Modern engineering design workflows are inherently iterative, with each iteration typically requiring new sets of computationally expensive evaluations while previously generated data is discarded. This thesis proposes a methodology for reusing data from previous design iterations within a multi-fidelity (MF) Gaussian Process (GP) surrogate modelling framework. The data from a previous design iteration, referred to as the source, is fused with a limited number current design evaluations, referred to as the target, using Multi-Fidelity Co-Kriging (MFCK). The source data serves as the low-fidelity input and the target data as the high-fidelity input within the MFCK. The squared Pearson correlation coefficient is adopted as an a priori metric for assessing LF source quality, while a post-hoc discrepancy share diagnostic is introduced to evaluate the contribution of the LF source from the trained model’s internal perspective.
The methodology is validated on two structural design problems of increasing complexity: a three-dimensional cantilever beam problem with a readily available analytical solution and a six-dimensional stiffened cylindrical shell problem, evaluated using finite element analysis. The findings suggest that the effectiveness of the approach depends primarily on the quality of the source data: fusion of well-correlated designs can substantially reduce the number of required target evaluations while maintaining comparable precision and improved robustness. Poorly correlated designs, on the other hand, can degrade the performance below the singlesource counterpart.
... Read more

Crashworthiness optimisation is a challenging problem, as a single simulation is expensive, the response is non-smooth, and the design space can contain continuous, categorical and conditional variables, the latter being variables that are only active under specific design choices. The current standard in structural optimisation is the Gaussian Process (GP); however, it assumes a smooth response and a fixed design space, and therefore struggles with exactly these properties. Tree-based models handle mixed-discrete problems natively, but they lack a reliable uncertainty metric to balance exploration and exploitation. This thesis sets out to determine if tree-based surrogate models can be used to efficiently identify and exploit feasible regions within a mixed-discrete design space, given a limited computational budget.

In order to do so, a constrained version of BOinG is created, which combines a random forest global model with a local Gaussian process. Three variants are proposed to differentiate between the routing and the trust region construction mechanism. BOinG-cEI acts as the baseline and adds a constrained Expected Improvement acquisition function, BOinG-inc replaces the global search with a Thompson sampling router centred on the incumbent, and FuRBO-Tree builds its local region from inspectors ranked by a gradient-boosted tree model. These methods are benchmarked alongside the native GP, Sequential Model-based Algorithm Configuration (SMAC) and Tree-structured Parzen Estimator (TPE) models.

On the synthetic and engineering benchmarks, the hybrid variants outperform the native models. FuRBO-Tree performs the best on the highly constrained problems. In the constrained 15D Ackley, it spends 75.9% of its budget inside the feasible region, against 24.0% for BOinG-cEI and 12.5% for the GP. On the conditional benchmarks, neither router dominates. The cEI router locks on to a single branch, while the Thompson router keeps exploring all of them. Only in the crash simulation does the GP show the most reliable performance, closely followed by BOinG-cEI, as the optimum lies on the constraint boundary, where the tree-based methods settle inside the feasible region. Of all the tree-based methods, BOinG-cEI performs most robustly, and can be considered an optimiser for constrained, mixed-discrete design problems. The most limiting factors remain the lack of an uncertainty measure driven by the model itself, and a trust region that is built on the objective alone.
... Read more

This work presents the analysis of on-edge impact loading of carbon fibre reinforced polymer(CFRP) material. On-egdge impact are rarely researched and they often consume a lot of time and resources to be tested. This work presents a possible easier set-up to analyze the response of the CFRP using a conventional droptower. First, quasi-static tests are performed to gain initial mechanical understandings of the material, including the energy absorbed by the specimen. Then efforts are dedicated to impact load the specimen at same energy using the droptower. Then the impact energies are increased to further investigate the effects of the strain-rate loading for on-edge loading. The damages are compared through the means of microscopy and ultrasonic C-Scan to assess the visible damages and the extent of the damages. While it was possible to perform reliable energy analysis and the damages were compared in a detailed manner per strain-rate used, the force and impulse data of the specimen were not able to evaluated to full extent. The nature of the droptower, measuring the force experienced by the impactor and the vibration caused due to the impact hinders the data significantly. For future, it is recommended to present a set-up which can measure what the specimen experiences directly, rather than relying on the information from the droptower to gain information about the specimen.
... Read more

Thermoplastic composites are increasingly adopted in the aviation industry due to their superior damage tolerance, shorter processing times, and recyclability compared with brittle thermosets. Complex geometries with curved radii, which have high energy absorption capability, are prone to unfolding failure under bending and have been studied only at a single corner using analytical methods and expensive numerical models. This thesis establishes an integrated testing and modelling framework to characterise the structural response of a curved, omega-shaped CF/LM-PAEK thermoplastic structure undergoing simultaneous multi-corner unfolding under bending, up to large deformations.

A building-block approach was adopted. At the coupon level, laminate quality assessments and mechanical tests were conducted on specimens extracted from the curved structure to establish baseline material quality and determine its in-plane material properties. These results showed good agreement with literature values and were used to build a validated LAW25/CRASURV material card with plasticity modelling to capture the matrix-dominated non-linear response.

At the element level, three-point bending tests performed on the curved omega structures showed excellent repeatability, both quantitatively and qualitatively, with consistent cross-sectional deformation, strain field, and load–displacement response across independent instrumentation. A shell-based finite element model, developed using the validated material card in Altair HyperMesh, closely captured this non-linear response at both global and local levels, while identifying limitations in modelling delamination. Sensitivity studies were also carried out to identify the governing parameters. These findings provide an experimental-numerical framework to build robust, validated models that predict the structural response at low computational cost during composite design iterations, and pave the way for efficient modelling using advanced elements to capture the complete structural behaviour. ... Read more

This report details the design, engineering, and commercial feasibility of the High-speed Unmanned General-purpose Observatory (HUGO), a 50 kg modular transonic flying laboratory. HUGO aims to address the gap between scale wind-tunnel testing and full-scale flight validation. Operating primarily at Mach 0.75, HUGO bypasses the wall-interference limitations of conventional transonic wind tunnels and the severe safety risks associated with manned flight testing. To meet its experimental objectives, the finalized airframe features a high-wing, T-tail configuration with a modular wing, canard and empennage ports and retractable steel landing gear. A defining feature of the platform is its customizable "Planform Envelope," which allows clients to interchange main wing configurations with aspect ratios ranging from 5 to 27, as well as various sweeps, chords and thickness profiles. To cost-effectively produce in low volumes, while maximizing structural efficiency, the airframe utilizes a composite monocoque structure capable of safely withstanding ultimate load factors of up to 9g. Propulsion is provided by twin fuselage-mounted JetCat P100-RX-BL micro-turbines, enabling HUGO to execute a 30-minute nominal mission that includes a five-minute Mach 0.75 test segment at FL270. Because high-fidelity data collection is the core value proposition of a flying laboratory, the aircraft is equipped with a comprehensive payload suite, including dual LIDAR wing deflection sensors, 16 surface-mounted strain gauges, and 6-axis load cells, to isolate and measure aerodynamic forces. Commercially, HUGO targets a €14.2 million Serviceable Obtainable Market within the Dutch wind tunnel testing ecosystem and is projected to operate profitably at a rate of €25,000 per flight day. Operations will be based out of Den Helder Airport under the EASA SC-Light-UAS High-risk classification, with regulatory compliance maintained through the simultaneous acquisition of a Type Certificate and Design Organisation Approval (DOA), which allows internal certification of modifications made to the aircraft on a per experiment basis. ... Read more

Accurate aero-structural modelling of leading-edge inflatable kites is essential for scaling up airborne wind energy systems. Yet, existing approaches make simplifications on either the wing or the bridle line system. This paper presents a fast finite element modelling framework that integrates the wing structure and bridle lines into a single model. The finite element framework captures material and geometric non-linearities using the Newton–Raphson scheme with a co-rotational formulation. The bridle line system is represented using non-compressive springs, which are also used to represent pulleys. Inflatable leading edge and strut tubes are modelled as Timoshenko beam elements, with beam properties tuned iteratively to match experimentally fitted inflatable beam equations. The canopy is represented using non-compressive spring elements, allowing for the resolving of billowing effects. The framework is applied to the V3 kite, and validation is performed against stereoscopic photogrammetry measurements obtained from static load tests under varying internal pressures and external loads. The results demonstrate good agreement with measured kite shapes for most cases, confirming the model’s ability to capture the dominant structural behaviour of leading-edge inflatable kites. The proposed framework provides a basis for aero-structural analysis in kite design, and enables future coupling with aerodynamic models for steady and quasi-steady kite simulations. ... Read more

Liquid hydrogen-powered aircraft are a promising candidate for reducing aviation’s climate impact, but safe integration requires crashworthiness considerations already at conceptual design. This work presents a parametric, knowledge-based methodology to design and analyse a fuselage-located cryogenic tank supported by a belly crash structure under a vertical drop test. The methodology is implemented in a parametric application that automatically generates geometry from readily adjustable inputs, enabling quick generation of a wide variety of configurations. A baseline configuration, consisting of 12 sub-tank X-beams per bay configured in six X-crosses, a stiff crossbeam design, and distributed Kevlar tank suspension ropes, exhibited favourable deformation kinematics. A design-of-experiments study was conducted and combined with a surrogate-model-based sensitivity analysis to identify the dominant design parameters governing the crash response. Results show that the coupled frame and crossbeam thickness, together with the crash coefficient, are the most influential. Including a crash structure substantially increases the required tank and fuselage length, and the associated mass penalty is dominated by the added tank and fuselage structure rather than by the crash structure mass. The developed model provides a basis for further research on structural mass optimisation and for providing data for a surrogate integration into a conceptual aircraft design tool. ... Read more

Growing competitiveness in the launcher market is driving metallic propellant tanks toward tighter mass limits, increasing the need for reliable buckling prediction under combined axial compression and internal pressurization. Thin-walled cylinders are imperfection-sensitive, and current practice relies on conservative knockdown factors derived mainly from unpressurized tests. This work evaluates alternative nonlinear imperfection modeling approaches for pressurized launcher tank segments, with an emphasis on reliability and welding-induced imperfection signatures.

Five imperfection-sensitive strategies were assessed using NASA Shell Buckling Knockdown Factor Project cylinders at 0, 2, and 4 bar, including SPLA, MPLA, EIA, MGI, and a distributed-force perturbation approach (DFPA). Reliability was quantified using the coefficient of variation, Kendall’s W, and the intraclass correlation coefficient.

Results show that pressurization reduces imperfection sensitivity by increasing geometric stiffness. Distributed and multiple-perturbation methods were the most stable across pressures, whereas localized approaches exhibited strong pressure sensitivity. Accurate pressurized buckling prediction, therefore, requires pressure-aware imperfection mechanisms.
... Read more

The aviation sector is responsible for approximately 2.5% of global CO2 emissions. While the CO2 emissions reduced at a compound annual growth rate (CAGR) of 2.8% between 2000 and 2019, the passenger demand is projected to more than double by 2044 as compared to 2024 at a CAGR of about 4.2%, outpacing the reduction in emissions. Unconventional aircraft (for example blended wing body aircraft) are projected to reduce emissions by more than 20% as compared to widebody aircraft of comparable mission and size, making them instrumental for reducing aviation emissions. However, these novel configurations also introduce new crashworthiness challenges that must be addressed early in the design process to prevent cost overruns and delays at later stages of the design process.

Historically, preliminary crashworthiness analysis has relied on linkage-kinematic joint representation of the aircraft, with characteristic curves for the kinematic joints derived using experimental or numerical campaigns. Even the most recent multi-body and FEM-kinematic joint hybrid models still depend on characteristic curves derived through simulations or experiments. Consequently, while these models are fast and ideal for preliminary crashworthiness analysis, design exploration is limited by the amount of component level test/simulation data, highlighting a clear research gap for analytical/semi-analytical models that can inform such multi-body models on the go with material and geometry data as inputs.

This thesis addresses that gap by focusing on the two dominant energy absorption mechanisms in metallic fuselage structures: axial crushing and plastic bending. For axial crushing of closed-section metallic tubular structures, existing models are evaluated and a generalized expression for various cross-section shapes is proposed. For bending collapse, a semi-analytical framework combining spring-linkage based elasto-plastic stage prediction, transition algorithm, and Kecman model for collapse stage prediction is proposed to obtain complete force–displacement curves. Although aircraft structures commonly use open sections, the scope of the present work is limited to closed sections in order to establish a robust and consistent methodological foundation within the available time frame.

Finally, the bending framework is extended to hybrid materials, specifically GLARE (Glass Laminate Aluminium Reinforced Epoxy). A maximum strain-based damage model based on the spring-linkage method is proposed, the approach predicts stiffness degradation and progressive central cracking failure for GLARE laminates. The results demonstrate that analytical and semi-analytical models can be further extended to hybrid systems, enabling fast, early-stage crashworthiness evaluation of unconventional aircraft designs utilizing hybrid materials. ... Read more

The structural design of composite aircraft wings in aeroelastic optimization frame-works is conventionally governed by a linear buckling constraint, which restricts the structural response to the pre-buckling regime across all prescribed load cases.This constraint has been shown to emerge as the active design driver in mass minimisation studies,yet it is conservative. Stiffened thin-walled composite panels are capable of sustaining load beyond the onset of buckling through stable postbuckling behavior, and this reserve capacity remains in accessible in conventional design practice. The consistent in corporation of postbuckling behavior within a computationally efficient aeroelastic optimization framework had not been realized prior to this work.
This dissertation addresses that gap through three sequential contributions.... ... Read more

Optimisation is at the heart of modern engineering. From reducing aircraft emissions to designing safer cars or tailoring drugs for specific diseases, the goal is to find the best solution among countless possibilities. Yet real-world systems are complex, and every design must meet strict constraints related to safety, performance, and physical laws. A design that performs well but violates just one constraint, such as structural failure during flight or non-compliance, is not desirable.

To assess a design’s performance, engineers rely on complex computer simulations that capture physical processes like drag or structural deformation of an aircraft. These simulations often behave like black boxes, as they are expensive to run and the relationship between inputs and outputs is typically non-linear and opaque. This makes exhaustive search of the design space impossible and necessitates data-efficient optimisation strategies.

Bayesian Optimisation (BO) has emerged as a state-of-the-art method for optimising expensive black-box functions, offering a principled way to make the most of limited data. It builds a probabilistic model of the system to guide evaluations efficiently, balancing exploration of uncertain regions with exploitation of promising designs. Although BO has been widely adopted across scientific and engineering domains, it continues to face significant challenges in scenarios that involve both high-dimensional input spaces and complex feasibility constraints. These settings form the primary focus of this thesis.

The first contribution of this work is to show why techniques that work in unconstrained settings, such as random subspace embeddings or simple model priors, often fail under constraints. To address this, the thesis introduces supervised subspace methods and revisits dimensionality-scaled priors that improve both robustness and feasibility discovery in constrained problems.

Second, it proposes scalable strategies to model thousands of constraints, which arise, for example, in structural or aerospace design. Rather than modelling each constraint separately, the thesis uses dimensionality reduction to reduce input and output dimensionality, making constrained optimisation tractable at scale.

Finally, it develops methods for multi-source optimisation, where both accurate and approximate models are available. A modelling framework captures their discrepancies and a novel acquisition strategy balances information gain, cost, and constraint satisfaction, accelerating convergence under tight budgets.

Together, these contributions extend the reach of BO to realistic, simulation-based engineering problems. The resulting tools are broadly applicable and help bridge the gap between theoretical advances in optimisation and the practical demands of high-stakes engineering design.
... Read more

The increasing demand for lightweight and cost-effective structures in aerospace and wind energy applications has led to widespread adoption of fibre-reinforced polymers (FRPs). While carbon fibre reinforced polymers (CFRPs) offer excellent stiffness-to-weight ratios, their high cost and poor damage tolerance pose challenges for structural applications such as wind turbine and helicopter rotor blades. Hybrid laminates combining CFRPs and low-cost, higher failure strain (compared to CFRP) Glass fibre reinforced polymers (GFRPs) present a promising alternative, balancing competing objectives of cost and stiffness while providing higher damage tolerance. Most existing optimisation studies of hybrid composite plates treat material selection and ply orientation as decoupled problems, lacking a unified formulation that can capture their interaction holistically. Furthermore, existing approaches prevent the application of gradient-based optimisation via lamination parameters(LP), efficient for stiffness-targeted optimisation, when attempting to unify material assignment and ply orientation within a single design framework. This thesis develops a novel method to enable the use of lamination parameters for hybrid composite laminates by introducing material-dependent dispersion parameters(DP), unifying material selection and ply orientation within a single design formulation. With the material properties of the laminate as a function of volume fraction, the dispersion parameters describe the through-thickness material distribution in a continuous, geometry-independent manner. The combined LP–DP formulation defines a continuous, convex design region amenable to gradient-based optimisation frameworks. The DPs and feasible region are further explored to create a combinatorics-based and deterministic method to constrain the region, creating a relation between LPs and DPs. Further, different optimisation strategies have been discussed to achieve an optimum solution for design problems. A two-level optimisation strategy is roposed, combining a gradient-based optimisation for target parameter identification and a genetic algorithm to retrieve manufacturable, symmetric, and balanced stacking sequences that comply with standard layup rules. ... Read more

This thesis presents a methodology for constitutive material modeling of flexible targets, with a specific emphasis on finite element analysis. The Johnson-Cook constitutive model is employed in combination with a customized Bao-Wierzbicki damage model, both implemented in Abaqus/Explicit through a user-defined material subroutine (VUMAT). The primary objective is to evaluate the accuracy of the finite element simulations by comparing them to experimental tensile test data. To this end, the dogbone specimen used in an experimental test was reconstructed in Abaqus, and stress-strain as well as reaction force-strain data were extracted from a single finite element located in the necking region. These data were then compared to strain measurements obtained via Digital Image Correlation (DIC) from the physical experiments. The analysis was performed across four mesh densities to assess mesh convergence.

The results indicate a reasonable agreement between the simulated and experimental force-strain and stress-strain responses. Minor discrepancies were observed, largely attributable to uncertainties in the experimental measurements. Overall, the simulations demonstrate the ability of the implemented material and damage models to capture key aspects of the observed mechanical behavior.

This validated simulation framework provides a foundation for future investigations involving high-deformation phenomena. It is particularly relevant for impact analyses, such as bird strike scenarios in aerospace applications. ... Read more

In this work, a conceptual design was created that integrates battery packs into the wings of a 90-seater electric aircraft. This was done under the challenging requirement of replacing over 35 tonnes of batteries more than once a year. The concept allows the batteries to move spanwise along a rail system and through the rib using a panel structure. The core of this concept is that it avoids the need to remove large skin panels of the wing. From the broader concept, a single component was analyzed in detail using non-linear finite-element models combined with fretting-fatigue analyses and physical testing. ... Read more

Bird strikes pose a significant threat to aircraft safety, yet accurately predicting their impact behaviour remains challenging due to many uncertainties, such as variations in bird properties, high impact velocities, and limited experimental data. While much research has focused on modelling the bird itself, understanding the structural response of materials, particularly aluminium alloys commonly used in aircraft structures, is just as important. This thesis investigates how different material model definitions in the plasticity regime affect the predicted response of aluminium 2024-T3 under bird strike conditions.
Moreover, it investigates the uncertainties that arise from limited quasi-static testing or usage of inconsistent material properties that can be found in the literature. The work aims to highlight the influence of modelling choices and identify key sources of uncertainty related to material characterisation. Using a surrogate model implementation for two chosen plasticity models, Ramberg-Osgood and Johnson-Cook plasticity models, the uncertainties are being assessed through one-factor-at-a-time sensitivity analysis together with Monte Carlo analysis. The results of this work contribute to improving the reliability of bird strike simulations and provide guidance for future model development. ... Read more

Inverse design is a concept where one can design a structure via a property-first approach, where the properties of a system act as precursor information to guide the search for a viable design. This concept is classified as one of two: indirect or direct. While the former is no different than a forward-based approach using traditional optimization, a direct inverse design framework attempts to invert the process and directly map the desired properties to a suitable design candidate, which is unconventional. As such, the design process can be positively radicalized, with regards to the required computational resources when designing multiple structures.

In this work, the use of Bayesian machine learning, more specifically Bayesian optimization using Gaussian process regression, is employed to construct a direct inverse design framework using the design of mass-optimized fixed-wing aircraft ribs against buckling as the validation case study in 1, 6, and 10 dimensions. The results are compared to an indirect approach using the same algorithms and demonstrate that barring certain limitations, which are not inherent flaws of the direct approach, the framework designed not only demonstrates sound potential in inverting the forward map, but outperforms the indirect approach when designing multiple structures.

This research equally provides a stepping stone towards future research possibilities in the same field, all culminating in the improvement of the multi-disciplinary design process.
... Read more

With the progressively more widespread exploitation of post-buckled skin-stiffened structures in the aerospace industry, aimed at making them more weight efficient, assessing the failure of such structures is of utmost importance. Additionally with the trend of moving towards fastener free connections, one of the potential failure modes of skin-stiffened structures is the skin-stiffener separation that leads to a sudden and drastic reduction in the load carrying capacity of the structure. Hence, it is imperative that structures are designed against such failure modes, being crucial to have predictive models for the initiation and propagation of delaminations at the connection interface.

Given this motivation, the present thesis investigates the development of a novel multi-domain semi-analytical cohesive zone approach to predict the initiation and propagation of damage along a two-dimensional interface. In the multi-domain method the main structure being analyzed is decomposed into multiple smaller domains for computational advantage, with each domain having its own set of continuous approximation functions based on the Ritz method. A penalty formulation governs the inter-connection of domains. To extend the multi-domain framework to model damage, the interface connection is modelled as a cohesive zone with the traction-separation law governing its response. The area connection from the multi-domain method is modified to include a damage-dependent penalty stiffness term, whose degradation is controlled by the traction-separation law of the cohesive zone model. Lastly, the energy dissipated to create a crack is included in the formulation. Since the framework is based on the Ritz approach, the appropriate choice of continuous approximations to model damage is investigated, presenting their limitations and the proposed methods to overcome them.

Finally, the results obtained from the multi-domain framework were verified against finite elements for a multitude of cases, which in turn had been validated by experimental data. The results were subsequently followed by discussions and the caveats of the developed framework. Results indicate that the developed multi-domain semi-analytical cohesive zone method predicts the results within 4.3% of finite element results, while taking 75% less time on average.

Finally, having somewhat successfully modelled the initiation and propagation of damage along a two-dimensional interface, this thesis establishes a robust basis for future research involving the use mixed-mode behaviours with multi-linear cohesive laws, whose formulations can just be slotted into the current framework. All of this forms a solid foundation to efficiently model damage initiation in larger structures and, more concretely, skin-stiffener separation. ... Read more

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. ... Read more