Validation of a Fractal Dynamic SGS Combustion Model by DNS of Turbulent Premixed Flame in Strong Shear Flow

Abstract

A fractal dynamic subgrid scale (SGS) combustion model for large eddy simulation (LES) of turbulent premixed combustion which was developed in our previous study (Yoshikawa et al., 2013) is evaluated through static tests on filtered data of direct numerical simulation (DNS) of H2—air turbulent V-shape premixed flame. The model is based on flamelet concept, fractal characteristics of turbulent premixed flames and scale separation in high Reynolds number turbulence and the accuracy has been demonstrated for freely propagating premixed flame in homogeneous isotropic turbulence (HIT). The results of the static tests validate that the present model has high accuracy and is applicable to conditions where strong mean shear exists and the interaction between flame and turbulence is complex. Furthermore, comparison with some conventional SGS combustion models indicates that the present model has superiority to these models in terms of accuracy. INTRODUCTION Large eddy simulation gains more and more attention as a powerful tool for simulating turbulent combustion. Recently, many attempts applying LES to realistic combustion systems have been made (Colin et al., 2000; Pitsch and Duchamp de Lageneste, 2002; Stone and Menon, 2002; Grinstein and Fureby, 2005; Fureby, 2005; Wang and Bai, 2005; Huang and Yang, 2005; Fiorina et al., 2010; Kuenne et al., 2011; Wang et al., 2013). In LES, physical quantities are filtered and divided into the grid scale (GS) and subgrid scale components. Then, large scale unsteady phenomena represented by the GS components are computed, while the contribution of SGS phenomena to those of GS is given by SGS models. Since combustion is intrinsically unsteady, LES is expected to be more accurate than the classical Reynolds Averaged Navier-Stokes (RANS) simulation for numerical simulations of turbulent combustion. In LES of turbulent combustion, tracking of the flame front is one of the difficulties, since the flame thickness is generally quite thin compared to the LES filter width (), where the filter width should be in the inertial sub-range of turbulent spectrum. For tracking the flame front, Gequation (Kerstein et al., 1988) is frequently used. In the approach, flame front is represented as an infinitely thin scalar iso-surface and its propagation is computed. For extending this approach to LES, G-equation is filtered. Then, a closure model (SGS combustion model) for turbulent burning velocity, ST, is introduced to solve the filtered G-equation. Since accuracy of LES strongly depends on the SGS model, highly accurate SGS models are required to acquire trustworthy results. However, almost all SGS combustion models used in this approach are those extended from RANS models (Pitsch and Duchamp de Lageneste, 2002; Stone and Menon, 2002; Wang and Bai, 2005; Huang and Yang, 2005) and their accuracy in LES is not assured. In our previous study (Yoshikawa et al., 2013), a fractal dynamic SGS combustion model based on flamelet concept, fractal characteristics of the flame surface and scale separation in high Reynolds number turbulence for LES of turbulent premixed combustion has been proposed. A series of static tests with DNS data of H2—air freely propagating flame in HIT has been conducted to evaluate the model, and has demonstrated the accuracy. In practical combustion chambers, however, strong mean shear exists and turbulent combustion characteristics become complicated compared with freely propagating flame in HIT. In this study, the fractal dynamic SGS combustion model is examined on filtered DNS data of H2 — air turbulent premixed flame with strong mean shear, where V-shape flame configuration (Minamoto et al., 2011) is considered, and the accuracy of the model is evaluated. For comparison, several conventional SGS combustion models are also investigated. August 28 30, 2013 Poitiers, France COMB2E 2 FRACTAL DYNAMIC SGS COMBUSTION MODEL Detailed derivations of the fractal dynamic SGS combustion model are presented in our previous study (Yoshikawa et al., 2013). The model consists of two parts that represent turbulence and dilatation effects. The former is modelled based on fractal characteristics of the flame surface of turbulent premixed flames and scale separation in high Reynolds number turbulence, while the latter is based on the flamelet concept. Then, the turbulent burning velocity, ST, is given as the sum of the two parts.