Multi-regime combustion modelling in high-fidelity numerical simulations of reacting flows
Nov 04, 2024
Anurag Surapaneni defended his thesis supervised by Dr. Daniel Mira on October 30, 2024 at the Barcelona Supercomputing Center (BSC). Entitled "Multi-regime combustion modelling in high-fidelity numerical simulations of reacting flows" the thesis focuses on developing high-fidelity numerical combustion models for simulation reacting flow with mixture inhomogeneity and under complex combustion regimes.
The quest to reduce carbon-based emissions has led to the use of alternative fuels and combustion systems. The aviation industry presents unique challenges due to the need for high-energy fuels, with e-fuels and hydrogen among the early candidates to facilitate net-carbon neutrality. To this end, LES simulations of turbulent reacting flows have emerged as key design and analysis tools. Capturing the complexities of realistic burner configurations poses a multi-scale, multi-physics problem. The inherent flow complexities and multi-fuel systems often result in practical systems operating in multi-regime combustion, a consequence of mixture inhomogeneity that promotes a complex response of burning rates and pollutant formation. It is therefore fundamental that the models reacting flows capture this complex response. Combustion can be solved using a variety of methods that balance generality and computational cost. Finite rate chemistry methods are the most general but also prohibitively expensive, while manifold-based methods offer a feasible option but with limited generality. This study addresses the problem of multi-regime combustion from both perspectives and advances the current state-of-the-art in these methods.
Acceleration in the FRC solver is achieved by focusing on the most computationally expensive task: chemical integration. The two fold strategy reduces chemistry and utilizes solvers that take advantage of the reduction. After briefly discussing chemistry reduction methods, a novel dynamic adaptive chemistry (TRAC), which is based on the dynamic tabulation of reactions in a low-order manifold space is presented. TRAC is analyzed in canonical combustion problems, where a speedup of around 30 % was achieved at a negligible loss in accuracy. Despite the gains in computational performance complete description of turbulent reacting flows in the limit of the current computational resources is unfeasible, this leads naturally to the other facet of the study, which is to include generality in manifold based methods.
Regarding manifold based methods, two strategies were analyzed, namely, the multi-regime flamelet and the multi- mode combustion model. The models were tested in 1-D and 2-D benchmark problems, where the multi-mode combustion method excels at highly stratified flows. Though, for low and moderately stratified flows both the methods show better prediction than conventional manifold methods. The multi-model combustion model is applied to the multi-regime burner (MRB), where it reproduces global and conditional flame statistics.
Overall, the objectives of realizing multi-regime combustion at various levels of complexity was achieved using the different strategies studied. Lastly, concluding remarks are given for the usability of the methods in context of current HPC scenario.
Share: