Publications

Recent legislation banning the sale of new petrol and diesel vehicles in Europe from 2035 has shifted the focus of internal combustion engine research towards alternative fuels with net zero tailpipe emissions such as hydrogen.

Research regarding hydrogen as a fuel is particularly pertinent to the so-called ‘hard-to-electrify’ propulsion applications, requiring a combination of large range, fast refuelling times or high-load duty cycles. The virtual design, development, and optimisation of hydrogen internal combustion engines has resulted in the necessity for accurate predictive modelling of the hydrogen combustion and autoignition processes.

Typically, the models for these processes rely respectively on laminar flame speed datasets to calculate the rate of fuel burn as well as ignition delay time datasets to estimate autoignition timing. These datasets are generated using chemical kinetic mechanisms available in the literature. However, these mechanisms have typically been developed with a focus on hydrocarbon oxidation – e.g., syngas, natural gas, biofuels, diesel, and gasoline - and their validation datasets feature a very limited number of hydrogen-specific targets.

Therefore, this study explores the predictive capability of six commonly used chemical kinetic mechanisms over a large dataset consisting of hydrogen-specific ignition delay time and laminar flame speed targets compiled using data available in the literature. Additionally, a sensitivity analysis was conducted to identify reactions that strongly affect the ignition delay time of hydrogen-air mixtures in the intermediate-temperature regime, where large ignition delay time deviations are observed compared to experimental results. The sensitivity analysis was followed by an exploratory study in ad-hoc mechanism adjustment.

Hydrogen combustion can decarbonise difficult-to-abate sectors. However, practical deployment depends on reliable prediction of combustion behaviour under transient conditions, which contrasts with the steady-state experiments typically used for combustion mechanism development. This study presents a fully optimised H2-NOx mechanism, calibrated against 118 fundamental combustion datasets containing 1695 datapoints, which shows significant improvements in the prediction of ignition onset in an internal combustion engine with nitric oxide injection into the intake system.

In contrast to prior single-objective approaches, this study introduces a fundamentally new approach to chemical kinetic mechanism optimisation, which leverages a Multi-Objective Particle Swarm Optimisation framework on a High-Performance Computing platform. The framework simultaneously balances accuracy and consistency across datasets, explicitly incorporates experimental uncertainty, and evaluates all candidate mechanisms with full chemical simulations. Prediction accuracy is quantified using the normalised root mean square error (nRMSE) to experimental measurements and the proportion of predictions within experimental uncertainty limits. Relative to the best existing mechanism, the optimised model achieves a 35 % reduction in nRMSE and a 19 % increase in the number of predictions within uncertainty bounds, demonstrating improved predictive performance for fundamental combustion targets.

When the optimised mechanism was applied to autoignition timing in a Homogeneous Charge Compression Ignition engine, significant improvements were found for data with nitric oxide. Nevertheless, the overall accuracy in autoignition prediction is insufficient for practical applications, indicating that transient engine conditions are not adequately represented by steady-state datasets. These findings underscore that even fully optimised mechanisms based solely on fundamental experiments will not deliver high-accuracy predictions under real-world, transient conditions and integration of transient combustion data into future development of chemical mechanisms is recommended.