Ph .D.(Engg.): Explicit Filtering LES of turbulent swirling flows and multi-regime hydrogen flames

The need to decarbonize power generation and aviation has established hydrogen as a promising fuel for next generation gas turbine combustion systems. However, its high chemical reactivity significantly increases the propensity for autoignition and flashback under lean premixed conditions commonly used in hydrocarbon combustion systems. New combustor concepts based on direct fuel injection address these problems but typically produce lifted flames. These types of flames exhibit multi-regime combustion, in which premixed and non-premixed combustion regions co-exist within highly turbulent swirling flow fields. Accurate numerical prediction of such reacting flows remains a major challenge as traditional models developed for hydrocarbon flames often fail to reliably predict the stabilization and combustion characteristics of multi-regime hydrogen flames. This thesis proposes a regime-sensitive combustion model using the thickened flame approach to extend the explicit filtering large eddy simulation (EFLES) method for computations of high Reynolds number (Re) reacting flows. A combustion regime index (CI) is defined based on the alignment between the gradients of mixture fraction and oxygen mass fraction to distinguish between premixed and non-premixed regions. The flame thickening treatment is applied selectively in premixed regions only. The model is implemented within an in-house high-order finite-difference flow solver that employs structured multi-block overlapping meshes to represent complicated geometries encountered in practical combustion systems. The predictive capability of the overall numerical methodology is systematically assessed through a series of increasingly complex flow configurations relevant to swirl-stabilized hydrogen combustors. The first case corresponds to an isothermal (Re ~ 27000) flow in the PRECCINSTA single nozzle swirl combustor. EFLES solutions show excellent agreement with the experimental data for flow field statistics and shape of the vortex breakdown bubble. Spectral proper orthogonal decomposition applied on time resolved velocity fields confirms that the EFLES method accurately reproduces the broadband turbulence structures and the coherent, narrow band, self-excited precessing vortex core oscillation. The combustion model performance is then evaluated using a canonical lifted slot jet flame configuration, comprising a nitrogen diluted hydrogen slot jet of Re=8000 at 400 K issuing into a co-flow of heated air at 850 K. Entrainment of air into the jet core results in a stratified rich premixed flame alongside the primary non-premixed flame branch. The LES predicted flame lift-off height and flow field statistics are in good agreement with companion direct numerical simulation (DNS) results. Differences between LES and DNS solutions are small in premixed flame regions but increase with mesh spacing in non-premixed flame regions. The CI field is shown to clearly distinguish the two flame branches. This result is further validated a-posteriori using chemically explosive mode analysis (CEMA). Combined with EFLES, the combustion model accurately predicts mixture fraction conditioned thermochemical statistics, combustion mode distributions and the downstream evolution of premixed and diffusion flame branches in the lifted slot jet hydrogen flame. Finally, a turbulent lifted hydrogen flame in the dual-swirl HYLON burner is investigated as a practical test case. The time-averaged mean flame shape and liftoff height is well captured in the LES solution. Good agreement is obtained between LES predictions and experimental measurements for statistics of velocity, temperature, and major-species mole fractions. CEMA reveals that a partially premixed flame branch is formed along the inner shear layer due to entrainment of hydrogen into the high-velocity annular air stream. The central diffusion flame branch is located within the vortex breakdown bubble between the fuel jet and the recirculating lean combustion products. The distribution of key thermochemical variables in mixture fraction space is also analyzed to characterize the two flame branches. Overall, the results demonstrate that the proposed multi-regime thickened flame model, coupled with explicit filtering LES provides a robust framework for high-fidelity LES of turbulent hydrogen flames. The methodology reproduces turbulent flow features, flame stabilization and mixed combustion regimes across canonical and practical configurations. These results make the case for evaluating the performance of EFLES at higher operating pressures towards realizing a reliable computational engineering method for the design and analysis of future hydrogen gas turbine combustors.

Speaker :  ANINDYA DATTA

Research supervisor : Santosh Hemchandra