The transition towards low-carbon energy systems has deepened interest in hydrogen as a supplementary or alternative fuel for internal combustion engines, particularly within the heavy-duty transport sector, where rapid electrification remains challenging. In this context, hydrogen–diesel dual-fuel (HDDF) combustion in compression ignition (CI) engines offers a reasonable transitional pathway, enabling significant emission reductions while leveraging existing diesel engine infrastructure. The overarching aim of this research was to systematically investigate the performance, combustion, and emissions characteristics of HDDF CI engines and to develop robust optimisation and prediction frameworks to support their efficient and reliable deployment. To achieve this aim, the study adopted a multi-scale and multi-disciplinary numerical methodology. Initially, a detailed one-dimensional (1D) engine model was developed and validated to evaluate the influence of hydrogen substitution level, engine speed, and exhaust gas recirculation (EGR) on brake power, torque, brake thermal efficiency (BTE), fuel consumption, and emissions. The analysis was extended to assess the impact of different hydrogen injection strategies at a fixed hydrogen energy share (HES), enabling a comparative evaluation of performance–emissions trade-offs. Subsequently, a high-fidelity three-dimensional (3D) computational fluid dynamics (CFD) framework was established to investigate the role of geometric parameters, including bore size, stroke length, and compression ratio, under varying HES and load conditions. The results demonstrate that hydrogen enrichment can significantly enhance engine performance at higher speeds and loads, while reducing carbon-based emissions; however, challenges remain in terms of combustion stability at low load and increased nitrogen oxides (NOₓ) formation at high hydrogen substitution. Injection strategy and engine geometry were shown to play critical roles in governing these trade-offs, with stroke length and compression ratio emerging as key design levers for balancing efficiency and emissions. To complement physics-based modelling, an interpretable data-driven framework based on Sparse Identification of Nonlinear Dynamical Systems (SINDy) was developed for engine performance prediction across a wide operating envelope. Coupled with evolutionary optimisation and multi-criteria decision-making, the proposed framework achieved high predictive accuracy while retaining physical interpretability. The original contributions of this thesis lie in the systematic generation of new insights into hydrogen–diesel dual-fuel CI engine behaviour through coordinated analysis of operating conditions, injection strategies, and engine geometry, together with the application of an interpretable SINDy-based framework to capture nonlinear performance and emissions relationships. The research establishes quantitative performance and emissions trade-offs and optimisation pathways that clarify how hydrogen can be effectively integrated into CI engines. Collectively, the findings advance understanding of hydrogen utilisation and provide practical guidance for the design and optimisation of low-emission engines.