Subgrid scale modeling for large eddy simulation of supercritical mixing and combustion

Abstract

Large eddy simulation (LES) is a widely used modeling and simulation technique in turbulent flow research. While the LES methodology and accompanying subgrid scale (SGS) modeling have been developed and applied over decades, primarily in the context of ideal gas conditions, their extension to complex multi-physics flows encountered in aerospace propulsion requires further refinement. In particular, the application of LES to turbulent flows at supercritical conditions presents several new modeling challenges and uncertainties. The scope of this dissertation is to investigate the theoretical LES formalism and SGS modeling framework for multi-species turbulent mixing and combustion at supercritical pressures. The goal is to identify the deficiencies with the current methodology and to establish a refined and consistent framework that accurately accounts for all the necessary physics.

In this dissertation, a consistent theoretical formulation of the filtered governing equations for LES is derived. Direct numerical simulations (DNS) are performed for spatially evolving non-reacting and reacting mixing layers at supercritical pressures. The complete set of terms in the filtered equations are quantified and analyzed using the DNS datasets. Based on the analyses, two new groups of subgrid terms are identified as important quantities to account in the LES framework. Parametric analyses are performed as a function of the filter resolution to derive resolution considerations for practical LES applications.

The performance and accuracies of two state-of-the-art subgrid modeling approaches for the traditional subgrid fluxes are assessed. The study demonstrates the better performance of scale-similarity based models over the eddy-viscosity based approaches. The study also reveals the deficiencies of conventional subgrid modeling approaches for LES of supercritical combustion. To address the additional modeling requirement for the filtered equation of state, novel subgrid modeling approaches are proposed. The performance of these models are tested and good improvements are demonstrated.