Large eddy simulation of bluff-body turbulent hydrogen/nitrogen flames using principal component transport models with differential diffusion effects

This study investigates the combustion characteristics of nonpremixed bluff-body stabilized H2/N2-air flames, with a specific focus on a fully cracked ammonia case. The motivation behind this exploration lies in the potential of such flames to contribute to low-carbon combustion technologies, serving as both a reference and a first step towards understanding the role of the recirculation zone on the in-situ cracking of ammonia bluff-body stabilized flames. Simultaneous Raman/Rayleigh measurements of temperature and major species are utilized to validate the numerical model, offering a detailed characterization of the flame structure. To achieve computational efficiency at high fidelity, the simulation algorithm employs a data-driven approach that combines principal component analysis (PCA) with deep neural networks (DNNs). The model incorporates differential diffusion and subgrid-scale effects. The proposed PC-DNN approach includes differential diffusion based on a rotation technique, utilizing the mixture-averaged (MA) transport model for the training dataset. To show the potential of the PC-DNN approach in modeling fully cracked ammonia flames, the study compares large eddy simulation (LES) results with recent Raman/Rayleigh scattering measurements obtained at KAUST. The findings reveal that the PC-DNN method accurately captures key flame features with just two principal components. Incorporating differential diffusion and subgrid-scale effects enhances predictions, despite some inconsistencies in fuel-rich areas, and in the far field of the burner exit where mixing is predominant, suggesting further room for improvement in the model. This research sheds light on the complex dynamics of fully cracked ammonia flames, providing valuable insights for advancing low-carbon combustion technologies.