Research Topic

Development of numerical methods for the simulation of turbulent premixed combustion

Introduction

Figure 1: Combustor inside a typical gas turbine for power generation © Siemens

Despite the important advances made in the field of renewable energies, the majority of the rising energy demand is still covered by combustion processes. Considering its negative impact onto the environment due to the formation of pollutants, the optimization of the combustion process itself and the corresponding devices is highly desirable. Therefore the majority of the land based gas turbines operate in premixed mode (fuel and oxidizer are mixed before entering the combustion chamber) which strongly decreases the formation of nitrogen oxides (NOx) due to the lower peak temperature when compared to non-premixed systems (fuel and oxidizer mix within the combustor and burn instantly). Unfortunately, unsteady phenomena like flashback and acoustic instabilities are known to appear prevalently within these premixed flames making them more difficult to control.

Therefore this work focuses on the prediction of these combustion processes via computational fluid dynamics. Such methods are expected to become a powerful tool during the design process of combustors. In figure 1 a typical power generation unit is shown consisting of compressor, combustor (glowing middle section) and turbine. The current industrial standard of a time averaged description of the combustion process has been found to be insufficient and hence a numerical technique that accounts for the intrinsically unsteady behavior of the underlying physics by using a high spatial and temporal resolution will be used within this work.

Method

Figure 2: Manifold of a methane-air flame defined by two controlling variables showing the source term of carbon dioxide colored with temperature

Within this work the so called Large Eddy Simulation (LES) will be used which explicitly computes the large structures of the flow resolved on the computational cell size while modeling the subgrid part. As shown in numerous publications this approach is very accurate in complex geometries and has been successfully applied to non-reacting flows in the past. In the context of combustion these technique is at a beginning stage but is a very promising approach for our work.

Unfortunately a splitting of scales which represent the turbulent energy cascade of the velocity field is not possible for the chemistry since it occurs completely below the cell size. As mentioned above, within premixed combustion the reactants (fuel and oxidizer) are already mixed before entering the combustor. After ignition there exists a sharp flame front which can be viewed as an interface separating the burnt from the fresh gases. This interface propagates into the fresh gas with a characteristic, also known as the laminar flame speed. Since this flame speed is the most decisive parameter within premixed combustion it needs to be reproduced by the simulation. However, the exact rate of fuel consumption is in general determined by complex chemical mechanisms which cannot be described in a three-dimensional simulation due to the high computational effort. Therefore, basically two steps are conducted within this work to enable the computation of combustion processes in complex geometries.

1. Chemistry reduction using one dimensional flamelet theories:
Chemical reactions consisting of complex intermolecular interactions which can only be accurately described by a large set of differential equations with high resolution requirements. These equations can be solved numerically by a detailed chemistry code in a one dimensional situation. Within the framework of Flamelet Generated Manifolds these results can be mapped onto a strongly reduced set of controlling variables to represent the flame structure in compositional space as shown in Figure 2.

2. Combustion modeling to use the reduced chemistry for three dimensional LES:
To use the reduced chemistry in simulations of complex geometries artificial thickening is applied to treat the relatively sharp flame on LES meshes. The procedure shown on the left of Figure 3 enables to spatially resolve the flame structure which is necessary to obtain the correct flame propagation speed. Doing so, an excellent agreement with the detailed chemistry solution exists as shown on the right of Figure 3.

Figure3: Left: Illustration of the thickening procedure. Right: Comparison of the flame propagation speed obtained for different equivalence ratios with the detailed chemistry simulation and the LES-Code using the FGM table.

Application

Besides the one- and two-dimensional test cases which served for numerical verification only, the method is currently applied to a 3-Dimensional reacting turbulent flow to test its applicability for burners of technical relevance. The unconfined natural gas swirl burner given in Figure 4 is well suited since a comprehensive set of measurements exists on velocity field, species and temperature distribution for the validation of the method.

To give an illustration of the flame stabilization and instantaneous burning behavior a snapshot is shown on the left. As one can see from the iso-surface of the chemical source term the flame has stabilized above the bluff body where the methane-air mixture issuing from the annulus is consumed. A slice colored with the temperature has been added to illustrate the burnt and unburnt state. As expected in a turbulent flowfield the isosurface is strongly distorted by vortices. Even though the flame turbulence interaction acting on the smaller scales is suppressed by the thickening procedure the increased flame surface due to wrinkling is significant. On the right the time averaged temperature extracted at the marked position is compared with experimental data. In general a good agreement with the measurements can be observed. The overestimated temperature can be attributed to heat losses which are currently neglected in the model but will be included in future work.

Figure 4: Left: Isosurface of the chemical source term an a slice showing the temperature ( K ). Right: Comparison of the computed temperature (lines) with measurements (dots) extracted along the red arrow.

Key Research Area

Multi-Physics; Large Eddy Simulation, Premixed Combustion Modeling, Stratified Combustion

Contact

Guido Künne
Dipl.-Ing.

Address:

Petersenstraße 30

D- 64287 Darmstadt

Germany

Phone:

+49 6151 16 - 24401 or 24402

Fax:

+49(0)6151 16-6555

Office:

L1|01 286

Email:

kuenne (at) ekt.tu...

 Print |  Impressum |  Sitemap |  Search |  Contact |  Privacy Policy
zum Seitenanfangzum Seitenanfang