A New Foundation For Performing LES of Turbulent Reacting Flows

Large Eddy Simulation (LES) has become the Computational Fluid Dynamics (CFD) approach of choice for simulation of reacting flows relevant to aerospace. This is despite significant limitations of the approach such as lack of reproducibility of results and limited predictive ability. In this paper we identify the root causes of these key reacting LES limitations and we begin the process of identifying methodologies for their mitigation. Extending our work of the past few years we clarify that the lack of reproducibility of results is primarily related to the inherent grid-dependency of LES as practiced today which links the LES filter to the computational grid (Grid-Filter approach). This results in an inability to execute solution convergence tests and to quantify the impact of numerical error. These issues are explained theoretically and verified by simulations relevant to planar bluff body stabilized premixed flames. It is proposed that the mitigation of this problem can be achieved by disconnecting the LES filter from the grid size (Specified Filter approach). The effectiveness of this approach is demonstrated through non-reacting and reacting simulations. The predictive limitations of LES are shown to be also related to the complexity of the physical phenomena to be modeled via subgrid modeling. It is argued that a key problem here is the heavy reliance of currently available subgrid models on Kolmogorov turbulence theory. Using physical and mathematical arguments it is explained that Kolmogorov theory is not relevant to exothermic reacting flows. Success of subgrid models based on this theory is partially accidental, relies on model features that are not related to Kolmogorov theory and also relies on the ability of the practitioner to tune the solution via iteration. It is argued that new ideas are needed in this area and it is proposed that exploitation of the fractal nature of high Reynolds number reacting flow could be one of them.