In developing Type V hydrogen tanks for energy storage and propulsion in commercial airliners, the key design criterion is maintaining tightness under cryogenic conditions. A concern is that anomalies in the laminate could cause microcracks which compromise the tightness. This study examined how resin flow, caused by an expanding mould during curing, creates a gradient in the local fibre volume fraction (FVF) across the laminate. The impact of resin system selection, autoclave cycle parameters, mandrel material, and fibre tension on the FVF gradient was investigated. Cylindrical specimens were manufactured in a process replicating automated fibre placement (AFP), with a piezoelectric sensor measuring contact pressure at the mandrel-laminate interface during the autoclave cycle as an indicator of resin flow and FVF spread. Significant correlations were found between resin flow and all parameters except fibre tension. The resin's viscosity profile and gelation characteristics had the most substantial impact on the FVF spread. Resins with high viscosity and short gelation times lowered the FVF spreads the most. Mandrel materials with lower thermal expansion coefficients (CTE) also reduced FVF gradients, though to a lesser extent. Adjusting the autoclave cycle to lower temperatures for longer periods resulted in the least significant reduction in FVF spread; however, this approach holds high potential for further improvement if more data on resin rheology and cure kinetics are available to optimise the autoclave cycle. These findings provide valuable insights for minimising FVF gradients in the design of carbon fibre-reinforced polymer (CFRP) tanks for liquid hydrogen.

Viscoelastic materials such as rubbers find ubiquitous use in engineering applications due to their outstanding ability to dampen vibrations and shocks. During a load cycle, the stress-strain curve of a viscoelastic material exhibits energy loss due to hysteresis. The structural response is damped because mechanical energy is dissipated and converted into heat, known as self-heating. The heating of the viscoelastic material results in degraded mechanical properties, and eventually, failure. This study focuses on the application of the viscoelastic material polyurethane in roller coaster wheels. Roller coaster wheels typically consist of a cast aluminium rim with a co-bonded polyurethane rubber bandage. The high forces and high speeds of a roller coaster ride lead to high self-heating in the rubber bandage. The high resulting temperatures in the bandage may cause failure of the bond line and of the rubber bandage itself. Replacement of the rubber bandage (re-treading) is a time-consuming and expensive process and ideally avoided. It is known that the design parameters of the wheel (diameter, width, bandage thickness, rubber hardness, etc.) influence the temperatures in the rubber bandage. At the same time, these design parameters have an influence on the damping capacity of the wheel, which is important for passenger comfort, noise and the longevity of the other mechanical components of the roller coaster vehicle. For example, the wheel with the maximum thermal performance (i.e., minimum temperature) will have minimum damping capacity, and vice-versa. To avoid this complication, this study compares the thermal performance of wheels with similar damping capacities. To evaluate the thermal performance of a wheel, a decoupled thermo-structural finite element analysis methodology is implemented. The method is validated using experimental measurements provided by the polyurethane manufacturer. Subsequently, a novel wheel design is proposed and compared to a regular roller coaster wheel. The 𝛽-ratio was introduced as the ratio of thermal to mechanical power. By comparing the finite element results with the experimental results, it was found that 𝛽 = 0.24 ensures that the numerical temperature matches the experimental temperature 𝑇num = 𝑇exp = 319.4 K. This value for the 𝛽-ratio was subsequently applied to finite element analyses of a regular roller coaster wheel and the novel wheel design. The standard wheel reached a temperature of 351.3 K on the surface and 376.7 K internally. The novel wheel design reached a maximum temperature of only 314.2 K. The maximum temperature of the novel wheel design is much lower than that of the standard wheel, because its structural finite element model is much stiffer, therefore providing less damping. Given the same loading cycle, the maximum temperature of the novel wheel design was found to be significantly lower than that of the standard roller coaster wheel. This indicates that the novel wheel will be more durable and have a longer lifetime. However, the novel wheel was also found to be significantly stiffer than the standard wheel. Toward the end of the report, various design iterations of the novel wheel design are considered which decrease this stiffness and improve damping.

The established theory of diagonal tension (i.e., post-buckling of skin-stiffened structures under shear load) for curved skin is based on experiments conducted on stiffened cylindrical shells at NACA in the 1940s-50s. However, the experiments were unable to capture all aspects of the phenomenon, leaving knowledge gaps in the understanding of the local distribution of the compressive loads generated on the stiffener, which were filled using the theory of diagonal tension for flat skins. Furthermore, due to the limitations of the time, the secondary bending effects on the stringers were not captured well.

This MSc thesis attempted to fill this gap by performing virtual tests using non-linear finite element modelling, which is enabled by the significant improvements in computational hardware of today. A virtual testing framework is developed based on the stiffened cylinders tested at NACA using the ABAQUS finite element modelling software. The skin and the stringers are modelled using continuum shell elements and the rivets are modelled using point-based fasteners. Furthermore, the generation of the models were parameterised using python scripting with provisions to model imperfections on the cylinders in the shape of the buckling eigenmodes. To validate the modelling framework, FE models of 8 cylinders were created and the results were compared with experimental data. Satisfactory correlation was found between the virtual testing results and the experimental results given the under-reported experimental data and the unknown imperfections in the original test

A custom post-processing method was developed that was able to capture the axial and bending loads that were generated in the stringers, which were used to develop a more detailed understanding of the load-transfer mechanism under diagonal tension. The results were also compared with the semi-empirical design method and the theory of diagonal tension, and significant discrepancies were found pertaining to the local variation of the axial loads and the secondary bending of the stiffeners.

This thesis explores finite element modelling of fibre-matrix debonding and frictional sliding in
Abaqus, with particular attention given to their influence on neighbouring fibres. It aims to improve model accuracy through advanced simulations and validation methods. A literature review underscores the need for robust finite element models in predicting composite material behaviour in fatigue. Theoretical equations for validating the finite element model are developed.
Single-fibre and multi-fibre models are utilised, with the former being used primarily for validation and the latter being used to simulate various realistic cases. Results demonstrate successful validation and provide insights into the effects of friction along the debond interface on stress concentrations in neighbouring fibres. Key findings indicate that reducing interfacial friction increases crack tip energy release rates, leading to stress fields that could potentially cause fractures in neighbouring fibres near the debond crack tip.

The implementation of morphing wings technology has the potential to significantly improve aircraft performance. However, due to the complex interplay between aerodynamics and structural integrity, it requires an aerostructural analysis. 3DExperience by Dassault Systèmes is a Computer-Aided Design (CAD) - based software with integrated structural and aerodynamic analysis tools. This software was employed to establish a methodology for conducting one-way aerostructural analysis of morphing trailing edge wings on a transonic transport aircraft. The methodology highlights the benefits of CAD-based modelling and the coupling between the aerodynamic and structural models. Nevertheless, the aerodynamic analysis results were affected by a node limit and the meshing options. Finally, a high-lift and an aerodynamically efficient test case were utilised to compare the impact of the trailing edge morphing to that of an unmorphed wing.

Due to the high aspect ratio and low induced drag of aircraft with strut-braced wings, they are being extensively studied due to their fuel-saving potential. This thesis aims to expand the fundamental knowledge in the design of aircraft with Strut Braced Wings (SBW) by achieving the following objectives.

The objectives of this thesis are twofold, with their primary focus on evaluating the effect of changing the typical design variables for an SBW, such as the wing span, root chord, spanwise strut attachment location, wing taper ratio, wing sweep and engine location on the aerodynamics, weights and performance of an SBW. The first objective was to evaluate the significance of including the propeller slipstream effects in the preliminary stage design optimisation of a low-speed, short-range SBW. This aligned with the hypothesis that the performance of an SBW could be enhanced by using swirl recovery. The second objective of this thesis was to investigate the sensitivity of various design variables to the aircraft’s fuel burn and other performance metrics at the optimum SBW design.

A Design of Experiments (DOE) approach was used to explore the design space involving the typical influential SBW design parameters. The geometry and mesh files were created using OpenVSP for every DOE point. This was followed by the aerodynamic analysis in a panel method-based software called Flightstream, which could capture the relevant aerodynamic flow phenomena with reasonable accuracy. The aerodynamic analysis was performed twice- first without considering slipstream effects and second by simulating them. Regression-based analytical equations, particularly designed for the weight estimation of the wing and strut, were used from the literature. Empirical equations from FLOPS were used for the rest of the aircraft components. The performance of the SBW was calculated iteratively using the Breguet range equation along with a few modifications. All the SBW designs were constrained by a maximum wing loading criterion.

From two sets of the DOE results (with and without propeller effects), it was observed that the two optimum SBW designs were identical in terms of external geometry and performance. A deeper investigation revealed that the variation in the spanwise engine positioning (to maximise the swirl recovery) resulted in a marginal change of the induced drag (less than two drag counts). It was concluded that the propeller slipstream effects could be excluded from the preliminary stage, design optimisation of a propeller-powered, short-range SBW to reduce computational expense. However, the results should be treated with a pinch of salt due to the limitations of panel methods. Moreover, the SBW was optimised only for cruise, not for other flight phases such as take-off and climb, which may benefit from swirl recovery.

Finally, a sensitivity analysis was performed to evaluate the trends in the performance metrics, such as fuel burn, lift-to-drag ratio, wing loading and maximum take-off weight when subjected to variations in the design variables at the optimum. The impact of the fuel burn was quantified, while the other performance metrics were qualitatively answered. The research findings revealed that certain design variables had a greater influence on fuel burn than others when varied by ±10% w.r.t their optimum value...

The concept of lightweight design is driving future aircraft to exploit all the available strength of materials and further reduce the weight of an aircraft leading to lower fuel consumption and more sustainable aviation. Current designs introduce rivets and bolts to join the structure and simultaneously create holes in the pristine material, which is not wanted due to the local stress concentration. If the design wants to exploit all the strength of the material, stress concentration should be avoided using suitable joining technologies that don't require holes in the structure. This is possible with the adhesive bonding technology. However, many factors, such as the effects of manufacturing and impact-induced damage are still not fully understood. Thus, the damage tolerance of the design cannot be guaranteed. This results in conservatory safety factors being prescribed in the design process, which possibly reduces the exploitation of all available material strength.

The occurrence of barely visible impact damage (BVID) in aircraft composite structures, especially in adhesive joints, is a serious issue that can jeopardise an aircraft's structural safety during operation. However, virtual testing and numerous numerical approaches allow high-fidelity simulation of damage initiation and propagation in adhesives and composite adherends to predict the residual strength of the bonding accurately. Although the development is fast, accurate impact simulation is still computationally very expensive and thus not applicable in industrial cases where behaviour needs to be analysed to assess the structures' design, maintenance and repair.

This thesis investigated two topics that allow robust and accurate simulation methods to assess the effect of manufacturing and impact damage on the residual strength of bonded joints. The first topic is related to the development of the composite material damage model, where the multiscale material model is created by the use of a representative volume element (RVE). The second topic of simulation methodology is the development of the quasi-static simulation approach for the damage tolerance assessment of bonded joints. Modelling approaches to simulate residual strength are investigated through the finite element modelling of three groups (i) pristine, (ii) artificially damaged during manufacturing, and (iii) impacted bonded joints. In the numerical models, an approach based on the observation of the fracture surface of single-lap joints in different geometry and layup configuration is proposed in which the damage assessment focuses on the bondline area up to the first 0° ply. For the modelling of the damage in joints, two modelling techniques were studied, first removing elements in the damaged area and second detachment of the elements. Utilising those techniques, simplified approaches to model damage resulting from the impact were studied, with the modelling of impact damage as a hole, where all through-thickness elements are deleted and as delamination, where interlaminar cohesive zone elements are deleted.

The developed multiscale material model allowed for an accurate representation of failure modes occurring in the composite material by the characterisation of elastic, damage and plasticity parameters of the fibre and matrix constituents in the homogenisation and inverse characterisation process. The results showed that the deletion of the elements can be used to represent defects and damage of different kinds in the composite bonded joints, both in the adhesive and adherend. The comparison of different representations of the impact damage in the single lap joint configuration revealed that the best prediction in terms of the ultimate load, failure mode and size of the model is obtained with the simplified representation as a single delamination positioned before the first 0° ply in the layup and in the studied adherend configuration that was between the first (45°) and second (0°) ply in the layup. The study ends with the conclusion that in different geometries of single lap joints and different damage types studied, a numerical analysis should focus on the region of the overlap edge and in the thickness direction from the bondline up to the first 0° ply in the lay-up. The results and numerical method developed during this thesis create a base for the further investigation of a variety of impact cases on bonded joints.

A Helically wrapped structure is a hollow pipe structure that is made from a metal strip and the overlapping interfaces are adhesively joined. It is yet unknown how to determine the quasi-static strength and predict the fatigue life of this structure. A method was developed to determine whether either the adherend or the adhesive is critical to determine the strength and fatigue life. A machine was made to create the helically wrapped structure and tested for validation. It was found that the adherend is critical instead of the adhesive.

The development of novel materials has been of great interest to the aerospace industry in the pursuit of efficiency and cost reduction. Fibre-steered variable stiffness laminates have promising characteristics due to the ability to better tailor designs for expected loading, as the fibre orientation can be varied within a layer. This thesis was created to develop a fibre placement planning algorithm for fibre-steered layers on moderately curved open shells for a given fibre angle distribution.

The fibre-steered layers are to be produced by Automated Fibre Placement (AFP) without any course overlaps, which means that several manufacturing defects have to be considered. Gaps between courses are inevitable due to course shifting in combination with the prescribed fibre paths, and tow cuts run the risk of tow straightening due to loss of steering control of the AFP roller head. Furthermore, steering limits have to be observed to prevent tow buckling.

The thesis objective was approached by dividing it in three parts. Firstly, the geometry of the mould and the course paths was defined parametrically in the form of NURBS surfaces and curves. These have been widely used in Computer Aided Modelling (CAM) and have extensive support in terms of existing algorithms and documented work. Additionally, performing most operations in the parametric space allows for greater accuracy compared to general finite element methods.

Secondly, the fibre angle distribution was converted into a fibre layer. This was initiated by generating initial course centre lines, which were then fit onto the parametric surface. Representations of tow courses were made by geodesically offsetting the course centre lines in the binormal direction. A curve-curve intersection algorithm was then used to accurately detect tow overlaps, which were handled according to the tow placement strategy. Two tow placement strategies have been developed, one for general use and one to specifically prevent tow straightening. In addition, curvature of the course centre line was registered to check for tow buckling, and the minimum cut length was enforced by evaluating the length of tow segments.

Lastly, fitness criteria for fibre-steered layers were defined. Layers were evaluated for gap percentage, fibre angle error and number of cuts. These can be used to objectively grade layers against each other, or to assess the effects of fibre steering against conventional laminates.

Over the past decades, fibre-reinforced composite structures have been increasingly adopted into mainstream manufacturing over conventional metals owing to their higher performance attributes, such as superior strength-to-weight ratios. Composite structures offer the unique advantage of tailoring reinforcements in correspondence with design load cases. This allows for more efficient and better performing structures.
Traditionally for composite laminate structures, the design space and freedom were vastly influenced by the accumulated experience in design and certification, as well as available flexibility in manufacturing processes; the available degree of freedom in aligning fibres dictated the design freedom available. With the advent of advanced processes such as automated fibre placement (AFP), reinforced tows can now be placed with more freedom and accuracy; even allowing for tows to be steered during deposition. These are called variable stiffness composite laminates. Fibre directionality within these laminates can be tailored to achieve an optimal load redistribution, thereby increasing its structural performance.
Novel additive manufacturing such as Fused Deposition Modelling (FDM) allow for more complex geometries to be manufacturable in a variety of materials. Such a process can allow more design space and freedom in existing optimization frameworks for designing variable stiffness laminates. Many reinforced thermoplastic materials can be processed using FDM. Short fibre reinforced materials are very readily available and can be used on all commercially available FDM platforms with minimal changes.
This research culminates the three key aspects in engineering – design, material, and process. First, a suitable design framework is chosen, which in combination with added the design freedom by virtue of a novel process such as FDM, is used to design laminates optimized for buckling performance. Additional design freedom is afforded to the optimization framework by means of relaxing the manufacturability constraint – which restricts the maximum allowable curvature of each individual path within the laminate. Secondly, these laminates are manufactured using short fibre reinforced thermoplastic material, for which the shear-induced alignment of the material is analyzed to predict the effective mechanical properties under the parameters used for printing the laminates. Lastly, to validate the effect of additional design space on the effective performance of these laminates, a suitable experimental protocol is devised and used. For the optimized laminates, two cases for maximum allowable steering curvature are considered – one low and one high, and an effective quasi-isotropic laminate is used as a benchmark for comparison. Finally, all the laminates are tested under compression and analyzed. The increase in buckling performance of optimized laminates corresponding to increase in allowable steering is verified, as well as insights are drawn from the processing and experimentation to suggest future recommendations.

The collapse of composite stiffened panels in compression is complex and involves the interactions of many different failure modes. However, when the panel is allowed to enter the post-buckling regime, skin-stringer separation due to the interaction of the post-buckling deformations with the skin-stringer interface is often the critical failure mode. The accurate prediction of skin-stringer separation in these types of panels is therefore crucial for their design. Due to the high computational costs associated with modelling damage in a large multistringer panel and the manufacturing and testing costs associated with testing such a panel, single stringer specimens that accurately represent the behaviour of critical regions in these panels have been proposed in literature. In this thesis, transversely loaded single stringer specimens are designed and verified and are used to predict skin-stringer separation in the critical regions of a specific multistringer panel. A specific composite stiffened panel is considered that contains a skin of a tape material and four co-cured stringers of a fabric material. At the intersections of the skin and stringers, resin-rich noodle regions are created. A buckling analysis is first performed on the panel to obtain its post-buckling deformations. Based on these, two types of regions where skin-stringer separation may occur are identified; a mode I dominated skin-stringer separation at the minima of the buckling waves and a combined mode II and mode III skin-stringer separation at the inflection points of the buckling waves. A seven point bending specimen is designed based on the panel deformations near the former critical region, while a four point twisting specimen is designed based on the deformations near the latter region. Finally, detailed damage models are set up for the single stringer specimens to predict skin-stringer separation and the models are verified with the multistringer panel. The analyses showed that a mode I dominated skin-stringer separation is critical for this panel. The detailed analysis of the four point twisting specimen showed that no damage occurred at the inflection point. To predict the skin-stringer separation at the minimum, a verified seven point bending specimen has successfully been obtained. To do so, it was important to accurately simulate the deformation of the skin at the stringer flange edge, since this proved to be driving for the initiation and propagation of skin-stringer separation. Next to that, a trilinear cohesive law was used to capture the R-curve effect in the tape/fabric interfaces. Finally, a first assessment of the noodle region behaviour showed that damage in this region will probably happen at the location of maximum bending. Recommendations for future research have been provided.

Disassembly of fusion bonded joints aided by induction heating was investigated. Single-lap shear experiments were performed while heating the joint with an induction coil to research the effect on force required and damage inflicted during disassembly. Co-consolidated CF/PEEK samples with and without a metal mesh susceptor and ultrasonically welded samples were tested and compared. An induction heating model was built to facilitate experimental design and to help analyse the effect of the susceptor on the heating process. Induction heating appeared successful in lowering the disassembly force. Large reductions were achieved, but this came at the cost of thermal damage in the disassembled adherends. For a reduction to 37% of the original strength, no thermal damage was inflicted during disassembly, except for one outlier. A susceptor facilitated disassembly for co-consolidated joints, while PEEK energy director remnants hindered heat development in ultrasonically welded joints. Further research is required to develop the method.

The demand for efficient lightweight structures grows rapidly in the aerospace sector. Topology optimization, introduced in the late 80s, is a method capable of producing such structures and has been mainly studied for isotropic materials. On the other hand, the performance of composite structures, which are already widely employed in the aerospace industry, can be now improved due to the latest advances in manufacturing. Automated Fiber Placement has paved the way for further exploiting the capabilities of composites and led to the introduction of an optimization method, namely three-step variable stiffness design, that alters the fiber orientation within each ply, leading to a variable stiffness laminate. In this thesis, a staggered optimization method that simultaneously improves the material and the structural performance of balanced and symmetric laminated composite structures under planar loading is developed by coupling the two aforementioned methods. A finite element code is developed based on the equations for a static and linearly elastic problem. For the topology optimization problem, the nodal density values are used as design parameters and the Solid Isotropic Material with Penalization approach is chosen, along with a density filtering. The first step of the variable stiffness design method is implemented, using the nodal lamination parameters as design variables. The coupling of the two optimization techniques is performed through a staggered optimization, where both techniques are implemented successively, at each iteration. A gradient-based scheme is used for both topology optimization and variable stiffness design and the steepest descent method is implemented for the design update. The sensitivity analysis is performed based on the continuous adjoint method. Three different examples are demonstrated in this thesis; the case of a flat composite plate, a lug and an aircraft chair bracket. The whole optimization procedure, including pre- and post-processing, is carried out using an algorithm developed in MATLAB. The meshes are externally created and imported in the code. Convergence studies and a comparison with indicative examples from the literature are performed in order to verify the obtained results. A parametric analysis of the plate studies the effects of the different topology optimization parameters. Finally, comparative analyses are executed for the three aforementioned designs investigating both the structural and the computational efficiency of the results obtained with each of the three optimization methods, i.e. topology optimization, variable stiffness design and staggered optimization.

L-AFP is able to place composite tapes in very precise directions, thereby allowing tailored stiffness designs. The laser is used to heat the thermoplastic matrix material to temperatures above its melting point. CF/PEEK is a popular choice for aerospace grade structures due to the high structural performance and high glass transition and melting temperatures. Current L-AFP research is driven by the desire to produce OOA laminates with a performance comparable to autoclave level. This will remove manual, labor intensive, and energy consuming steps, leaving a repeatable and automated process that is more predictable. A key phase for the performance of a laminate during L-AFP is the development of intimate contact. Recent research showed the effects of thermal deconsolidation due to rapid laser heating and questioned the current practice of characterizing the surface of the composite tape material for the intimate contact development models: instead of characterizing the pristine tape surface, the laser deconsolidated surface should be characterized. Additionally, another dominant physical mechanism, driving the intimate contact development was proposed based on the observations of thermal deconsolidation. These ideas are novel and require more extensive understanding of what thermal deconsolidation due to rapid laser heating is, how it affects intimate contact development. The effects of rapid laser deconsolidation were studied by performing (1) a rapid laser deconsolidation experiment where CF/PEEK samples were prepared and (2) an intimate contact development experiment where the specimens were compressed under constant pressure while being subjected to a temperature profile. Three degrees of deconsolidation were manufactured with the laser set-up: (1) zero deconsolidation by using the as-received tape directly as specimens, (2) slightly laser deconsolidated tapes which experienced maximum temperatures between the glass transition and melting region, and (3) highly laser deconsolidated tapes that experienced temperatures above the melting point. All samples were characterized on void content, roughness, and waviness; all of which significantly increased which showed that the state of the tape right before the nip-point during L-AFP is significantly different than the pristine state of the CF/PEEK tape. The intimate contact development experiment was carried out with pressure levels of 10, 50, 100, and 300 kPa reaching maximum temperatures of 363°C. Decreasing the degree of deconsolidation or increasing the pressure resulted in better consolidation. At 10 kPa, most of the characterization showed very little difference from a pristine tape, but a significant amount of DEIC was developed with 36% for the highly laser deconsolidated tape to 46% for the as-received degree of deconsolidation. At the highest pressure of 300 kPa, no significant difference between the degrees of deconsolidation was observed as the average DEIC values are between 85 to 87%. Additional intimate contact development temperature settings below melt showed that no intimate contact was developed below the melting temperature. During the glass transition, a significant compaction occurred which eliminated most of the void content. Between the glass transition and melting temperature, however, void content, thickness, and roughness all remained constant. At the melting point, a second thickness compaction occurred together with the development of DEIC and decrease of roughness which appeared to also relates with resin percolation.

Due to an ever-growing demand for sustainable transportation, there is an increased need for the development of hydrogen as an energy carrier. Based on mass, hydrogen contains about three times as much energy as conventional gasoline. However, the volumetric energy density, which is the amount of energy per volume unity, is much lower for hydrogen. Therefore, it needs to be stored either at cryogenic temperatures or at hyperbaric pressure. The most common nominal working pressures (NWP) for hydrogen are either 350 or 700 bar. Because of these excessive pressures, the composite pressure vessels (CPVs) require to become thick-walled. This makes the out-of-plane stress components significant which requires calculation models that consider these effects.

This thesis evaluates, whether the use of tow-preg material, could reduce the amount of material necessary to attain a predefined burst pressure. Currently, the production of CPVs is mainly executed using wet winding. The use of this process results in varying material properties due to differences in fibre volume fraction. Also, the friction coefficient is minimal because the resin has a low tackiness. This friction is required for the production of winding paths, which deviate from the traditional geodesic angle. Alternatively, tow-preg material introduces constant material properties and possibly an elevated friction coefficient. This friction coefficient provides an opportunity to vary the winding angle through the thickness, by which the material strength can be made more uniform in every lamina. However, the challenge during the design and production of CPVs is that several effects are present which alter the mechanical response such as varying stiffness on the dome-cylinder interface, non-geodesic winding angle on the dome and resin-rich areas. This thesis will attempt to describe the mechanical behaviour of the CPV, by including several phenomena in numerical models and to minimise the material usage.

The structural properties of a composite laminate can be tailored by allowing the fibres to vary their orientation within the individual plies. This class of composites is termed variable stiffness and they are able to enhance the structural performance over traditional laminates by redistributing the internal load of the structure. Therefore, it is of interest to pinpoint the structural properties that may benefit most from this concept and the extent thereof. To that end, a multi-objective optimisation of a variable stiffness cylinder was performed which identifies the Pareto front for objectives such as buckling, natural frequency, stiffness and strength. Manufacturing considerations were also incorporated by accounting for the allowable tow curvature and the resulting overlaps during the process. The stiffness variability was formed by means of a Lagrange polynomial, the non-linearity of which is determined by a specified amount of control points. An investigation on the suitable degree of non-linearity of the functional representation of the cylinder was thus carried out. In order to complete the optimisation in a practical time frame, machine learning algorithms were utilised to act as surrogate models and perform predictions on the optimisation objectives. Different algorithms were implemented for that task, namely Radial Basis Function, Kriging and Artificial Neural Network models. The Artificial Neural Network surrogate model outperformed the others when the dimensionality of the problem was high, although the opposite was observed for low dimensional problems. Cylinders with and without cut-outs were optimised separately and compared with their optimised constant stiffness counterparts. It is found that the performance gains for the cylinder without cut-outs are not significant when the additional mass due to the overlaps is accounted for. On the other hand, the cut-out cylinder greatly benefits from the stiffness variation for multiple objectives. Additional structural improvements were witnessed for higher degrees of non-linearity of the utilised Lagrange polynomial.

Variable Angle Tow designs have shown to improve structural performance by providing a better stiffness distribution. However, additional manufacturing constraints are involved compared to straight fibre laminates: mainly the maximum fibre steering curvature that can be achieved. Moreover, if a gap free laminate is constructed, overlaps will be formed depending on the relative fibre orientation when the continuous tows are bent to follow a reference path in Automated Fibre Placement and a one-sided irregular thickness profile is created. To counter the latter effect and still obtain uniform thickness laminates, a cut and restart strategy of the tows is often used in industry, leaving the final product with small gap areas. However, previous research has hinted both numerically and experimentally on further buckling improvements of variable stiffness laminates incorporating overlaps, resulting in a variable thickness profile. To investigate these possible benefits and their extent, the thickness build-up and overlap locations are specially considered in this proposed framework on Variable Angle Tows. In first instance, a virtual manufacturing surrogate of the laminate is produced to represent the discrete thickness profile due to the overlaps. This is performed by plotting each tow graphically and subsequently retracing the total thickness based on the superimposed opacity over the laminate. The tow paths are obtained from the interpolated fibre orientation, which simultaneously incorporates the steering limit of the manufacturing process. The surrogate information is then compared to a smeared approximation of the thickness build-up, based solely on the steering angle. The linear buckling simulation is performed by means of a semi-analytical model on a plate formulation, where the one-sided thickness profile variation is modelled by incorporating an offset from the geometrical varying midplane to a common one, around which the laminate stiffnesses are calculated. These adapted stiffnesses are then used in solving the neutral equilibrium buckling problem, where the displacements are approximated by means of the Ritz method, with a set of Legendre polynomial shape functions and numerical integration of the stiffness matrices. The semi-analytical smeared model correlates to within ± 5% of the discrete thickness Finite Element Model for a range of geometries, loading and boundary conditions. This verifies the thickness approximation for linear buckling of small tow widths compared to the laminate's dimensions, yet was inconclusive for larger tow width ratios. Finally, laminates with a varying thickness profile are optimised alongside the variable stiffness for different cases and compared to uniform thickness counterparts at isomass. The results show small improvements in the symmetric problem, but a higher gain is obtained for unsymmetrical conditions with an unrestricted layup sequence, as the thickness increase can be highly tailored together with the variable fibre orientation to locations requiring higher stiffness.

Near free edges of fibre reinforced composite laminates modelled with homogeneous layers, the presence of high gradient interlaminar stresses has been found theoretically. Over a span of nearly 50 years, various methods, including numerical and analytical have been employed for analysis of the high gradient stresses which sometimes exist to an extent of a singularity. The analysis of such free edge stresses is found to be computationally expensive. However, since the interlaminar stresses have been found to play a role in initiating delamination in composite laminates, it becomes imperative to get an estimate of the interlaminar stresses near the free edge to be able to predict delamination initiation through suitable criteria.

The abrupt material discontinuity between dissimilar layers in a homogeneously modelled laminate is believed to be a reason for the appearance of high gradients or singularities theoretically. The mitigation of material discontinuities at such interfaces could be done by modelling the laminate heterogeneously so that the interface is modelled by matrix material and hence, a discontinuity of material could be avoided. This idea of modelling the layers of laminate with explicit definition of fibre and matrix materials has inspired the thesis project with a view to investigate the difference in free edge stresses between the two models and to further draw a correlation between the stresses from the two models to predict delamination initiation.

In previously published literature, the use of average interlaminar stresses up to a certain characteristic distance from the free edge through criterion for delamination initiation prediction is reported. Further it is also found that a single averaging distance can be proposed for a combination of laminates made of the same material. This encourages the idea of determination of averaging distance for a material and find its application on a laminate susceptible to delamination initiation. In the course of the current work, a correlation is proposed between the stresses of homogeneous and heterogeneous model [0/90]s cross-ply laminate of equivalent stiffness made of T300/934 material to determine the averaging distance and apply the determined averaging distance on [±25/90]s laminate to predict delamination initiation due to high gradient free edge stresses.

An averaging distance of 0.125 mm is determined for [±25/90]s laminate made of T300/934 material to find average stresses for incorporation into criterion for delamination initiation. It is found that the determined averaging distance predicts delamination initiation successfully for experimentally reported delamination initiation strain range for the [±25/90]s laminate. Further, the convergence of average of high gradient interlaminar stress is found to be computationally more efficient than the convergence of high gradient interlaminar stress themselves. This indicates that use of average of high gradient interlaminar stresses is an efficient means for predicting free edge stress induced delamination initiation.