Data and Computational Materials for “Physics-Guided and Uncertainty-Aware Surrogate Modelling of NH₃/H₂ Combustion: Multi-Output Prediction of Reactivity, NOx, N₂O and Ammonia Slip under Kinetic and Operating Uncertainty”
Description
This dataset contains the computational data, processed analysis outputs, reproducibility materials and supporting results associated with the manuscript “Physics-Guided and Uncertainty-Aware Surrogate Modelling of NH₃/H₂ Combustion: Multi-Output Prediction of Reactivity, NOx, N₂O and Ammonia Slip under Kinetic and Operating Uncertainty.” The study integrates experimental benchmarking of NH₃/H₂ kinetic mechanisms, multi-mechanism Cantera simulations, machine-learning surrogate modelling, out-of-distribution validation, grouped conformal uncertainty quantification, machine-learning and elementary-reaction sensitivity analysis, and uncertainty-aware multi-objective optimization. The computational production dataset comprises 99,057 complete machine-learning-ready simulation rows generated using three validated kinetic mechanisms: 45,000 ignition-delay simulations, 45,000 jet-stirred-reactor simulations and 9,057 complete freely propagating flame simulations. The repository includes processed computational outputs and supporting analyses for experimental mechanism validation, mechanism-disagreement quantification, surrogate-model benchmarking, hydrogen- and pressure-OOD evaluation, leave-one-mechanism-out transfer, conformal prediction, five-seed neural-model ablations, controlled hydrogen-response analysis, kinetic sensitivity, mechanism-disagreement/reliability analysis, five-seed NSGA-II optimization, direct kinetic verification and numerical-convergence testing. Experimental benchmark observations originate from the published studies cited in the associated manuscript. Users should consult the original publications for source-specific experimental data rights and provenance. The accompanying README describes the repository organization and the relationship between repository files, manuscript results and Supplementary Information.