Multi-Fidelity Modeling of a Lean Premixed Swirl-Stabilized Hydrogen Burner with Axial Air Injection

Premixed hydrogen burners offer promising results in reducing pollutant emissions but are susceptible to flashback, posing significant safety risks and requiring high experimental costs. This study introduces a multi-fidelity modeling approach to address the challenges posed by the scarcity of high-fidelity data, leveraging the assumption of a linear correlation between high-fidelity and low-fidelity data. The model is tested on predicting the axial flame distance from the mixing tube, an indicator of flashback susceptibility, in a lean premixed swirl-stabilized hydrogen burner. Experimental results serve as high-fidelity data, while 2D steady axisymmetric RANS simulations provide low-fidelity data. The results demonstrate the potential of 2D RANS to approximate burner behavior accurately and the capability of the multi-fidelity model to enhance low-fidelity predictions with a severely limited set of training points.