The Large Eddy Simulation / Filtered Mass Density Function Approach Applied to High Pressure Turbulent Non-Premixed Flame

Large Eddy Simulation (LES) coupled with the transported Filtered Mass Density Function (FMDF) approach has shown great potential in simulating turbulent flames. In the beginning of this work, we conduct a priori analyses of the conditional diffusion in LES/FMDF that focus on the velocity scalar mixing time scale ratio (CO,alpha) which is a widely used parameter in LES combustion modelling. The analysis is based on an existing database of direct numerical simulations (DNS) of high pressure temporally developing shear flames. The DNS includes generalized heat and mass diffusion, a real fluid equation of state (EOS), pressure and temperature dependent properties and detailed H 2/O2 and H2/Air kinetics. Results show CO,alpha for the high pressure hydrogen flames in this work is approximately 3. However, the optimal value is found to vary both for different species and for different locations within the flame. This is particularly evident for the light species (e.g. H and OH), for which CO,alpha is observed to be typically twice that of major species such as oxygen.;Then we extend LES/FMDF to simulate the DNS temporally developing reacting shear flames at high pressure (up to 100atm), and Reynolds numbers (up to 4500). Our LES/FMDF solver includes all of the high pressure physics of the DNS. The standard LES equations are solved on an Eulerian mesh while the FMDF equation is solved with the notional particle technique and is used to close the filtered chemical source terms for the LES. Reasonable agreement is found for the majority of first and second order flow statistics for both the velocity and scalar fields. We also conduct a comparative study of several conditional diffusion models used in current FMDF simulations, including the Interaction and Exchange with the Mean (IEM), Euclidean Mean Spanning Tree (EMST), and Modified Curl's model. The most recent and relatively untested Shadow Positioning model is also assessed at high pressure. All four models predict temperature and scalar fields fairly well; however, the EMST model provides the best overall results; particularly so for the H2/Air flames with their associated large levels of local extinction. The EMST model is also shown to be the least sensitive to the choice of the modeling constant relating the velocity scalar mixing timescale.;To further incorporate generalized diffusion in FMDF, an effective diffusion coefficient model is derived. The model is found to be able to accurately reproduce the exact generalized diffusion vector. With this effective diffusion coefficient, the counter-gradient diffusion (caused by different directions among the pressure gradient, temperature gradient, and other species gradients) is able to be captured. In addition, the consistency between FMDF and LES Eulerian transport equations is better obtained. A posteriori analyses show this model has significant improvement in minor species prediction in high pressure reacting flows.;In addition, the subgrid pressure is studied in a priori analyses. Results show the subgrid pressure and its gradient are significant in reacting flows at high pressure; especially in specific local regions (e.g. regions of large subgrid temperature fluctuation). Therefore, the subgrid pressure may need to be modelled. To close the subgrid pressure, a dynamic similarity subgrid pressure model is proposed. The model is tested by both a priori and a posteriori analyses. Results show the model can improve the estimation of the flow fields in LES.