Mechanism-Reduction-Driven clustering of the operating parameter space for NH3/Hz premixed flames

The increasing need for clean and efficient combustion processes has intensified the exploration of alternative fuels such as ammonia (NH3) and hydrogen (H2). While these fuels offer zero carbon emissions, understanding their complex combustion kinetics remains a challenge, with special concerns about NOx production. This work presents a systematic approach for generating and selecting skeletal mechanisms for premixed flames of NH3/H2 blends across numerous operating conditions, reducing the complexity of detailed kinetic models while preserving accuracy in key observables. Given the vast operating parameters space (blending ratio, equivalence ratio, dilution, pressure), our goal is to identify a minimal set of skeletal mechanisms that offer an acceptable size/ accuracy trade-off over as wide as possible regions of such space. We employ Cantera for strained counterflow flame simulations under multiple operating regimes and PyCSP for threshold-based and minimal-user-expertise mechanism reduction, followed by constrained agglomerative clustering to identify representative skeletal mechanisms.