Optimisation of the experimental information for kinetic modelling of mineral carbonation reactions

The cement industry contributes 5–8% of global anthropogenic CO2 emissions (2.7–2.9 Gt annually), requiring urgent decarbonisation to meet the Paris Agreement target of at least 16% emission reduction by 2030. Mineral carbonation of cementitious waste offers permanent CO2 sequestration while producing supplementary cementitious materials (SCMs) that partially replace clinker. However, carbonation experiments are costly, time-intensive, and limited by destructive sampling, while existing modelling approaches neglect systematic optimisation of experimental information. This thesis addresses that gap by developing digital tools, experimental design methodologies, and mechanistic models to maximise information from limited data. An open-source Python package, MIDDoE, is developed to implement Model-Based Design of Experiments (MBDoE) for model discrimination and parameter precision. A key contribution is constraint-adaptive experimental design, ensuring information-optimal experiments remain practically feasible under equipment limitations. Two novel MBDoE criteria are introduced to address ill-conditioned kinetic systems with strongly correlated parameters, either by sequentially deflating the sensitivity matrix or promoting uniform information distribution. Both outperform classical criteria under equal experimental budgets, enabling identification of more parameters per experiment. Using this framework, aqueous carbonation of industrial recycled concrete fines (RCFs) is studied at 25–85 °C under severe data constraints. A modified shrinking core model based on parabolic diffusion shows superior predictive performance among solid-state formulations. Estimability analysis prevents overparameterisation, while global sensitivity analysis identifies particle size and CO₂ partial pressure as dominant variables, with temperature showing negligible net influence. Carbonation achieves 81% efficiency (95 kg CO2 per tonne), and the carbonated products meet compressive strength requirements at 10% cement replacement. Despite good predictive ability, the solid-state model oversimplifies aqueous chemistry and depends on destructive calibration. To address these limitations, an ionic-state mechanistic model is developed, explicitly resolving aqueous chemistry through particle-scale diffusion, thermodynamically consistent dissolution kinetics, and carbonate equilibria. This formulation accurately reproduces experimental trajectories and, when calibrated solely with pH data, predicts carbonation efficiency from continuous pH monitoring, eliminating reliance on thermogravimetric analysis and increasing information density at lower cost. Overall, the thesis demonstrates that reliable kinetic model identification under resource constraints is achievable through systematic optimisation of experimental information. By integrating constraint-aware design, robust MBDoE criteria, estimability analysis, and mechanistic reformulation, a coherent strategy for efficient model development is established, with applicability beyond mineral carbonation to broader reactive systems governed by coupled dissolution and precipitation processes.