On the Impact of CFD Turbulence Models for Premixed $${\text{NH}}_{3}$$/$${\text{H}}_{2}$$ Combustion on Emissions and Flame Characteristics in a Swirl-Stabilized Burner

Ammonia combustion is gaining interest as a feasible alternative to traditional fossil fuels because of to the low environmental impact and as hydrogen and energy carrier. This study used Computational Fluid Dynamics (CFD) simulations to compare various turbulence models for premixed ammonia/hydrogen combustion in a swirl-stabilized burner. The primary aim was to identify the best turbulence model for accurately predicting the flow dynamics, combustion behaviour, and emissions profiles of ammonia/hydrogen fuel blends. The turbulence models evaluated were Large Eddy Simulation (LES), Realizable k- $$\epsilon$$ ϵ , Renormalization Group (RNG) k- $$\epsilon$$ ϵ , k- $$\omega$$ ω SST, and Reynolds Stress Model (RSM). On the LES side, a further comparison of two subgrid models (Smagorinsky-Lilly and WALE) was investigated. The Flamelet Generated Manifold (FGM) method was utilized with a detailed chemistry scheme taking into consideration all $$\hbox {NO}_x$$ NO x reactions. To improve the prediction of $$\hbox {NO}_x$$ NO x emissions, additional scalar transport equations for NO and $$\hbox {NO}_2$$ NO 2 were included. This methodology aimed to be a balance between computational efficiency and the accuracy expected of detailed chemistry models. Validation was done with a swirl burner from Cardiff University’s Gas Turbine Research Centre. Results showed that all turbulence models accurately captured flame characteristics in terms of exhaust temperature and axial velocity with minor differences in the recirculation zones, where only the RSM model can predict the velocity trend as the LES simulation while other RANS models differ by at least 7 m/s. The temperature reached by the LES resulted 100 K higher than the other models in the flame zone. LES simulation can predict the emission value with an error of less than 10 $$\%$$ % . Moreover, the error related to emissions derived from the RANS simulations was not negligible, underestimating $$\hbox {NO}_x$$ NO x emissions by about 35 $$\%$$ % . However, RSM model produced results that were closer to those derived from the high-fidelity LES when compared to the others RANS models, particularly in terms of flame thickness and emissions. It was concluded that it is mandatory to perform an unsteady analysis to reach reasonable results.