Theses

Alex's doctoral thesis is concerned with the development of finite element methods for coupled systems of partial differential equations (PDEs) relating to problems in charge transport. The main focus is on structure-preserving approaches, which imitate at the discrete level properties possessed by the governing PDEs. This can include phenomena such as mass conservation, preserving the positivity of density variables, and energy decay laws. Aside from being desirable from a physical interpretation standpoint, structure-preserving methods can also benefit from inherent stability properties as a result of working within the natural framework of the problem.

In recent years, electrification of automotive powertrains has been increasingly considered and implemented as a key element in attaining new affordable, more efficient, and potentially sustainable forms of transportation. Since they represent a defining element of any electrified vehicle, energy storage systems are receiving an increasing amount of attention from the research community as well as the automotive industry. This is most often aimed towards Li-ion batteries and pursues to improve their performance and lifespan while reducing cost and environmental impact through advancements in the development and manufacturing of the system as well as its operation and recycling.

One of the most effective tools in supporting all lifecycle stages of a battery system is represented by simulation models, ranging from static data-driven representations to complex multidimensional virtual depictions of electrochemical cells. Due to the trade-off between the resources required to develop and the ability to replicate the voltage response as well as the vast array of applications, the equivalent circuit model is one of the most popular options employed for simulating battery systems. While the usefulness of any model is always determined by its structure, accurate and effective parametrisation of the structure in relation to the system to be replicated also represents a highly important prerequisite for all applications.

The aim of this thesis is to discuss the development and evaluation of a novel parametrisation methodology suitable for equivalent circuit models and exemplified for the Li-ion battery use case. The research presented fills a substantial research gap for accurate and effective procedures capable of parametrising the specified model in an automated manner. These attributes reliably allow the use of unconstrained time-domain datasets capturing the voltage response of the system and further strengthen the ability to complement the dataset by using prior models as alternative sources of information.

The development and evaluation of the proposed methodology were supported by a readily available equivalent circuit model supplied by the industrial sponsor. The first part of the research focuses on the assembly of the procedure, which creates an estimate of the model and iteratively improves it by refining the values of its parameters through a series of steps. The second part evaluates the resulting methodology using a wide array of cases defined by the time-domain data and the prior model used for training. The two main aspects investigated are the ability of the parametrisation procedure to efficiently convert time-domain information into an accurate model as well as its ability to reduce the requirement for data by using information supplied by the prior model.

The accuracy and computational efficiency demonstrated endorse the proposed methodology as a state-of-the-art solution. The process was highly consistent in achieving model MSE values below 1E−3 V² across large individual data files used for training as well as validation but also presented values as low as 2.83E−5 V². To obtain an accurate evaluation of the computational efficiency, a partially streamlined version of the proposed methodology was implemented. Despite the non-optimised code, the parametrisation of 409 individual model parameter values distributed across 6 lookup tables using 24.543 hours of data was completed in 999.1 seconds. A comparison study published by Savu et al.[1] in a separate article targeting smaller scale versions suitable for small segments of data demonstrates that the proposed parametrisation principle requires 75 times less computational effort than a readily available global solver, while the expansion to the complete model is also likely to increase this factor exponentially. Further work aimed at computational time optimisation also suggests a margin of improvement, potentially by an order of magnitude, that could be achievable through optimisation of the code and parallelisation techniques.

Outside accuracy and computational efficiency, the results also support the ability of the proposed parametrisation methodologies to include prior information towards improving interpolation and extrapolation capabilities of the model outside its trained regions as well as reducing the amount of data required and, implicitly, testing efforts. To exemplify, a model representative of a Samsung 30Q cell was used as prior information for the parametrisation of a model set to replicate a similar Sony VTC6 cell. An accuracy improvement of over 80% that of a model parametrised using the complete training dataset was achievable by using as low as only 31% of the data in the presence of the informed prior. The consistency of the result was also confirmed by reverted cases employing the model parametrised for the Sony VTC6 as the prior information for models replicating the Samsung 30Q cell. Key factors determining the effectiveness of the technique include the level of similarity between the system producing the prior and the system to be modelled, as well as the distribution of the data relative to operating conditions.

The outcome of the research presents a significant positive impact on multiple elements associated with simulation models. The methodology provides a gateway to efficiently obtaining models in an automated manner with limited time, effort and resources required, hence supporting their adoption in research and development activities as well as frontloading these activities in virtual environments. The process also represents a showcase of compiling unconstrained (‘as found’) datasets into accurate models, reducing the requirements imposed on training data and becoming directly suited to effectively using data lakes. The ability to include prior models as alternative sources of information will enhance the model obtained while employing only an unconstrained dataset and will also reduce the amount of data required and, implicitly, testing efforts.

Lastly, the structure of the methodology can also be interpreted as a prior model update process using new data in a computationally efficient manner, which makes it highly compatible with the concept of digital twins. 

[1] V.-I. Savu, C. Brace, G. Engel, N. Didcock, P. Wilson, E. Kural, and N. Zhang, “Linear Regression-based Procedures for Extraction of Li-ion Battery Equivalent Circuit Model Parameters,” Batteries, Vol. 10, No. 10, 343, 2024, doi: 10.3390/batteries10100343

The output from this PhD project offers a valuable contribution to the transportation research and policymaking fields by providing a location-independent methodology and computational model that consider the geographical complexity of the transport system and determine its distributed environmental and socioeconomic impacts across a region. iTRIPP is a replicable, flexible, representative and scalable methodology contributing to the automated simulation of multiple different regional transport strategies. Moreover, by providing the outputs in numerical and mapped formats, this model will facilitate the analysis and dissemination of the results and, therefore, assist with decision making and public consultation processes on proposed transport strategies.

The automotive industry is a major consumer of plastics, relying on virgin polymers from dwindling crude oil resources and contributing significantly to climate change. While some biopolymer–fibre composites have been adopted, these are often derived from food crops, raising sustainability concerns. Terpenes, by contrast, are abundant, lightly oxygenated hydrocarbons with simple scaffolds that can be integrated into existing infrastructures and, within a biorefinery framework, provide diverse value-added products.

Due to the environmental concerns, strict regulations and requirements enforced by legislative bodies, the aviation industry is actively looking to undertake a transition from kerosene fuelled aircraft to zero-emission aircraft. Numerous options have been evaluated to identify a suitable candidate to replace kerosene.

Hydrogen was identified as a potential candidate to replace and assist kerosene in future aircraft powered by cryogenic propulsion systems. Hydrogen is an abundant element which is found naturally in gas form. Gaseous hydrogen (at 700 Bar and atmospheric temperature) has around three times higher gravimetric energy density compared to kerosene (at atmospheric conditions) [1] [2] [3]. However, the volumetric energy density of gaseous hydrogen (at 700 Bar and atmospheric temperature) is lower than kerosene (at atmospheric conditions) [1] [2] [3]. To overcome this, hydrogen must be stored as a liquid at 20 K. In liquid form (at atmospheric conditions), the volumetric energy density of hydrogen is around four times lower than kerosene (at atmospheric conditions) [1] [2] [3]. This means that compared to kerosene, to travel the same distance with a hydrogen fuelled aircraft, around four times larger storage tanks are required.

The storage tanks must also be compatible to operate under cryogenic conditions. Moreover, sufficient insulation is required to prevent any heat entering the storage tanks leading to evaporation and loss of hydrogen. Finally, the tanks must be designed to prevent components apart from the fuel system encountering with hydrogen.

Road safety has been a persistent challenge in the mobility sector for over a century, with road traffic crashes (RTCs) accounting for significant numbers of fatalities and injuries worldwide each year. Beyond the devastating human toll, RTCs impose a substantial socioeconomic burden on individuals, their families, and national infrastructure through vehicle damage, medical expenses, and the loss of productivity. A substantial body of research indicates that human factors such as fatigue and distraction are primary contributors to these incidents, prompting a growing interest in the development of Driver Monitoring Systems (DMS) aimed at improving driver attentiveness and road safety.

Incumbent DMS technologies predominantly rely on vehicle dynamics and observable driver behaviours, often captured internally from the vehicle itself or externally via an in-cabin camera. Despite their reasonable effectiveness in driver distraction detection, these systems typically lack the capability to directly measure a driver’s physiological state – a factor that is proven to be strongly correlated with fatigue, resulting in limited robustness and accuracy when it comes to driver fatigue detection, particularly under real-world driving conditions. Hence, real-time monitoring of physiological data has the potential to improve driver fatigue detection and reduce the resultant accidents. However, conventional methods of physiological measurements often require body-attached electrodes and are designed for controlled, quasi-static scenarios, contrasting to the noisy and highly dynamic driving environments. Moreover, the use of body-attached electrodes will further introduce distractions and restrict body movements, all making them impractical for integration into real-world DMS applications.

To address this gap, this thesis focuses on the development of the noncontact driver attentiveness detection system by advancing noncontact driver physiological measurements techniques under driving conditions, ultimately contributing to the unresolved fatigue detection challenge in the current DMS regime. Noncontact physiological sensing in real-world driving conditions presents significant challenges due to factors such as the variable illumination, motion artefacts, and environmental interference found in vehicular environments. The work begins with a comparative study of the two leading noncontact modalities in this field – computer vision and radar – for heart rate monitoring, highlighting the characteristics of each modality, and identifying radar as a more robust alternative under realistic driving scenarios. On this basis, the thesis then explores the use of Frequency Modulated Continuous Wave (FMCW) radar, which is the most promising radar type in vehicular sensing, for noncontact driver cardiorespiratory monitoring, analysing the key challenges such as motion cancellation trade-offs, phase ambiguity, and radar cross-section (RCS) variability restricting its practical implementation and the corresponding reasons through implementation of state-of-the-art techniques. To address some of these issues, a novel noncontact cardiorespiratory monitoring system for drivers using FMCW radar is proposed, incorporating a phase continuity tracking algorithm and a signal quality index, enabling continuous and reliable extraction of physiological signals for drivers even under typical driving movements and environmental constraints. Recognising driver behaviour represent the state-of-the-art indicators for driver fatigue detection, the final part of the thesis compares the two modalities and complements the newly enabled physiological measurements with behavioural analysis, providing insights for multimodal DMS development.

Overall, this research addresses the critical challenges for driver fatigue detection in existing DMS technologies by enabling accurate, unobstructive, and real-time physiological measurements and linking them to existing behavioural domains. The work presented provides a foundation for next-generation DMS, thereby contributing to the long-standing road safety challenge.

Accurate state of health (SoH) estimation is essential for advancing lithium-ion battery technology and optimizing operating strategies. However, this remains challenging as internal ageing mechanisms are not fully understood, difficult to measure, and require time-intensive experiments. This work addresses these challenges by proposing a novel methodology that improves the accurate parametrisation of physics based models and presents an efficient design of experiments.

An in-depth experimental ageing study demonstrates that degradation is usage-history dependent, with prior operating history influencing future behaviour and the onset of nonlinear ageing phenomena such as the knee-point. Depending on conditions, the knee-point was delayed by up to 50\% (400 full-cycle equivalents) or did not occur at all. Reversible short-term capacity fade and recovery are also observed, diminishing with over the cell's lifetime. These findings highlight the complexity of ageing processes and confirm that capacity alone is an inadequate health metric.

Incremental capacity (IC) analysis is evaluated as a non-invasive diagnostic tool, with simulations and experiments showing that IC-derived features provide deeper insight into degradation and can be linked to specific mechanisms, offering reliable indicators of SoH ($R_{\text{ave}}^{2}$ = 0.96). These features are leveraged to improve the parametrisation of physics-based degradation models, addressing identifiability issues and enabling their integration into a hybrid machine learning framework for robust SoH estimation.

To identify the most informative operating points for model training, a novel design of experiments methodology was developed. This approach combines synthetic datasets, an IC-based surrogate model, and data pruning to reduce the number of test cells required by 86\% (to 35 cells), thereby lowering experimental cost and duration. Overall, this work advances SoH estimation by integrating diagnostic techniques, physics-based modelling, and efficient experimental design.

The automotive industry is undergoing a paradigm shift from hardware-centric engineering practices to software-defined, Cyber-Physical Systems (CPSs). This evolution is driven by consumer demand for increasingly automated and intuitive vehicle features, such as Advanced Driver Assistance Systems (ADAS) and Autonomous Driving (AD), alongside the need for faster time-to-market and continuous functional updates. These trends have introduced unprecedented levels of system complexity, challenging traditional development and validation methodologies. In response to this paradigm shift, this thesis presents a model-based framework for managing complexity and enabling automated testing in the development of automotive CPSs.

The research uses Model-Based Systems Engineering (MBSE), which replaces traditional document-centric approaches with structured, queryable system models. A central contribution of this research is the development of a set of metrics computed by transforming static SysML models into Functional Dependency Graphs (FDGs), enabling graph-theoretic analysis of the system’s architecture. These include System Complexity, System Modularity, and System Test Effort.

Building on this evaluation, the research introduces a model-driven approach to test planning and execution. Falsification, where the system is deliberately challenged with scenarios designed to expose weaknesses or violations of requirements, is applied to facilitate the automatic generation of high-impact test cases, improving coverage and reducing the likelihood of undetected faults. Test prioritisation is introduced for large systems and guided by two indicators: the Function Maturity Rating and the Risk Rating, to manage large and complex systems. These metrics ensure that testing efforts are focused where they are most needed, minimising bias and improving Verification and Validation (V&V) efficiency.

Together, these contributions form a methodology for enhancing MBSE in the development of complex automotive CPSs. By bridging the gap between high-level system modelling and low-level testing, the framework supports early problem detection, continuous validation, and informed decision-making

Stricter emissions regulations are accelerating the shift toward “net-zero” powertrains, including battery-electric, fuel-cell and zero-carbon-fuelled internal combustion engine powertrains. In hard-to-abate sectors such as heavy-duty transport, marine, and construction, hydrogen and ammonia have emerged as promising alternatives due to their potential for carbon-neutral combustion and compatibility with existing internal combustion engine infrastructure.

Effective use of these fuels requires high-efficiency engines, the development of which depends on predictive combustion models. These models rely on chemical kinetic mechanisms to accurately simulate key processes such as flame propagation, autoignition, and species evolution. Nevertheless, chemical kinetic mechanisms are generally validated against experimental data at a broad range of thermochemical conditions and few studies in the literature focus on developing mechanisms specifically for engine applications – relevant pressures, temperatures, mixture dilutions and stoichiometries. Therefore, this remains an underexplored research area that can deliver significant improvements in the predictive accuracy of engine models.

This thesis focuses on improving the predictive accuracy of chemical kinetic mechanisms by developing a fundamentally new approach to chemical kinetic mechanism optimisation and its application to the optimisation of mechanisms for hydrogen and ammonia. The optimised mechanisms were further validated against experimental engine data.

A comprehensive data collection of hydrogen experimental data at engine-relevant thermochemical conditions was assembled and used to identify the literature mechanism that observes the best predictive fit to the data. A new optimisation framework was then developed, which utilises a multi-objective particle swarm optimisation algorithm to search the uncertainty hyperspace of selected pre-exponential rate coefficients, aiming to reduce simulation errors, while maximising the number of experimental datapoints modelled within experimental uncertainty. The optimised mechanism, “Bath-H2”, achieved a 35% reduction in the normalised root mean square error between model predictions and experimental measurements, while increasing predictions within experimental uncertainty by 19% compared to the baseline mechanism. Post-optimisation flux analyses revealed changes in the production and consumption pathways of hydroxyl radicals that led to the improved prediction of experimental data at conditions observed in internal combustion engines.

The optimised mechanism was further validated against experimental data from a hydrogen-fuelled Cooperative Fuel Research (CFR) engine operating in HCCI mode. The modelling results revealed that the mechanism matches the predictive accuracy of literature mechanisms for conditions without NO injection in the intake manifold, while significantly outperforming them in the cases involving NO addition. Uncertainty analyses showed that the optimised mechanism predicts autoignition within experimental uncertainty.

The optimisation framework was further applied to the development of an ammonia oxidation mechanism by including additional experimental data on hydrogen-ammonia blends and ammonia-only systems. The optimised mechanism, “Bath-NH3”, exhibited a 2.9% reduction in the normalised root mean square error between model predictions and experimental measurements at the expense of a 1.0% decrease in the number of predictions within experimental uncertainty against the baseline mechanism. Post-optimisation flux analyses revealed changes in the production and consumption pathways of OH radicals that led to the improved prediction of experimental data at intermediate temperatures and pressures for hydrogen-ammonia fuel blends. Further validation against the CFR experimental engine data revealed that Bath-NH3 predicted autoignition in the CFR engine with increased accuracy compared to the baseline.

The growing cost and time requirements of powertrain design and optimisation, due to the increasing complexity of vehicle powertrains, put an ever-greater emphasis on the need for truly predictive combustion models to streamline powertrain development processes. This work highlights the importance of chemical kinetic mechanisms for the progression of predictive combustion modelling, develops two mechanisms with improved predictive accuracy and identifies several promising directions that can enhance both the applicability of the optimised mechanisms and the efficiency of the optimisation framework presented.

It has been demonstrated that electric buses are a low carbon alternative to traditional diesel Internal Combustion Engine (ICE) vehicles – primarily due to their zero tailpipe emissions and less carbon intensive fuel source. For this reason, local and national authorities have been keen to expand their rollout. However, despite evidence of lower operating expenditures from energy consumption and vehicle maintenance, the UK government has relied on several rounds of generous capital grants alongside a higher rate of Bus Service Operators Grant (BSOG) to encourage operators to electrify their fleets.

In this thesis we investigate the evolution of the economic barriers to adoption that explain the private sector’s hesitance to decarbonise their fleets without financial support. Specifically, through the use of Total Cost of Ownership (TCO) simulations and financial statement modelling, we find that operational complexity (relating to vehicle scheduling, charging, and infrastructure requirements), fleet sizing, and vehicle maintenance are all factors with material impact on the economic cost and financial risk of fleet electrification. By analysing which barriers may remain or diminish, we translate our findings into policy implications relevant in the context of a once novel technology that has since matured and continues to experience improvements in cost and performance.

Our outlook suggests that the economics of operating electric bus fleets has evolved such that the UK government can begin to reduce and withdraw financial support over the coming years. However, policymakers must do so while being more conscious of the heterogeneity between operators. Policies effective in relation to urban services provided by large operators may leave small and/or rural operations behind.