Composite materials with 3-dimensional (3D) reinforcement were manufactured and corresponding simulation models were created in parallel. The used simulation approach has earlier been shown to produce close to authentic geometrical representation of the yarn architecture in 3D reinforcement. It is shown that although the as-woven reinforcement pattern can be modelled quite reliably, significant distortion from the nominal fibre arrangement might take place later in manufacturing, primarily related to compression during moulding. Such effects have earlier received significant attention for composites with 2-dimensional reinforcement but not as much for their 3D counterparts. The yarns in the real and the simulated materials are studied and compared, and some of the discrepancies and the mechanisms behind are discussed. The distortions are partly attributed to the relatively sparse weave that allows yarns oriented in the through-thickness direction, in particular, to deviate from their original positions.

Composites with three-dimensional (3D) reinforcement have several benefits, compared to laminated composites. Their through-thickness reinforcement increase the out-of-plane properties significantly and could eliminate problems with delaminations. Also, these composites have proven to have great damage tolerance and energy absorption. However, their complex yarn architectures make it challenging to predict their mechanical response and performance.

In this thesis different aspects of composites with 3D woven reinforcement are explored. The focus is on a specific yarn architecture, called fully interlaced 3D weave. The results are however not only limited to that specific 3D reinforcement but could to a certain extent also be applicable to 3D reinforcement in general.

Preforms with fully interlaced 3D weaves were manufactured and impregnated with epoxy. These were then examined in great detail with computer tomography (CT) to study the internal yarn architecture after impregnation. Analysis showed that the yarns were quite significantly distorted by the through-thickness compression during impregnation. The distortion was attributed to the relatively sparse weave, not supporting the through-thickness reinforcement, which therefore distorts and brings the rest of the yarns along with it. In parallel, a simulation model of the internal geometry of the manufactured material was developed. The simulation model was however not designed to include the distortions encountered in the physical material.

The manufactured material and its corresponding model was tested in a tensile test setup. Two different thicknesses of the material was manufactured as well as a corresponding composite with two-dimensional (2D) reinforcement. Results showed that the material with 2D reinforcement was stiffer and stronger than the ones with 3D reinforcement, which was attributed to the lower crimp in the 2D reinforcement. A difference in stiffness between the two 3D weaves was also found and addressed to the larger amount of surface layer in the thinner weave, where vertical weft yarns are aligned with the warp direction and contributing to the overall stiffness in that direction. Failure analysis of specimens tested in the warp direction showed that initial cracks form in the boundaries of vertical weft yarns, close to the material surface. For specimens tested in the horizontal weft direction, initial cracks were found through the vertical weft yarns at the surfaces. Both these findings were supported by results from the simulation model.

An application for composites with fully interlaced 3D weave was also explored, where it was integrated as a fillet in a composite T-joint. The scope here was to make a 3D reinforced fillet, having low transverse thermal expansion which would decrease the residual stresses in the fillet after curing. T-joints with conventional fillets and fillets with 3D woven reinforcement were manufactured and tested in a pull-off test. Results showed that T-joints with conventional fillets had higher strength, but also higher spread, than the ones with 3D reinforcement. The higher strength of T-joints with conventional fillets was attributed to their better ability to adapt to the T-joint cavity, while the fit was not as good for the 3D fillets. The lower spread in strength of the T-joints with 3D fillets was attributed to their lower sensitivity to minor flaws such as voids inside the fillet.

Kompositer med tredimensionell (3D) armering har flera fördelar jämfört med laminerade kompositer. Deras fibrer genom tjockleken ökar ut-ur-planet-egenskaperna avsevärt och kan eliminera problem med delaminering. Dessa kompositer har också visat sig ha stor skadetolerans och energiabsorption. Men deras komplexa fiberarkitektur gör det svårt att förutsäga deras mekaniska respons och prestanda.

I denna avhandling utforskas olika aspekter av kompositer med 3D vävd armering. Fokus ligger på en specifik 3D vävd armeringsarkitektur, där varpen är sammanflätad med både horisontell och vertikal väft. Upptäckterna är dock inte bara begränsade till den specifika 3D-armeringen utan kan i viss mån även tillämpas på 3D-armering i allmänhet.

I arbetet har 3D-vävar tillverkats och impregnerats med epoxi. Dessa undersöktes sedan mycket detaljerat med datortomografi (CT) för att studera den interna armeringsarkitekturen efter impregnering. Analysen visade att trådarna var markant distorderade av kompressionen genom tjockleken vid impregneringen. Distortionen tillskrevs den relativt glesa väven, som inte hade förmåga att stödja vertikalväften tillräckligt, som därför snedställts och därmed också distorderar de överiga trådarna. En simuleringsmodell av det tillverkade materialets inre geometri utvecklades parallellt. Simuleringsmodellen var dock inte utformad för att inkludera de distortioner som uppstår i det fysiska materialet.

Det tillverkade materialet och dess motsvarande modell testades sedan i en dragprovsuppställning. Två olika tjocklekar av materialet tillverkades samt en motsvarande komposit med tvådimensionell (2D) armering. Resultaten visade att materialet med 2D-armering var styvare och starkare än de med 3D-armering, vilket tillskrevs den lägre vågigheten i 2D-armeringen. En skillnad i styvhet mellan de två 3D-vävarna hittades också, vilket tillskrevs den större kvot ytskikt i den tunnare väven som bidrar med ökad styvhet i varpriktningen då vertikalväften på ytan är i linje med den riktningen. Brottanalys av prover som testats i varpriktningen visade att initiala sprickor bildas i gränsskiktet mellan vertikalväft och matris, nära ytan på kompositen. För prover som testats i den horisontella väftriktningen hittades initiala sprickor genom vertikalväften vid ytorna. Båda dessa upptäckter bekräftades av resultaten från simuleringarna.

En applikation för kompositer med 3D-vävd armering undersöktes också, där den integrerades som ”noodle” i ett T-förband av kompositlaminat. Syftet här var att göra en 3D-armerad ”noodle”, med låg transversell termisk expansion som skulle minska restspänningarna i ”noodle” efter härdning. T-förband med konventionella ”noodles” och ”noodles” med 3D-vävd armering tillverkades och testades i en dragprovsuppställning. Resultaten visade att T-förband med konventionella ”noodles” hade högre hållfasthet, men också högre spridning, än de med 3D-armering. Den högre hållfastheten hos T-förband med konventionella ”noodles” tillskrevs deras bättre förmåga att anpassa sig till T-förbandets hålighet, medan passformen inte var lika bra för 3D-”noodles”. Den lägre spridningen i hållfasthet hos T-förband den med 3D-”noodles” antas komma från att de är mindre känsliga för defekter i ”noodle”.

The use of 3D-woven composite materials has shown promising results. Along with weight-efficient stiffness and strength, they have demonstrated encouraging out of plane properties, damage tolerance and energy absorption capabilities. The widespread adoption of 3D-woven composites in industry however, requires the development of efficient computational models that can capture the material behaviour. The following work proposes a framework for modelling the mechanical response of 3D-woven composites on the macroscale. This flexible and thermodynamically consistent framework, decomposes the stress and strain tensors into two main parts motivated by the material architecture. The first is governed by the material behaviour along the reinforcement directions while the second is driven by shear behaviours. This division allows for the straightforward addition and modification of various inelastic phenomena observed in 3D-woven composites. In order to demonstrate the applicability of the framework, focus is given to predicting the material response of a 3D glass fibre reinforced epoxy composite. Prominent non-linearities are noted under shear loading and loading along the horizontal weft yarns. The behaviour under tensile loading along the weft yarns is captured using a Norton style viscoelasticity model. The non-linear shear response is introduced using a crystal plasticity inspired approach. Specifically, viscoelasticity is driven on localised slip planes defined by the material architecture. The viscous parameters are calibrated against experimental results and off axis tensile tests are used to validate the model.

Three dimensional (3D) fibre-reinforced composites have shown weight efficient strength and stiffness characteristics as well as promising energy absorption capabilities. In the considered class of 3D-reinforcement, vertical and horizontal weft yarns interlace warp yarns. The through-thickness reinforcements suppress delamination and allow for stable and progressive damage growth in a quasi-ductile manner. With the ultimate goal of developing a homogenised computational model to predict how the material will deform and eventually fail under loading, this work proposes candidates for failure initiation criteria. The criteria are evaluated numerically for tensile, compressive and shear tests. The extension of the LaRC05 stress based failure criteria to this class of 3D-woven composites is one possibility. This however, presents a number of challenges which are discussed. These challenges are related to the relative high stiffness in all directions, which produce excessively high shear components when projected onto potential off-axis failure planes. To circumvent these challenges, strain based criteria inspired by LaRC05 are formulated. Results show that strain based failure predictions for the simulated load cases are qualitatively more reasonable.

Three dimensional (3D) fibre-reinforced composites have shown weight effi- cient strength and stiffness characteristics as well as promising energy absorp- tion capabilities. In the considered class of 3D-reinforcement, vertical and horizontal weft yarns interlace warp yarns. The through-thickness reinforce- ments suppress delamination and allow for stable and progressive damage growth in a quasi-ductile manner.

With the ultimate goal of developing a homogenised computational model to predict how the material will deform and eventually fail under loading, this work proposes candidates for failure initiation criteria. It is shown that the extension of the LaRC05 stress-based failure criteria for unidirectional lami- nated composites, to this class of 3D-reinforced composite presents a number of challenges and leads to erroneous predictions. Analysing a mesoscale rep- resentative volume element does however indicate, that loading the 3D fibre- reinforced composite produces relatively uniform strain fields. The average strain fields of each material constituent are well predicted by an equivalent homogeneous material response. Strain based criteria inspired by LaRC05 are therefore proposed. The criteria are evaluated numerically for tensile, compressive and shear tests. Results show that their predictions for the simulated load cases are qualitatively more reasonable.

A study of T-joints made of CFRP prepreg is presented where the joints contain either conventional uni-directional (UD) fillets or fillets with three-dimensional (3D) woven reinforcement in the joint cavity. Both pristine and impacted specimens are tested experimentally in a pull-off load case. The T-joints with UD fillets are stronger but also show greater spread in strength than T-joints with 3D fillets. The higher strength is attributed to the UD fillets' ability to deform transversely to their length direction and efficiently adapt to the T-joint cavity before curing. The 3D fillets do not admit the same level of transverse shape adaptability and if their cross sections do not fit the geometry of the T-joint cavity sufficiently well, local stress concentrations could emerge that reduce the strength of the T-joint. The UD fillets on the other hand are believed to be sensitive to manufacturing flaws causing the greater spread in strength. That in turn is attributed to a lack of crack-arresting capability in the UD fillet. The 3D fillets however have excellent crack-arresting properties due to their multidirectional fibre architecture. With a few exceptions the impact damages did not significantly affect the strength of the T-joints tested in this study.

Flat specimens made of carbon/epoxy composite material are manufactured by resin transfer moulding, using 3D-woven carbon fibre preforms with a grid of warp yarns interlaced with both horizontal and vertical wefts. The aim of the study is to bring more light to the coupling between the mechanical properties of the composite material and the internal fibre architecture of its 3D-woven reinforcement. Factors that are varied in the fibre architecture are the amount of fibres in the through-thickness reinforcement (vertical weft) and the warp's wavelength. Tensile, compressive, in-plane and out-of-plane shear and peel tests are performed for the mechanical characterisation. Tensile and compressive properties are found to decrease when the crimp of the warp yarns is increased, and even more so in compression. The in-plane shear strength is evaluated through use of a new test specimen, designed for the purpose. Results show that the strength is higher when the shear load is applied across the warp than across the weft, where the difference is attributed to varying fibre content in the two in-plane directions. The out-of-plane shear properties are compared through short beam shear tests and the inter-laminar shear strength (ILSS) is determined. It is shown that the ILSS increases with increasing yarn thickness in the vertical weft, which is intuitive. The peel strength is evaluated by the opening mode I interlaminar fracture toughness (G(Ic)) through double cantilever beam tests. It is shown that G(Ic) is greatly dependent on the amount of reinforcement in the vertical weft.

The differences between surface and interior parts of a composite material with 3D reinforcement is studied. It is done through experiments on composites with different thicknesses, and thereby varying ratios of interior and surface structure. The influence of the surface layers on the overall mechanical response under tensile loading is investigated and discussed. A finite element model of a meso-scale representative volume element of the material is built and used to relate measured properties to the internal reinforcement topology. Details of the reinforcement structure and the FE model are presented and discussed aiming to identify where in the composite material failure initiates, and how that is related to the surface/interior structure. It is shown that the properties of the surface and interior parts differ significantly. It is also found that failure in tensile loading initiates at the surfaces, that they are matrix dominated and can be well predicted by basic isotropic failure criteria.

Composite materials with 3-dimensional (3D) reinforcement were manufactured and corresponding simulation models were created in parallel. The used simulation approach has earlier been shown to produce close to authentic geometrical representation of the yarn architecture in 3D reinforcement. It is shown that although the as-woven reinforcement pattern can be modelled quite reliably, significant distortion from the nominal fiber arrangement might take place later in manufacturing, primarily related to compression during moulding. Such effects have earlier received significant attention for composites with 2-dimensional reinforcement but not as much for their 3D counterparts. The yarns in the real and the simulated materials are studied and compared, and some of the discrepancies and the mechanisms behind are discussed. The distortions are partly attributed to the relatively sparse weave that allows vertical yarns, in particular, to deviate from their original positions.