Elsevier

Combustion and Flame

Volume 156, Issue 1, January 2009, Pages 25-36
Combustion and Flame

Large eddy simulation and laser diagnostic studies on a low swirl stratified premixed flame

https://doi.org/10.1016/j.combustflame.2008.06.014Get rights and content

Abstract

This paper presents numerical simulations and laser diagnostic experiments of a swirling lean premixed methane/air flame with an aim to compare different Large Eddy Simulations (LES) models for reactive flows. An atmospheric-pressure laboratory swirl burner has been developed wherein lean premixed methane/air is injected in an unconfined low-speed flow of air. The flame is stabilized above the burner rim in a moderate swirl flow, triggering weak vortex breakdown in the downstream direction. Both stereoscopic (3-component) PIV and 2-component PIV are used to investigate the flow. Filtered Rayleigh scattering is used to examine the temperature field in the leading flame front. Acetone-Planar Laser Induced Fluorescence (PLIF) is applied to examine the fuel distribution. The experimental data are used to assess two different LES models; one based on level-set G-equation and flamelet chemistry, and the other based on finite rate chemistry with reduced kinetics. The two LES models treat the chemistry differently, which results in different predictions of the flame dynamic behavior and statistics. Yet, great similarity of flame structures was predicted by both models. The LES and experimental data reveal several intrinsic features of the low swirl flame such as the W-shape at the leading front, the highly wrinkled fronts in the shear layers, and the existence of extinction holes in the trailing edge of the flame. The effect of combustion models, the numerical solvers and boundary conditions on the flame and flow predictions was systematically examined.

Introduction

Swirling lean premixed flames are frequently used in modern gas turbine combustors since they offer a possibility of controlled flame temperature and thus favorable thermal NOx emissions. However, these flames pose a challenge to engineers as they are inherently unstable; they exhibit not only turbulent motions but also low frequency large-scale coherent structure dynamics [1], [2], [3]. This unstable motion could trigger not only noise, but also combustion oscillations and structural damage. Many studies have been carried out to obtain deeper understanding and control of swirl combustion systems, e.g. experimental measurements of swirling lean premixed systems [4], [5], [6], and numerical simulations of premixed and stratified premixed swirling flames [7], [8], [9], [10], [11], [12], [13]. Gouldin et al. [4] found that under high swirl conditions, two modes of low-frequency non-turbulent large-scale oscillations can exist, one is attributed to the structure of the swirl generator inside the burner and one is caused by the Precessing Vortex Core (PVC), owing to the breakdown of vortices imposed by the swirling inflow. For high swirl, Benjamin [14] showed that the flow transits from a supercritical to a subcritical mode, where downstream disturbances can travel upstream and modify the combustion process. Escudier et al. [15], [16] experimentally demonstrated the subcritical flow nature: by varying the combustor outlet contraction, the flow structures changed from an on-axis bubble-like breakdown to an off-axis annular breakdown. Owing to the complexity of the flow and combustion physics, accurate predictions of swirling flames are challenging. Reynolds Average Navier Stokes (RANS) models with the kε type model were shown to be inadequate to capture the low-frequency rotational flow [17]. In this regard, Large Eddy Simulation (LES) and similar methods [18] offer a greater possibility due to its built-in ability to resolve the coherent structure dynamics.

LES has been successfully used to predict swirling isothermal flows and premixed flames [7], [8], [9], [10], [11], [12], [13], [19], [20], [21], [22]. It is important to note that rather diverse modeling approaches have been applied in these reported LES studies, for example the flamelet models based on level-set G-equation (e.g. [8], [13]), the artificially thickened flame model based on a reaction progress variable (e.g. [11]), the flame surface density models, and models based on direct coupling of finite rate chemistry (e.g. [7]). This reflects the fact that the physical process is complex and there is a lack of generally acceptable and universally valid model. It is therefore desirable to systematically investigate the performance of the different models. This can be done by applying different models to simulation of the same flame cases.

A common difficulty in LES of turbulent combustion is the modeling of the reaction zones. The chemical reaction zones are usually below the Taylor micro-scale and are generally not resolved in LES. The resolved flow acts as a boundary condition to the reaction zones, which has led to the development of different subgrid combustion models. One such model is the G-equation flamelet model, which combines a flamelet approximation, with the reaction layers tabulated and stored in a flamelet library [23], [24], and with a level-set G-equation [23], [24], [25], [26], [27] predicting the motion of the flame front. The main advantage of this method is the possibility of coupling detailed chemical kinetics to LES, and the drawbacks are the static coupling between the chemical kinetics and the flame front, and issues regarding the capability of predicting extinction and re-ignition. Another method is to employ the chemical kinetics to directly simulate the propagation of the reaction fronts at the resolved scales. This has the potential to capture the flame dynamics more precisely (including extinction and re-ignition) but at a higher computation cost—in particular if detailed kinetics is necessary. An issue with this type of model is how to model the integrated kinetics rates across the reaction zone, taking into account the different boundary conditions imposed on the reaction zones by the resolved flow. These two models are examined in the present paper, with the objective of assessing their performance in predicting swirling flames.

A limitation for the development and validation of combustion models for swirling flames is the lack of well-documented reference data, especially for the inflow and outflow conditions. As pointed out earlier [19], [28], boundary conditions can be decisive to the accuracy of the simulations. In view of this, a new laboratory measurement rig was developed [5], [8] in which a premixed CH4/air mixture is injected through a low swirl burner into to a co-flowing unconfined atmosphere. The purpose of having an unconfined flame is to separate the influence of the confinement induced outer toroidal vortices off the burner axis and the outlet geometry on the flame from the inflow swirl effect. The basic burner design is similar to that of Cheng et al. [29], [30], [31] with an annular swirler and inner perforated plate to generate nearly homogeneous turbulence in the central region and a weak vortex breakdown downstream. To quantify the inflow, 3-component PIV velocity measurements are carried out in the proximity of the burner nozzle orifice. The effect of the swirler geometry is evident in the PIV data and its effect on the flame is examined using LES.

Section snippets

Burner and experimental setup

The low swirl burner used in this investigation is an enhancement of the original design proposed by Bedat and Cheng [30] for studying premixed flames propagating in turbulence. The goal with the new design is to produce a suitable generic burner for stabilizing lean premixed flames and to create a validation database. The burner used here, Fig. 1, is one of several that were developed and manufactured by the Technical University of Darmstadt [5]. Premixed fuel and air are fed to a mixing

LES combustion modeling

The equations governing reactive flows are the conservation equations of mass, momentum and energy describing convection, diffusion and chemical reactions, which after filtering become,t(ρ¯)+(ρ¯v˜)=mρ,t(ρ¯Y˜i)+(ρ¯v˜Y˜i)=(ji¯bi)+w˙i¯+mi,t(ρ¯v˜)+(ρ¯v˜v˜)=p¯+(S¯B)+ρ¯f˜+mv,t(ρ¯h˜)+(ρ¯v˜h˜)=tp¯+S˜D˜+ε+(h¯bh)+ρ¯σ˜+mh. Here, ρ is the density, v the velocity vector, p the pressure, S the viscous stress tensor, f the body force, h the enthalpy, h the heat flux vector, D=12(v+v

Results and discussions

To investigate the impact of the two LES models on the numerical simulations, we consider comparing the numerical results in the following two aspects: the prediction of the basic structures of the flame and the prediction of the statistics of the flame. Experimental data obtained previously [5], [8] and in this study (e.g. simultaneous acetone and OH PLIF) are used to validate the simulations.

Concluding remarks

Experimental and LES studies are carried out on a low swirl stratified premixed flame to evaluate two different LES models, and also to deepen the understanding of stratified premixed turbulent low swirl flames. The two LES models are respectively based on a G-equation level-set formulation with pre-tabulated chemistry, and direct coupling of two-step finite rate chemistry with the flow. Two different inflow boundary conditions are also used in order to examine the influence of the details of

Acknowledgments

This work was supported by the Swedish research funding organizations (STEM, SSF, CeCOST, and VR), European Union Large Scale Facility in Combustion (contract no. HPRI-CT-2001-00166) and the Swedish Armed Forces.

References (64)

  • O. Lucca-Negro et al.

    Prog. Energy Combust. Sci.

    (2001)
  • N. Syred

    Prog. Energy Combust. Sci.

    (2006)
  • K.J. Nogenmyr et al.

    Proc. Combust. Inst.

    (2007)
  • W.P. Jones

    Comput. Phys. Commun.

    (2002)
  • Y. Huang et al.

    Proc. Combust. Inst.

    (2005)
  • R. Weber et al.

    Int. J. Heat Fluid Flow

    (1990)
  • J. Janicka et al.

    Proc. Combust. Inst.

    (2005)
  • C. Fureby et al.

    Proc. Combust. Inst.

    (2007)
  • P. Nilsson et al.

    Proc. Combust. Inst.

    (2002)
  • H. Pitsch et al.

    Proc. Combust. Inst.

    (2002)
  • P. Wang et al.

    Proc. Combust. Inst.

    (2005)
  • R.K. Cheng

    Combust. Flame

    (1995)
  • B. Bedat et al.

    Combust. Flame

    (1995)
  • J.H. Frank et al.

    Combust. Flame

    (1999)
  • I.G. Shepherd et al.

    Proc. Combust. Inst.

    (2002)
  • H.G. Weller et al.

    Proc. Combust. Inst.

    (1998)
  • C. Nottin et al.

    Proc. Combust. Inst.

    (2000)
  • F. di Mare et al.

    Combust. Flame

    (2004)
  • C. Fureby et al.

    J. Comput. Phys.

    (2002)
  • L. Selle et al.

    Combust. Flame

    (2004)
  • R.I. Issa

    J. Comput. Phys.

    (1986)
  • T.J. Poinsot et al.

    J. Comput. Phys.

    (1992)
  • S. Roux et al.

    Combust. Flame

    (2005)
  • A.K. Gupta et al.

    Swirl Flows, Tunbridge Wells

    (1984)
  • F.G. Gouldin et al.

    Proc. Combust. Inst.

    (1984)
  • P. Petersson et al.

    Appl. Opt.

    (2007)
  • N.D. Joshi, M.J. Epstein, S. Durlak, S. Marakovits, P.E. Sabla, ASME Paper 94-GT-253, June...
  • F.F. Grinstein et al.

    Proc. Combust. Inst.

    (2004)
  • W.W. Kim et al.

    Combust. Sci. Technol.

    (1999)
  • P. Schmitt et al.

    J. Fluid Mech.

    (2007)
  • C. Duwig et al.

    AIAA J.

    (2007)
  • T.B. Benjamin

    J. Fluid Mech.

    (1962)
  • Cited by (119)

    • Pilot impact on turbulent premixed methane/air and hydrogen-enriched methane/air flames in a laboratory-scale gas turbine model combustor

      2022, International Journal of Hydrogen Energy
      Citation Excerpt :

      The numerical predictions are denoted with continuous lines in Fig. 9. Based on the level-set G-equation flamelet model for turbulent premixed flames [15,16], the combustion products in the exhaust gas may be predicted using an ensemble average of different flamelets, taking into account the effect of the flame stretch. The present 1D adiabatic unstretched flame results can be considered as one realization of the turbulent flame.

    • Structure and dynamics of highly turbulent premixed combustion

      2021, Progress in Energy and Combustion Science
    View all citing articles on Scopus
    1

    Currently at Sandia National Lab.

    View full text