Theses - Engineering
Permanent URI for this collection
Browse
Recent Submissions
Item Open Access Photothermal and Photoelectric Effects in Lithium-Ion Batteries: Mechanistic Insights and Performance EnhancementTan, LifuThe growing demand for reliable, sustainable off-grid power solutions is especially significant for Internet of Things (IoT) devices. Solar energy, as a widely available renewable resource, has advanced energy-harvesting and storage technologies. Traditional solar-to-electricity setups rely on separate components, leading to larger device footprints and challenges with output voltage mismatches. This research explores an optimized, integrated system that combines light-harvesting and energy storage using a shared electrode, reducing device size and enhancing efficiency. The key innovation lies in the material selection for the shared electrode, balancing electrochemical performance and photoactivity. To start with, graphitic carbon nitride and bismuth vanadate were selected as potential photo-active materials for photo-enhanced batteries. However, they were proven to be not suitable for the application from their electro- chemical measurement due to the low capacity and degradation. Prussian blue analogues (PBAs) were then selected for their photothermal heating efficiency and compatibility as cath- odes in photothermal-enhanced Li-ion batteries. Electrochemical testing under illuminated and dark conditions demonstrated that light-induced heating boosted battery performance, increasing capacity by up to 38% at high current densities. EIS measurements further con- firmed reduced charge transfer resistance with illumination, underscoring the critical role of photothermal effects. A novel measurement technique, named impedance-based internal temperature estimation, clarified the impact of photothermal effects on battery performance. This study provides insights into how material properties and bias voltages can balance pho- tothermal and photo-generated charge effects, advancing our understanding of photo-induced processes. Furthermore, an in-depth examination of band alignment between the photoelectrode and counter electrode highlighted the mechanisms of charge transport, differentiating between photothermal and photoelectric effects through ultraviolet photoelectron spectroscopy (UPS) and UV-Vis spectroscopy in Li-ion batteries by using semiconducting metal oxide including anatase/rutile TiO2 and Fe2O3. Results indicated that photothermal effects dominated at the applying voltage which is below the energy gap between the conduction band minimum and Li plating/stripping potential, while a higher bias voltage activated photoelectric effects by band-bending at the interface, revealing the influence of band alignment on charge transport. Overall, this work illustrates that the processes taking place in photo-batteries are intricate, and it offers new electrochemical protocols and techniques to gain insight into the mechanisms that govern the changes in behaviour when illuminating photo-batteries. In conclusion, this work contributes valuable insights into integrating light-harvesting with energy storage systems, emphasizing the importance of material selection, photothermal effects, and precise measurement techniques. These findings support future advancements in photo-rechargeable batteries and sustainable energy technologies.Item Open Access Probabilistic machine learning algorithms for molecule discoveryTripp, Austin; Tripp, Austin [0000-0002-0138-7740]Discovering new molecules empowers humanity to solve problems in health, agriculture, energy, and more. The key challenge of molecule discovery is that the space of all possible molecules is vastly larger than the amount of molecules which we can test experimentally with our limited resources. Given this challenge, arguably the best option is to judiciously select which molecules to test based on both our current knowledge and our expectation of what information will be gained from each test. In machine learning, this approach is typically called Bayesian optimisation and has been studied for many other problems, such as tuning hyperparameters of machine learning models. Although in principle Bayesian optimisation can be straightforwardly applied to the problem of discovering new molecules, the discrete nature of molecules means that new models and algorithms are needed to make Bayesian optimisation work in practice. This thesis presents various probabilistic machine learning algorithms which could be used within a Bayesian optimisation loop to discover new molecules. Latent space optimisation with weighted retraining (chapter 3) and adaptive deep kernel fitting with implicit function theorem (chapter 4) are both algorithms which use a Gaussian process with a deep neural network kernel function to model the relationship between molecular structure and some property of interest. Tanimoto random features (chapter 5) allows an established cheminformatics model to be applied (approximately) to large datasets. Finally, retro-fallback (chapter 6) uses a novel probabilistic formulation of the retrosynthesis problem to estimate whether a molecule can be synthesized, and thereby determine whether it should be considered by Bayesian optimisation. Together, these algorithms form a suite of tools which could be used to discover new molecules automatically and intelligently.Item Embargo Navigating Global Shifts: Organisational Behaviour and Strategic Responses to Shifting Business EnvironmentsTsai, Tsung-Yu; Tsai, Tsung-Yu [0000-0002-0414-7005]Outsourcing has become integral to international manufacturing over the past few decades, enabled by technological advances that exploit cost differences between economies. However, a notable trend is emerging where firms are shifting back to in-house production or local subcontracting. In this context, this thesis explores the dynamic interplay between global shifts and organisational behaviour, focusing on strategic responses to evolving business environments. It analyses firm behaviour from two perspectives: earlycomer international manufacturers and latecomer contract manufacturers, aiming to provide a comprehensive understanding of organisational behaviour under shifting business conditions. The thesis first explores how location decisions are made amid global economic shifts. It analyses how firms that previously offshored production evaluate economic variables, navigate uncertainties, and capitalise on opportunities. It also underscores the complexities and impacts of strategic decisions influenced by operational and geopolitical factors. Next, the thesis shifts to an empirical analysis from the perspective of earlycomer firms. It explores how these firms make relocation decisions in response to geopolitical shifts and evolving global dynamics. It also highlights the growing importance of geopolitical considerations over traditional economic factors, showing that earlycomer firms adopt a phased approach to relocation that balances market logics with state logics to mitigate risks and ensure long-term stability. The thesis then presents the perspective of latecomer subcontractors, focusing on their strategies for upgrading and surviving within global value chains (GVCs). It identifies dynamic capabilities that enable these firms to thrive despite resource constraints and power asymmetries. It integrates concepts from business ecosystems and dynamic capabilities to explain how latecomers navigate and adapt to changing GVCs. Finally, the thesis synthesises insights and discusses the interconnectedness of its findings, stressing the relevance of earlycomer strategies for latecomers and vice versa. It highlights how understanding and aligning with each other’s strategic priorities can enhance resilience and adaptability. The thesis contributes to academic discourse and offers practical implications for management and policy by providing a holistic understanding of corporate decision-making in the context of shifting global dynamics.Item Open Access Achieving excellence in managing customer experience – An exploration of smaller details that make a differenceJha, GautamEstablished firms acknowledge the importance of a customer experience (CX) as a leading strategic approach but face challenges in managing customer experience compared to digitally born organisations. However, in theory and practice, the concept of customer experience management is fragmented across contexts and approaches. Therefore, this study to explores a comprehensive and holistic approach to managing customer experience. Drawing upon the theory of the mundanity of excellence, this study introduces a fresh approach to manage and operationalise customer experience based on a framework (ACM): attitudes (A) towards customer experience centricity, capabilities (C) that integrate customer journey management and mastery of methods (M) to optimise customer experience. Integrating insights from indepth interviews with fifty senior leaders, the study identifies nine smaller detail sub-themes encapsulated within three overarching categories of ACM. Furthermore, the study identifies five operational stages of customer experience management - siloed, reactive, cohesive, data-enabled and culture-driven, each distinguished by their focus on specific attitudes, capabilities and methods. Finally, as a practical implication, the study offers guidance on how firms can evolve within the operational stages to better implement the ACM approach providing valuable insights for executives, CX leaders and practitioners.Item Open Access A study on the properties of graphene nanoribbons for electronicsLiu, XiaoGraphene has the potential to lead the construction of next-generation electronics due to its high mobility, high thermal conductivity, and atomically thin two-dimensional structure. However, graphene’s intrinsic zero bandgap property restricts its use in a wide range of applications. Thus, the controllable engineering of the band structure of graphene is both important and necessary for many future applications. Nano-lithography has facilitated research into band structure engineering of graphene, as it has been shown that a size-dependent bandgap opens for graphene nanoribbons with widths below around 40 nm. To date, photo- and electron-beam lithography (PL & EBL, respectively) have routinely been used to etch graphene to fabricate graphene nanoribbons with sizes down to around 15 nm. However, these processes are complex and time-consuming, involve the use of solvents, resists, metals, and often oxygen plasma, and they result in devices with potentially significant amounts of residues and chemical contamination. In this thesis, we explore AFM-based scanning probe lithography (SPL) as an alternative method for the minimization of graphene-based field effect transistors (G-FETs), as it is single-step, potentially higher resolution, and does not involve the use of any chemicals other than water, so is essentially a direct-write process. We investigated three kinds of AFM-based SPL, namely (i) mechanical, (ii) DC voltage and (iii) AC voltage, and demonstrated their capabilities for patterning graphene at the micro- and nano-scale. Finally, we developed the process to such a degree that we were able to demonstrate a sub-10nm G-FET device by AFM-based SPL with AC voltage, which signifies the narrowest graphene nanoribbon (GNR) fabricated by a top-down lithography method. In addition, we have observed that the charge neutrality point (CNP) shifts as the width of these GNRs reduces below 80nm. This is in good agreement with previous studies on the CNP shift in G-FET deices fabricated by EBL in our group. To explore the mechanisms underlying this phenomenon, we performed simulations using Cambridge Serial Total Energy Package (CASTEP) to compute the band structure of GNRs with different widths and both with and without H-passivation at the edges, as a first step towards taking edge effects into account. Combined with our experimental findings, we propose that the effect of dangling bonds at the edges is to lower the energy levels of bands in GNRs, resulting in a shift of the CNP. Our investigations contribute to a better understanding of edge states in graphene nanodevices.Item Open Access Resolving oxidation phenomena in layered semiconductors and scaled heterogeneous integration for low-power electronicsChirca, IrinaEfficient materials development demands high-throughput experimental workflows and prompt feedback loops across all stages, from materials growth to device fabrication and quality control. There is a crucial need for characterisation approaches that can resolve both intricate structure-property relations at the atomic layer level and enable intelligent, cost-effective screening at high throughput. This is particularly pertinent to 2D materials, where there continue to be many unexplored layer- and stacking-dependent properties. This work details the development and implementation of a spectroscopic imaging ellipsometry (SIE) approach for the multi-scale analysis of layered semiconductors, focusing on oxidation phenomena. A versatile fitting algorithm is adapted for the rapid determination of the material’s complex dielectric function (ε), enabling optical modelling of multi-layer heterostructures. Combined with the various modes of SIE operation, it has the potential to unlock fast, high-throughput, large-area capability to better understand material stability and oxidation mechanisms, as well as accelerate process development. This methodology is adjustable to multiple material systems and advantageously agnostic to the underlying substrate, as exemplified through the analysis of HfS2, GaS, and monolayer WS2 systems. Given the facile extraction of their ε, we demonstrate the construction of optical models for accurate layer-thickness determination in partially oxidised samples. This can be scaled to the non-destructive 3D mapping of semiconductor / oxide heterostructures, from 1 μm lateral resolution to wafer-scale processing. When applied to the study of HfS2 oxidation, this methodology enables ready access to buried HfS2 layers, oxide quality, and lateral and vertical uniformity. The SIE analysis of the native oxide layer as a function of oxidation conditions reveals the large variation in thickness and optical properties achievable through fine-tuning of the reaction parameters. Moreover, operando capability is demonstrated for thermal oxidation up to 400 ◦C, providing insights into the temperature- and time-dependent nature of self-limiting oxide growth, and the trapping and eventual release of sulphur reaction products. Finally, this SIE methodology is used to inform the fabrication conditions of semiconductor / oxide heterostructures for future integration into resistive switching devices.Item Restricted Item Open Access Approximate Message Passing for Matrix RegressionTan, Thong Cai NelvinApproximate message passing (AMP) algorithms have become popular in various structured high-dimensional statistical problems. Previous AMP algorithms for generalized linear models (GLM) typically require the signal to be in the form of a vector, and the design matrix to have independent and identically distributed Gaussian entries. In this thesis, we explore estimation problems where previous assumptions of AMP algorithms are no longer satisfied. To relax the vector signal assumption, we introduce the matrix GLM model which accounts for matrix signals, and propose an AMP algorithm for estimation and rigorously characterize its performance in the high-dimensional limit. We consider prominent cases of the matrix GLM model to illustrate its usefulness, namely mixed linear regression, max-affine regression, and mixture-of-experts. For max-affine regression, we propose an algorithm that combines AMP with expectation-maximization to estimate intercepts of the model along with the signals. To relax the Gaussian design assumption, we extend the matrix GLM to have a generalized white noise design matrix (instead of Gaussian), and propose an AMP algorithm for estimating the matrix signal and rigorously characterize its performance. We apply this model to the pooled data and the quantitative group testing (QGT) problem, where the design matrix is binary-valued. For the noiseless pooled data setting, we show that the studied AMP algorithm is equivalent to one recently proposed by El Alaoui et al. Our results provide a rigorous version of their performance guarantees, previously obtained via non-rigorous techniques. For comparison, we propose estimators based on convex relaxation and iterative thresholding, without providing theoretical guarantees. To further improve the performance of AMP algorithms for QGT and pooled data, we introduce the spatially coupled Bernoulli test matrix and an AMP algorithm. We rigorously characterize its asymptotic performance in both the noiseless and noisy settings, and prove that in the noiseless case, the AMP algorithm achieves almost-exact recovery with a number of tests sublinear in the number of items. For both QGT and pooled data, this is the first efficient scheme that provably achieves recovery in the linear regime with a sublinear number of tests, with performance degrading gracefully in the presence of noise.Item Open Access Stochastic modelling for control and interconnections of biological systemsDonchev, Tihol IvanovBiochemical reactions often occur in small volumes within a cell, restricting molecule numbers to the hundreds or even tens. At this scale, reactions are inherently discrete and stochastic. Traditional models based on differential equations and mass-action kinetics fail to capture the behaviour of such systems. In contrast, stochastic models can suffer the curse of dimensionality, complicating system analysis and design in this low-copy number setting. The first part of this dissertation explores a novel decision-making mechanism in small cell compartments by introducing a novel self-regulating signalling motif: the Size-Regulated Switch (SRS). The SRS allows reliable switching at small system sizes and transitions to a stable behaviour as the size increases. Since many cellular compartments grow in size, the system size can act as a feedback signal to self-regulate the switching behaviour. These results are then generalised to different network topologies with two or three species, allowing both inhibition and excitation, finding that inhibitory connections robustly acquire size-dependent multimodality, while excitatory connections tend to suppress this behaviour. These results provide a potential solution to the contradictory findings surrounding the CaMKII/protein phosphatase-1 pathway, showing that bistability can be sensitive to the absolute quantity of reactants present. The second part of this dissertation develops a novel input/output framework for ap- proximating the stationary relationship between dependent species within a feedforward biochemical reaction network. These networks can consist of many species, including nonlinear interactions and combining series and parallel interconnections, making simu- lation methods computationally expensive and approximation methods inaccurate in the low copy number regime. The Conditional Poisson Approximation CPA method allows the approximation of the stationary behaviour of these stochastic systems, accounting for the inherent discreteness and stochasticity of the molecular species, as well as the nonlinear interactions such as Hill-type functions. Overall, the CPA method provides a scalable and accurate approximation of nonlinear feedforward biochemical networks, including both series and parallel interconnection. Furthermore, its potential extension to feedback systems is illustrated through a simple example.Item Open Access Probabilistic Methods for Monocular 3D Human ReconstructionSengupta, Akash; Sengupta, Akash [0000-0002-5162-8696]Remarkably, humans can infer the approximate 3D surface geometry and appearance of complex articulated objects, such as people and animals, given but a single 2D RGB image. In fact, we can reason about the whole range of 3D poses and body shapes, surface colours and textures that can plausibly explain a given image. Thus, we instinctively acknowledge and account for the ill-posed nature of the problem of monocular 3D reconstruction, where a single 2D input image gives rise to multiple reasonable 3D solutions. In recent years, great progress has been made towards computer vision algorithms that can imitate a human's ability to reconstruct in 3D from partial 2D observations. These 3D reconstruction methods facilitate impactful applications in healthcare, robotics, virtual retail and entertainment. However, most approaches in the contemporary research literature are deterministic, and estimate a single 3D ``best-guess'' solution given an input image -- ignoring the inherent ambiguity in monocular reconstruction. Blindly assuming that a single 3D estimate matches the true 3D geometry of the subject, which is infeasible in such an ill-posed setting, can result in failures in reconstruction-reliant downstream applications. Moreover, a deterministic approach may also curtail the quality of monocular 3D human geometry and appearance estimates, resulting in blurry colours and over-smooth surfaces in ambiguous regions of the body. This thesis develops probabilistic approaches to 3D human reconstruction, which predict probability distributions over 3D reconstructions conditioned on a single 2D RGB image. This enables us to sample any number of plausible 3D hypotheses during inference, and quantify and visualise prediction uncertainty, indicating the level of confidence our methods have in different reconstructed regions of the body. Specifically, Chapter 5 presents a selection of model-based probabilistic reconstruction methods, which involve predicting distributions over the parameters of a statistical body model. The extra information present in a predicted 3D distribution, beyond a single 3D point estimate, is valuable in downstream tasks. For example, it facilitates probabilistic fusion of 3D solutions from multiple images, or model fitting with an image-conditioned prior probability distribution -- both of which are demonstrated in Chapter 5. Moreover, Chapter 6 introduces a model-free probabilistic reconstruction method that yields photorealistic 3D samples with sharp colours and fine geometric details, even in unseen regions of the body. We predict probability distributions using deep neural networks that are trained via supervised learning. This requires suitable training data -- i.e. images of humans with diverse poses, body shapes and scene conditions that are accurately labelled with the subject's ground-truth 3D geometry. Before detailing our probabilistic reconstruction methods, we present a synthetic training data generation pipeline for 3D pose and shape regression in Chapter 3, which overcomes a trade-off between 3D label accuracy and data diversity exhibited by contemporary real training datasets. In a similar vein, contemporary evaluation datasets for 3D human reconstruction also feature limited diversity of body shapes. To this end, Chapter 4 introduces an evaluation dataset for parametric body shape estimation -- Sports Shape and Pose 3D (SSP-3D) -- which contains 311 RGB images of 62 sportspeople with a wide range of body shapes. These works are used through this thesis to train and evaluate our probabilistic reconstruction methods.Item Open Access Hydrogen Enrichment Effects on Thermoacoustics of Turbulent Partially Premixed FlamesDilip Kumar, AnkitPower generation and aircraft gas turbine combustors are frequently operated under fuel lean conditions to mitigate harmful emissions, rendering them susceptible to undesirable thermoacoustic instabilities. Additionally, meeting decarbonisation targets require incorporation of hydrogen into the existing combustion infrastructure. This doctoral thesis addresses the challenges of hydrogen combustion and thermoacoustic instability in Gas Turbine Model Combustors (GTMC). A bimodal thermoacoustic instability where two different modes are excited are investigated using Large Eddy Simulations (LES) for a partially premixed swirl stabilised methane-air flame in a GTMC. Simulations are also performed for cases with methane-hydrogen blends revealing a similar bimodal instability, where the frequency of thermo acoustic oscillations differs significantly from the cavity acoustic mode frequency due to the presence of intrinsic thermoacoustic modes. These modes are analysed using a 1D low-order model which offers insights on the appropriate area-ratio required at different junctions of the geometry to mitigate these modes. This thesis further explores pure hydrogen combustion in a Lean Direct Injection (LDI) combustor. Experimental results demonstrating thermoacoustic instabilities in this combustor are characterised using dynamical systems theory, revealing the presence of period-1 LCO, period-2 LCO, intermittent, quasi-periodic, and chaotic states, as either bulk velocity or equivalence ratio is varied. The unstable acoustic modes and their spatial behaviour are investigated using a reduced-order model. This thesis also addresses the challenge of modelling direct combustion noise, essential for mitigating its harmful effects on frequent flyers and residents living close to airports. Direct combustion noise is closely related to the Heat Release Rate (HRR) spectra, which this work investigates through LES analysis of three different configurations operating with various CH₄-H₂-air mixtures at atmospheric pressure. The global HRR spectra reveal an atypical spectral decay of f⁻⁵ at high frequencies across all cases. Local HRR spectra are qualitatively similar to the global spectra but are influenced by local mixing and coherent hydrodynamic structures. This analysis facilitates modelling local HRR spectra from non-reacting velocity spectra.Item Open Access Mechanisms of hot gas ingestion in turbine disk cavitiesDawson, MichaelIn a gas turbine engine, hot gas ingestion occurs when gas from the mainstream flow path migrates into the cavities between stator and rotor disks. In order to mitigate the damage caused by these high-temperature gases, the cavities are purged with colder air from the compressor, and rim seals are used at the entrance from the mainstream. Various mechanisms of ingestion have been proposed, and studied in great detail, but other potential mechanisms have been largely neglected. This thesis aims to test the sensitivity of ingestion to features of flow and geometry about which little is known, and to investigate how their effects might be predicted by engine designers. The thesis uses experimental testing, together with two- and three-dimensional modelling methods, to demonstrate that long wavelength, small amplitude variations in mainstream pressure are the dominant cause of hot gas ingestion for an engine-representative test case. This mechanism creates a flow pattern in the rim seal that drives hot mainstream gas into the cavity in discrete locations, increasing the risk of localised thermal damage. CFD simulations incorporating these variations predict average cavity seal effectiveness to within 2% of measured values, and capture local variations in seal effectiveness. A two-dimensional model of ingestion is developed and tested, based on the linearised Euler equations. The model uses a parametric representation of the mainstream pressure variation to model the flow in the seal, and tracks the transport of mainstream gas into the disk cavity. Results from the model show good agreement with experimental data for different mainstream pressure distortions and purge flow rates. The model is sensitive to several key geometric and flow parameters, and it is used to estimate the effects on ingestion of changes to mainstream pressure, rim seal geometry, and engine operating conditions. The final chapter of the thesis reports an experimental study of the effects of rotor leakage flows on hot gas ingestion. It is shown that the presence of leakage flows increases the purge flow required to maintain desired cavity temperatures, and amplifies local variations in surface temperature. A simple, semi-empirical model for the flow in the cavity is presented that estimates the effect of leakage on ingestion, based on data with no leakage present. The model shows good agreement with experimental measurements, and provides valuable insight into the structure of the flow in the disk cavity.Item Open Access ON THE AEROSOL SYNTHESIS OF HIERARCHICAL HYBRID CARBON NANOTUBE NANOMATERIALS FOR LITHIUM-ION BATTERIESLeclerc de la Verpillière, Jean; de la Verpilliere, JeanAchieving the ambitious performance and cost goals for Li-ion batteries set by the industry at the 2030 horizon requires significant progress in the design and mass-production of active electrode materials. The aim of this dissertation is to progress the field of advanced battery materials by investigating novel materials designs and characterisation tools enabling the potential scale-up of their production process. In particular, we focus on the aerosol-based production of a novel hierarchical hybrid carbon nanotube (CNT) nanomaterial, the suitability of common aerosol neutralisers to monitor its production process, and its application to Li-ion batteries. This material is an ‘urchin-like’ nanostructure, whereby CNTs are grown radially from a central metal oxide core, which we refer to as Carbon Nanotube Sea-Urchin (CNTSU). We investigated the aerosol synthesis and characterisation of CNTSUs, providing new insights into the mechanisms of CNTSU synthesis following a continuous aerosol production process combining spray-pyrolysis to form CNTSU metal oxide cores, and growth of CNTs from these cores by Chemical Vapor Deposition (CVD). Further insights were gained on the physical and chemical characterisation of CNTSUs, which may be useful for their application as Li-ion battery anodes. Furthermore, it was found that mass-mobility relationship in-line aerosol measurements of the CNTSU process could generate valuable insights into process scale-up and fundamental CNTSU properties. However, these measurements rely on the use of an aerosol neutraliser, operated with the assumption that charging equilibrium is reached within the neutraliser. That assumption has not been challenged by existing academic literature in the context of purposeful nanomaterial production, where, typically, aerosol particle concentrations are of the same order as neutraliser ionic concentrations, particles are small, and neutraliser residence times are low. We find that the commonly-accepted neutraliser equilibrium indicator, the 𝑛 ∙ 𝑡 product, rule fails to predict equilibrium behaviour in that context. We propose a nondimensional approach to predict equilibrium behaviour as a function of two nondimensional groups: 𝑁̂ the non-dimensional ion concentration, and 𝜏̂the non-dimensional residence time. This approach was successfully applied to the CNTSU synthesis process discussed above, and it was found that, fortunately, the specific neutraliser used in this work was likely to generate an equilibrium charge distribution, but that it may not be the case of other common aerosol neutralisers. Finally, we investigate the application of CNTSUs as conversion anode for Li-ion batteries. Our results confirm the relevance of CNTSUs as interesting candidates for next-generation Li-ion battery anode materials, thanks to their high specific capacity, high lithiation average potential avoiding dendrite formation, encouraging specific capacity retention at high rates, and initial indications of their compatibility with a common commercial cathode material. Some of these features may be attributed to hierarchical nature of the CNTSU electrode conductive network compared to a network of randomly-mixed, weakly contacted CNTs, such as those reported in conventional electrodes where CNTs are externally added as a dry powder during the electrode slurry mixing step. Similarly, we suggest that using CNTSU-type structures for electrode preparation may be a viable route to addressing the known issue of CNT segregation during electrode mixing and coating. However, CNTSUs do not address other common issues associated with conversion anodes, including a sloping voltage profile with high average potential vs. Li/Li+ , voltage hysteresis, and excessive Solid-Electrolyte Interface (SEI) formation leading to poor cell efficiencies and cycle life in a full cell. We suggest that further investigating the lithiation mechanism of CNTSUs, and modifying the CNTSU core composition may be a valuable future research direction.Item Open Access A Data-Centric Approach to Loss Mechanisms for Compressor Preliminary DesignSenior, AlistairIn the preliminary stages of design, low-fidelity models, which describe the underlying design space, are used to guide designers to optimal points in the design space. However, the majority of models currently used in compressor preliminary design are often limited by the human ability to find physical patterns in complex multidimensional data. Detecting patterns in complex multidimensional data is exactly the task for which machine learning and other statistical techniques have been developed. The aim of this thesis, therefore, is to try and use machine learning to augment the human ability to find physical relationships and to consequently develop a more physically accurate and general preliminary design system for compressors. In this work, a data-centric approach is proposed that utilises the strengths of machine learning to highlight patterns in complex data, while also utilising the strengths of human wisdom to ensure the models are physically meaningful. This approach is applied to datasets populated by large numbers of computational simulations for compressor and turbine blades. It is shown using 2D data that a new profile loss model can be developed which halves the error in preliminary design correlations and generalises well to turbines as well as compressors. The reason for these improvements is shown to be the result of a more physically accurate decomposition for mixing loss than current industry standard models. This new mixing loss model utilises flow properties local to the wake and is shown to explain the limiting cases for the mixing of two flows, the mixing out of a trailing edge blockage and the form drag of an aerofoil. It is argued that the reason this model is more accurate and general than existing models is that it retains the physical form for the source of loss creation at the fundamental scale. The data-centric approach is also applied to a 3D RANS compressor linear repeating stage dataset and shown to help develop a model for the total loss in a compressor blade row that is composed of physically accurate models for each of the key loss sources in a blade row. This model is demonstrated to halve the error in loss predictions compared to industry standard preliminary design models, and capture the key sensitivities to trailing edge thickness, lean and sweep, not previously accounted for at the preliminary stages of design. In particular, new, more accurate and general models for tip loss and wake mixing loss are highlighted that have the same physical form as the source of loss creation. The resulting model is finally used to generate Smith Charts to explain how the optimal point on the Smith Chart is very sensitive to the choice of preliminary design parameters and how these parameters can be used to move the optimal point. In particular, it is highlighted that there is a need to consider 3D effects in the preliminary stages of design to move the design point to higher stage loading coefficients.Item Open Access The Influence of Propeller Position and Size on the Aerodynamics of Blown WingsHawkswell, GeorgeThe drive to decarbonise the aviation sector is pushing aircraft designers towards more-electric propulsion architectures. One of the key benefits of electric propulsion is that multiple motors can be positioned around the airframe, without loss of motor efficiency. By placing many motors along the wing leading edge, each powering a propeller, increased lift can be achieved by blowing high velocity jets over the wing. This allows the wing to be reduced in size, decreasing its cruise drag, whist still maintaining the same lift during take-off and landing. It is shown that the use of small diameter jets allows high velocity jets to be used for a fixed landing thrust, increasing the lift coefficient. It is also shown that the wing must be fully immersed in the jet to avoid premature stall. This is only possible by placing the jet below the wing centreline, due to the wings’ up-wash. However, the use of small propellers placed below the wing centre-line requires many large nacelles, with a large associated drag. When the effect of nacelle installation drag is included, the optimum blown wing geometry to minimise drag is shown to be when the horizontal offset of the propeller is 25% wing chord in front of the wing leading edge, with no vertical offset of the propeller from the wing centre-line and with a jet diameter of 64% wing chord. This optimum is true for the wing and aircraft operating conditions studied in this thesis. This geometry achieves a 30% reduction in wing profile drag at cruise. By comparing 2D CFD modelling methods to 3D CFD and experimental data, it is shown that the lift coefficient of a blown wing is primarily driven by two things: the jet thrust, and jet position relative to the wing. 2D modelling of the blown wing is therefore able to capture the lift and drag without the need to use more complex models. 2D models can also be modified to account for the effect of propeller spacing. It is found that the propeller spacing can be increased, with a minimal reduction in the blown lift, provided that the total thrust of the propellers remains constant. Increasing the propeller spacing reduces the number and hence drag of the nacelles. The mechanism of blown wing stall when a flap is deployed is identified. It is shown that the flap stalls due to boundary layer diffusion resulting from a combination of the propeller jet and flap incidence. Moving the propeller down vertically increases the maximum lift coefficient, provided the jet does not pass below the wing, as the wing remains fully immersed in the jet up to higher flap angles. Increasing the propeller spacing leads to boundary layer separation at lower flap angles, however the growth of the separation with flap deployment is slowed when compared to closely spaced propellers, as less of the flap is affected by the propeller jet; as a result, spacing the propellers out also increases the maximum lift coefficient. 2D modelling of the blown wing lift coefficient captures the trends in maximum lift coefficient with different vertical offsets, but is unable to capture the increased un-stalled flap angle range afforded by increasing the propeller spacing. A power balance model for the aircraft shows that a blown wing can achieve a 12.2% reduction in power consumption in cruise if the leading edge propellers are stowed away after take-off. The optimum propeller position to achieve this is an axial offset of 25% chord upstream of the wing, a vertical offset of 5% chord below the wing chord-line and with a propeller diameter of 64% chord. If the leading edge propellers are not stowed away, and instead are also used in cruise, a 10.5% reduction in power consumption can be achieved. For this case, the propellers are positioned in the same horizontal location, but with no vertical offset from the wing chord-line and a larger diameter of 90% chord. Spacing propellers out offers a further 0.5% reduction in power consumption at cruise.Item Open Access Multimodal Soft Sensor Design through Single-Material Information StructuringHardman, David Samuel; Hardman, David [0000-0002-5102-0541]Human skin provides flexibility, compliance, pressure sensing, temperature sensing, and stimulus localisation while also being able to detect and heal damages. As technologies move towards the design of increasingly general-purpose robots, we require soft sensors which replicate these essential properties without sacrificing robustness, longevity, or straightforward fabrication. This thesis proposes a design framework in which multimodal soft sensors are considered as information-generating systems, the outputs of which can be structured through material, morphology, and electrical excitation co-design. The effects of each are experimentally explored, with the ultimate goal of single-material soft multimodal sensor fabrication. Their outputs are analysed through the learning-based processing of large physically-collected datasets, using architectures capable of adapting to the nonlinear responses of functional materials such as self-healing polymers. In doing so, numerous novel contributions are demonstrated: designing at the material level yields a 3D-printable sensorised hydrogel which responds to various multimodal stimuli; morphology explorations lead to a temperature & strain-sensitive membrane inspired by the distribution of receptors in the human skin; and multiplexed electrical excitations enable the sensorisation of a full-size 3D hand using electrodes positioned around its wrist. The thesis's experimental work culminates in the demonstration of a single-material multimodal soft skin, which holds great potential for the implementation of soft robotic sensors: the final chapter explores the vast future directions expected to stem from this work.Item Open Access Practical, High-Speed, Gaussian Modulated Coherent State Continuous Variable Quantum Key Distribution with Real-Time Post ProcessingWeerasinghe, WeerasingheQuantum key distribution is proven to achieve unconditional security based on the laws of quantum physics. Quantum key distribution was originally proposed using discrete variables. Later, continuous-variable quantum key distribution has been introduced. Continuous variable quantum key distribution with Gaussian modulated coherent states has gained interest due to its security and compatibility with classical coherent optical fibre networks. For successful system deployment, it is essential to implement practical, high-speed, systems that distils secret keys in real-time. Most existing demonstrations of continuous variable key distribution systems utilise bulky laboratory equipment such as arbitrary waveform generators and oscilloscopes to generate and record signals. To build practical systems that can be deployed in the field, it is important to investigate the use of commercially available digital-to-analogue converters and analogue-to-digital to converters for data generation at the sender and real-time data recording at the receiver respectively. Moreover, most demonstrations of Gaussian modulated coherent state continuous variable quantum key distribution to date, estimate the secret key rates without real secret key distillation. Therefore, it is crucial to build fully functional, practical systems with high-speed post processing toolchains that can distil secret keys in real-time. In this thesis, a Gaussian modulated coherent state continuous variable quantum key distribution system with a repetition rate of 50 MHz, the highest repetition rate for a similar practical system, is presented. This fully functional system consists of front-end optical hardware and the back-end real-time post processing toolchain. A real-time parameter monitoring software module is developed, which continuously measures excess noise and detects unstable operation immediately. With the measured excess noise, asymptotic key rates of 9.1 Mb/s, 6.8 Mb/s, 5.2 Mb/s, 3.8 Mb/s, 2.0 Mb/s, and 1.1 Mb/s are estimated for transmission distances of 15 km, 20 km, 25 km, 30 km, 40 km, and 50 km respectively. These are record asymptotic key rates for similar practical Gaussian modulated coherent state continuous variable quantum key distribution systems. Slice reconciliation is applied to convert continuous variables into bit strings within continuous variable quantum key distribution systems. In Gaussian modulated coherent state continuous variable quantum key distribution systems, even at moderately high signal-to-noise ratios observed after metropolitan transmission distances, the bit error rates remain high. High bit error rates result in high computational complexity for error correction. Most of the bit errors occur at the decision points applied within the Gaussian distribution of data during the slice reconciliation procedure. Therefore, in this thesis, for the first time, ‘guard bands’ are applied around these decision points, so that data lying within the guard bands can be discarded. This lowers the overall bit error rates at the expense of removing a small portion of the data blocks. An optimisation method to determine the widths of guard bands that maximises the final secret key rates is also presented. For instance, with optimised slicing, secret key rates (under finite-size effects and after privacy amplification) of 6.1 Mb/s, 4.9 Mb/s, and 3.7 Mb/s are achieved for transmission distances of 20 km, 25 km, and 30 km. These are record secret key rates for practical Gaussian modulated coherent state continuous variable quantum key distribution systems. Post processing for quantum key distribution consists of error correction and privacy amplification. Here, low density parity check codes are used for error correction, and fast Fourier transform-based method with circulant matrices are used for privacy amplification. Most previous studies on post processing focus on either error correction or privacy amplification without integrating them into a full continuous variable quantum key distribution system. Here, the back-end post processing toolchain is integrated with the complete system to distil secret keys in real-time. To achieve real-time post processing, the throughput rates of error correction and privacy amplification must support the raw asymptotic key rates achieved from the continuous variable quantum key distribution system. The asymptotic key rates achieved from the system presented here are on the order of a few Mb/s. Post processing speeds up to a few Gb/s have been achieved using graphical processing units and field programmable gate arrays in previous studies. However, in this practical system, more user-friendly central processing units are used. To achieve error correction speeds in the order of a few Mb/s, software techniques such as multi-threading and sparse matrix storage for parity check matrices are utilised on the central processing unit. Error correction speeds of up to 9.6 Mb/s have been achieved here. With the fast Fourier transform based privacy amplification method, throughput rates up to 9.3 Mb/s have been realised. These post processing speeds can support the record real-time secret key distillation for transmission distances ranging from 15 km to 50 km achieved from this system.Item Open Access Stability of thin-film PEDOT:PSS electrodes for neuromodulationOldroyd, PoppyImplantable electrodes that can reliably measure brain activity and deliver an electrical stimulus to a target tissue are increasingly employed to treat various neurological diseases and neuropsychiatric disorders. However, their long-term stability has yet to be proven in neuromodulation applications where electrical stimulation over months to years is desired. This thesis addresses this critical challenge. An accelerated aging platform enabled high-throughput screening of various electrode configurations. Metal-free PEDOT:PSS electrodes exhibited the best stability, with a failure rate of only 4.3% after 800 days. They significantly outperformed pristine and PEDOT:PSS-coated gold electrodes, which failed at rates of 89.7% and 82.1%, respectively. Delamination and gold corrosion were identified as key degradation mechanisms. Further aging revealed limitations with traditional polymeric encapsulation materials during stimulation. Introducing a PDMS elastomeric substrate significantly reduced device failure rates (<1%) and prolonged the device lifetime twenty-fold. This device architecture, consisting of stretchable PDMS/PEDOT:PSS electrodes fabricated using lithographic techniques, conformed to dynamic tissues like the gastrointestinal tract and simultaneously recorded electrical activity and mechanical strain. This holds immense potential for studying and treating complex gastrointestinal disorders. Beyond the gastrointestinal system, the long-term stability of metal-free PEDOT:PSS electrodes opens doors for chronic disease monitoring and treatment, eliminating the need for repeated surgeries. These electrodes represent advancements in both the nanofabrication, and bioelectronics fields. This expands prospects for long-term neuroscience studies such as mapping neuromodulatory pathways, improving our understanding of neural communication and treatment of neurological disorders.Item Open Access An Acoustic Pulse Electro-Kinetic SensorGlauser, Antony RobertThere are many circumstances under which it is important to study the biochemical or biological activity of a surface in contact with a fluid. Protein adsorption is of particular interest - the affinity of proteins for a surface will determine the suitability of the underlying material for use in anything from kitchen utensils to biomedical implants. A surface that sheds dirt easily may be effective in preventing food-poisoning, but one that sustains a coat of non-denatured proteins is vital for producing a successful and long-lasting vascular implant. Numerous techniques exist for studying protein-surface interactions, each having their own strengths and weaknesses. Established methods often rely on the prior modification of the protein with some kind of label (e.g. radioactive iodine, or a fluorescent marker). More recently, efforts have focussed on improving techniques which do not, such as Surface Plasmon Resonance. In this dissertation, a new real-time, label-free system is described, which has notable advantages over existing methods. In particular, it offers flexibility and experimental simplicity at a substantially lower cost than many of its counterparts. The method involves detecting tiny electrical signals generated at the solid-liquid boundary when ultrasound strikes at an oblique angle; the information so obtained is a function of the charge distribution and the visco-elastic properties of the fluid at the interface. Two distinct electro-kinetic effects are identified and characterized, referred to as Double-Layer Compression (DLC) and the Parallel Vibration Potential (PVP). A DLC signal is generated when the distance between fluid-borne ions and the solid surface to which they are attracted is modulated by the acoustic variations in pressure. The Parallel Vibration Potential is generated when the same ions are caused to slip over the surface, in an oscillatory fashion. Apparatus has been designed and constructed for characterizing these effects. One effect (DLC) is shown to be useful for studying the electrochemical nature of the solid surface itself. The other (PVP) yields information on the type and density of proteins adsorbed at the surface; both protein-surface and protein-protein interactions have been monitored in this way.Item Open Access Dynamic Fleet Maintenance ManagementCrespo Del Castillo, Adolfo; Crespo Del Castillo, Adolfo [0000-0002-1151-3253]The increased prevalence of digitalisation in today’s industry allows a comprehensive understanding of the value of the assets for the organisations. In this context, new technologies enable predictive maintenance to cover all activities from data acquisition and processing to maintenance decision-making advisory as output. This particularly expands to a wider view of ‘health management’ as opposed to a focus solely on maintenance at the fleet level. Consequently, decisions related to workload determination and operational scheduling must be aligned with asset condition assessments and maintenance strategies. Traditional fleet maintenance strategies are often reactive or rely on predetermined schedules, which can lead to inefficient resource allocation and increased operational costs. The paradigm shift towards a data-driven approach enables fleet management to dynamically respond to issues identified through sensors and algorithms insights. Furthermore, it allows fleet management to make integrated maintenance and operations decisions. However, despite the rapid technological change, a notable deficiency exists in the integration of predictive maintenance with predetermined preventive maintenance, existing limitations for maintenance resources in each depot, and fleet operational scheduling. This is due to the lack of a holistic strategic approach, and a criterion to stablish the optimal solution, thereby impeding value creation for businesses. The thesis presents a three stage model that (i) defines the operating context and maintenance resources (ii) evaluates feasible opportunistic maintenance timeslots to integrate predictive maintenance and (iii) allocates the assets of the fleet to operation, preventive and predictive maintenance, or being idle for a certain planning horizon. In addition, the thesis explores how the optimal allocation, and hence the value of this integrated approach is affected by the criticality of the components that are monitored, the quality of RUL prognosis, and the balance of maintenance costs and service risks. The applicability of the approach is demonstrated through a case study using a real industrial scenario of a fleet of high speed trains from Talgo in Spain. The thesis concludes that the proposed approach presents a holistic solution that allows to solve the research gap and integrate predictive maintenance scheduling with preventive maintenance, depot resources, and workload balance. Generally, highly critical assets justify the integration of predictive maintenance into the decision making process, as the savings considering the risk of failure accurately would compensate the cost of condition monitoring. This was proven with the model, and besides, components of medium criticality can also be quantified and justified for condition monitoring. Furthermore, nowadays due to the high demanding service contracts fleet managers may assume higher risks in order to fulfil service. This challenging balance is quantified by the model proposing the optimal fleet scheduling recommendation.