Large-Eddy Simulation of the lean-premixed PRECCINSTA burner with wall heat loss
Introduction
Swirl-type combustion is widely used for industrial burners as it provides several advantages. When the geometrical swirl number is high enough, i.e. over 0.6, the burner exhibits a Central Recirculation Zone (CRZ) and Outer Recirculation Zones (ORZ). In these zones, hot burnt gases recirculate helping flame stabilization and extending flammability limits [1]. In premixed swirl burner, the flame structure is influenced by wall heat transfer [1]. Indeed, it has been observed that strong heat loss may influence the ORZ, weakening the stabilization process [2] up to local extinction [3].
Numerous numerical studies have been performed on the so-called PRECCINSTA burner [4], a swirl lean-premixed methane-air combustor. Most of them focus on Large-Eddy Simulations (LES) with adiabatic wall conditions. These computations were not able to correctly capture the flame structure. Moureau et al. [5] and Franzelli et al. [6], [7] agree to attribute this effect to the adiabatic wall condition or to partial premixing without being able to validate it. However, these studies demonstrated the strong influence of the mesh resolution [5] and of the chemistry description [7] on the flow and the flame topology. Indeed, Franzelli et al. [7] explained the need of analytical or skeletal schemes to accurately predict the consumption speed, the intermediate species concentration and the flame structure.
In this paper, LES of the PRECCINSTA burner are carried out with a skeletal scheme using four different grid resolutions and with adiabatic and non-adiabatic wall conditions. The objective is to study grid resolution and heat loss influence on the flame structure and pollutant formation. First, the experimental set-up and the numerical modeling are presented, as well as the strategy established to obtain the non-adiabatic condition is exposed. Then, the LES results are analyzed and compared to experimental data.
Section snippets
Experimental set-up
Designed within the EU project PRECCINSTA, the combustor is derived from an industrial design by SAFRAN Helicopter Engines [4], [8], representative of a real aeronautical gas turbine combustor. It was widely studied experimentally [9] or numerically [5], [6], [10], [11], [12], [13], [14].
As described by Fig. 1, the geometry consists of three parts. The premixed methane-air mixture is injected into the plenum and swirled through the injector by twelve radial veins before entering the chamber
Numerical modeling
Large-Eddy Simulations are performed using the in–house solver YALES2 [15]. This massively parallel finite-volume code solves the low-Mach number Navier–Stokes equations using a projection method [16] for variable density flows [17]. Density, momentum and scalar equations are solved using a 4th-order centered scheme in space and a 4th-order Runge–Kutta-like scheme in time [18]. It is able to handle unstructured meshes up to billions of elements.
Finite-rate chemistry is employed: all species
Results and discussion
All the performed LES computations are listed in Table 3, combining the different meshes and wall thermal conditions. As shown in Table 2, the time-averaged maximum value of the thickening factor decreases at each grid refinement. However, even with the most refined mesh composed of 877 million elements, the SGS flame-turbulence interaction model is still necessary but is expected to have a very small impact on the results. A maximum thickening of around 3 means that the equivalent laminar
Conclusions
In this work, a numerical strategy to take into account wall heat loss on the lean premixed methane-air PRECCINSTA burner has been proposed. This methodology relies on high-fidelity LES, a skeletal kinetic scheme, the DTFLES model and a wall temperature Dirichlet boundary condition. The results were compared to adiabatic computations and to experimental data. Different grid resolutions were tested, with meshes containing from 1.7 up to 877 millions of elements. The results show that the flame
Acknowledgments
This work was granted access to the HPC resources of CINES and CEA under the allocations A0032B06880 made by GENCI and 2016163999 made by PRACE and of CRIANN under the allocation 2012006.
References (28)
- et al.
Combust. Flame
(1974) - et al.
Appl. Therm. Eng.
(2004) - et al.
Combust. Flame
(2012) - et al.
C. R. Mec.
(2013) - et al.
Combust. Flame
(2007) - et al.
Combust. Flame
(2005) - et al.
J. Comput. Phys.
(2007) - et al.
Combust. Flame
(2010) - et al.
Combust. Flame
(2016) - et al.
Combust. Flame
(2017)