technique for preforming of RTM-parts,the question arises if the fibre undulations as a result of the stitch-ing process will influence the bearing strength of bolted joints.

The bearing strength of the non-stitched NCF composites showsan average value of 502 ± 20 MPa, which is close to that of refer-ence composites prepared from the standard prepreg material(Hexcel M18/1 G939) with a similar lay-up (commonly used foraircraft applications) of 533 ± 31 MPa. Fig. 13 shows a representa-tive stress-deformation curve for each of the materials underinvestigation in this study (prepreg, unstitched and polyamide-

Fig. 11. Representative ultrasonic C-scan images showing the damage areas of the various composites after a 25 J impact.

Fig. 12. Comparative plot of the CAI strength of the various composites followingan impact at 25 J.

Fig. 13. Load-displacement curves of the bolt bearing test.

U. Beier et al. / Composites: Part A 39 (2008) 1572–1581 1579

stitched NCF composites), whereas the deformation on the hori-zontal axis is denoted as hole deformation s divided by the originaldiameter D. The initial peak of these curves represents the bearingstrength, while the curve progression beyond this point representsthe amount of energy necessary to maintain further hole deforma-tion. According to AITM 1.0009 not only the bearing strength givenby the initial peak in the stress-deformation curve but also thebearing stress for some specific points of irreversible hole deforma-tion were determined. These values are summarised in Fig. 14.

The polyamide-stitching seams orientated perpendicularly tothe load direction appear to have no influence on the bearingstrength, as the average bearing strength (504 ± 19 MPa) is almostidentical to that of the unstitched NCF reference. These specimenshave only one stitch close to the loaded edge of the clearance hole,so that the induced fibre undulation is of negligible influence. Incontrast, seams orientated in the load direction lead to a loss inthe bearing strength of almost 12% (446 ± 35 MPa).

Microscopic analysis showed that all NCF specimens fail as a re-sult of kinking of the fibres orientated in the load direction (0�-

Fig. 14. Characteristic bearing strength values according to AITM 1-0009.

1580 U. Beier et al. / Composites: Part A 39 (2008) 1572–1581

layer) close to the loaded edge of the clearance hole; an effect thatfinally leads to the induction of inter-fibre fractures in the 90� and±45�-layers, demonstrated in Figs. 15–17. Due to stitching in the

Fig. 16. Light microscopy micrograph of a bolt bearing specimen with a seam orienta

Fig. 15. Light microscopy micrograph of a

load direction, the a = 0� fibres are displaced by the polyamide yarnand no fibres with 0�-orientation can be found in the seam’s plain,Fig. 17. This effect leads to a locally reduced fibre content of fibresorientated in load direction (0�-fibres) in the area of the highestcompressive stress, consequently leading to a reduced compositebearing strength.

Compared to the prepreg and the unstitched NCF reference sys-tems, the stitched NCF composites can generally sustain higherloads beyond the initial peak. This behaviour suggests that thepresence of the polyamide yarn and the locally elevated matrixcontent lead to a slightly more ductile interlaminar failure, as ver-ified by Fig. 16 representing a NCF composite stitched perpendicu-larly to the load direction. Thus, more destructive energy isdissipated before delamination will occur, leading to the conclu-sion that stitched NCF composites (depending on the seam orienta-tion) do not necessarily show a reduced bearing strength but havean improved ‘‘fail-save” performance.

4. Conclusions

The use of thermoplastic stitching yarns that either melt or soft-en during preforming and/or the subsequent RTM process or ofyarns that are soluble in the matrix are shown to present a feasibleway of utilising stitching technologies for the future automatedmanufacture of textile performs used for liquid composite mould-ing of high-performance polymer composites. The experimental

ted perpendicularly to the load direction, with region of ductile matrix fracture.

n unstitched bolt bearing specimen.

Fig. 17. Light microscopy micrograph of a bolt bearing specimen with a seam orientated in load direction showing the deflection of the 0�-orientated fibres.

U. Beier et al. / Composites: Part A 39 (2008) 1572–1581 1581

results presented in this study demonstrate a generally highmechanical property level for stitched composites as comparedto the non-stitched non-crimped carbon fibre reference. Novelstitching yarn materials such as polyamide or phenoxy clearly leadto an improved mechanical composite performance as comparedto those achieved with a standard polyester yarn. Evaluating theprovided data in more detail, it becomes apparent that yarnparameters such as linear density as well as the base thermoplasticmaterial both influence the microstructure of the composite andthe resulting overall properties; thereby providing an effectivemeans to optimise the mechanical performance of stitchedcomposites.

For a further improvement of mechanical composite properties,soluble yarns appear to be quite promising due to possible localtoughening effects. As demonstrated here, the mechanical proper-ties of composites stitched with soluble phenoxy yarns were supe-rior to those obtained with other thermoplastics, independent ofthe significantly greater linear density of the phenoxy yarn. Yetfurther progress is expected by tailoring these yarn materials aswell as the RTM–process in order to achieve greater control overthe local morphological and concentration aspects of the thermo-plastics in the cured matrix.

Acknowledgement

The authors thank the Bavarian Research Foundation (GrantFlexnaht–Strukturen) for financial support.

References

[1] Weimer C, Mitschang P. Aspects of stitch formation process on the quality ofsewn multi-textile-preforms. Comp Part A 2001;32(10):1477–84.

[2] Mouritz AP, Leong KH, Herszberg I. A review of the effect of stitching on the in-plane mechanical properties of fibre-reinforced polymer composites. CompPart A 1997;28(12):979–91.

[3] Mouritz AP, Cox BN. A mechanistic approach to the properties of stitchedlaminates. Comp Part A 2000;31(1):1–27.

[4] Moll KKU. Nähverfahren zur Herstellung von belastungsgerechten Fügezonenin Faserverbundwerkstoffen. Aachen: Shaker-Verlag; 1999.

[5] Larsson F. Damage tolerance of stitched carbon/epoxy laminate. Comp Part A1997;28(11):923–34.

[6] Beier U, Fischer F, Sandler JKW, Altstädt V, Weimer C, Buchs W. Mechanicalperformance of carbon fibre-reinforced composites based on stitchedpreforms. Comp Part A 2007;38(7):1655–63.

[7] Beier U, Fischer F, Sandler JKW, Altstädt V, Weimer C, Spanner H, et al.Evaluation of preforms stitched with a low melting temperature thermoplasticyarn in carbon fibre-reinforced composites. Comp Part A 2008;39(5):705–11.

[8] Aymerich F, Priolo P, Sun CT. Static and fatigue behaviour of stitched graphite/epoxy composite laminates. Comp Sci Tech 2003;63(6):907–17.

[9] Dransfield KA, Jain LK, Mai YW. On the effects of stitching in CFRPs-I. Mode Idelamination toughness. Comp Sci Tech 1998;58(6):815–27.

[10] Jain LK, Dransfield KA, Mai YW. On the effects of stitching in CFRPs-II. Mode IIdelamination toughness. Comp Sci Tech 1998;58(6):829–37.

[11] Mouritz AP, Jain LK. Further validation of Jain and Mai models for interlaminarfracture of stitched composites. Comp Sci Tech 1999;59(11):1653–62.

[12] Pang FY, Wang CH. A predictive creep model for un-stitched and stitchedwoven composites. Comp Sci Tech 2000;60(2):255–61.

[13] Mouritz AP. A simple fatigue life model for three-dimensional fiber-polymercomposites. J Comp Mater 2005;40(5):455–69.

[14] Zhang M. (2003) A review of the epoxy resin toughening. (http://writing.syr.edu/~cslipson/419619/LiteratureReviewrevised.pdf).

[15] Lo Faro CRM, Pederson C. Recent developments in preform assembly andtextile technology, SAMPE International Conference 2005, Paris.

[16] Abusafieh ARB, Doyle M, Lo Faro C, Maskell R, Price R. Process and performanceenhancements in resin infusion using an epoxy-soluble non-woven veil,SAMPE International Conference 2006, Paris.

[17] Kim J, Robertson RE. Toughening of thermoset polymers by rigid crystallineparticles. J Appl Polym Sci 1992;27(1):161–74.

[18] Andres MA, Garmendia I, Valea A, Eceiza A, Mondragon I. Fracture toughness ofepoxy resin modified with polyethersulfone: Influence of stoichiometry on themorphology of the mixtures. J Appl Polym Sci 1998;69(1):183–91.

[19] Hodgkin JH, GP S, Varley RJ. Thermoplastic toughening of epoxy resins: acritical review. Polym Adv Tech 1998;9(1):3–10.

[20] Francis B, Thomas S, Jose J, Ramaswamy R, Rao VL. Hydroxyl terminated poly(ether ether keton) with pendent methyl group toughened epoxy resin:miscibility, morphology and mechanical properties. Polymer2005;46(26):12372–85.

[21] Girard Reydet E, Vicard V, Pascault JP, Sautereau H. Polyetherimide-modifiedepoxy networks: Influence of cure conditions on morphology and mechanicalproperties. J Appl Polym Sci 1997;65(12):2433–45.

[22] Martinez I, Martin MD, Eceiza A, Oyanguren P, Mondragon I. Phase separationin polysulfone-modified epoxy mixtures. Relationships between curingconditions, morphology and ultimate behavior. Polymer 2000;41(3):1027–35.

[23] Siddhamalli SK, Kyu T. Toughening of thermoset/thermoplastic composites viareaction-induced phase separation: epoxy/phenoxy blends. J Appl Polym Sci2000;77(6):1257–68.

[24] Anon. Polymer matrix composites – vol. I. Guidelines for characterization ofstructural materials. MIL-HDBK-17-1D U.S. Department of Defense, 1994.

[25] Baron Ch. Mechanische eigenschaften kohlenstofffaserverstärkter kunststoffe(CFK) bei variation der Matrixduktilität und der Bruchdehnung der Faser.Dissertation, Bremen, 1990.

[26] Baron C, Schulte K, Harig H. Influence of fibre and matrix failure strain on staticand fatigue properties of carbon fibre-reinforced plastics. Compos Sci Technol1987;29(4):257–72.

[27] Curtis P. An investigation of tensile fatigue behaviour of carbon fibrecomposite laminates. ICCM-6 Proceedings, Imperial College of Science andTechnology, vol. 4. London, UK: Elsevier Applied Science Publishers; 1987. p.54–64.

[28] Bibon GA, Hogg PJ, Backhouse R, Mills A. Carbon-fibre non-crimp fabriclaminates for cost-effective damage-tolerant structures. Comp Struct1998;58(1):129–43.

[29] Ehrenstein GW. Handbuch kunststoffverbindungstechnik. München: Carl-Hanser-Verlag; 2004.

[30] Park HJ. Effects of stacking sequence and clamping force on the bearingstrengths of mechanically fastened joints in composite laminates. Comp Struct2001;53(2):213–21.

[31] McCarthy MA, Lawlor VP, Stanley WF, McCarthy CT. Bolt-hole clearance effectsand strength criteria in single-bolt, single-lap, composite bolted joints.Compos Sci Technol 2002;62(10–11):1415–31.

[32] Stellbrink K. Dimensionierung von Krafteinleitungen in FVW-Strukturen. DLR-Mitteilung 93-12, Deutsche Forschungsanstalt für Luft- und Raumfahrt e.V.,Köln, 1993.