(273e) Numerical Modelling of Industrial Burners for Reduction of NOx Emissions Using Flamelet Methods in Combination with a Newly Developed Postprocessor for Fast and Accurate Emission Prediction

Current trends of research on combustion systems show increased interests on prediction of pollution. To meet the demands of modern combustion systems such as industrial burners detailed studies of their behaviour is needed. One of these pollutants which is commonly of interest is NO_x. The aim of the work is to develop a numerical toolbox that is able to predict NO_x emissions of industrial burners by numerical simulation using Computational Fluid Dynamics (CFD).

Numerical simulation is applied to many multiphysics problems such as combustion simulation. A set of numerical equations is solved for flow, temperature, pressure and species. Using CFD it is possible to model a combustion system using different approaches like flamelet methods or detailed chemical reaction systems. To predict the emission of NO_x from a flame a detailed chemical reaction system is necessary such as the GRI 3.0. mechanism [1]. The problem of using this mechanism is that it can be applied only for small geometries to avoid long calculation time or processing costs. Using a detailed chemical reaction system for every species a few chemical reactions have to be solved. Using the GRI 3.0. mechanism 325 reactions for 53 species have to be solved [2]. Also with modern computers it is not affordable to model industrial burners using this approach which would calculate NO_x emissions well.

The second main possibility for simulation of burners is to use flamelet methods [3][4]. Choosing this approach the combustion system is pre-calculated and reaction data will be stored in so called lookup tables. During simulation this reaction data will be accessed by reading them from the lookup tables. This approach allows fast calculation of temperature and flow field of gas burners as well as fast chemical reactions like formation of CO₂, H₂O or CO. This is the standard procedure for simulation of industrial burners for analysis of flame shape and temperature fields. The problem is that the described approach is not applicable for slow chemical reactions such as formation of NO_x. Therefor a detailed chemical reaction system is necessary. The problem is that on the one hand there is the detailed chemical reaction system which would predict NO_x emissions very well but takes too long for calculation. On the other hand there is the flamelet method which is fast but delivers only for flow fields, temperature fields as well as some species fields right results but models the formation of NO_x not very well. The idea now is to combine the benefits of both approaches and to postprocess the results out of the flamelet calculation using a detailed chemical reaction system. This means temperature and flow field are calculated by the fast flamelet calculation. After convergence of this calculation the solution for the flow field and temperature field as well as for some species fields are known. Subsequently for slow chemical reactions such as NO_x production, which cannot be predicted by the flamelet method, the postprocessor calculates the reactions using the already existing temperature and flow field from the flamelet calculation. This combination enables a much faster calculation of NO_x formation in the flame and combines the speed of flamelet method with the accuracy of using detailed chemical reaction systems with reasonable calculation costs [5].

For a more precise description of radiation effects it was necessary to implement the Discrete Ordinate Model (DOM) into the flamelet solver of OpenFOAM [6]. The standard radiation model is only valid for free stream flames with a constant surrounding wall temperature. The radiation model was used in combination with a Weight Sum of Gray Gases Model (WSGGM) which calculates the optical density of the gas by applying a mixture of temperature dependent absorption coefficients for calculation of the radiation beams [7].

For flamelet calculation a steady laminar flamelet for non-adiabatic combustion was used out of literature [3][4]. By defining a range for enthalpy defects in flamelet generation heat transfer in the system and out of the system can be taken into account. Flamelets were generated for different scalar dissipation rates which takes mixing time scales into account and for different enthalpy defects.

For modelling of turbulent effects the k-ε-model was used with a slight change according to literature data. For a more precise prediction of flame acceleration the constant C_2ε was modified to 1.8. [8]. Out of data from Sandia Flame D boundary conditions for turbulence were generated for turbulent kinetic energy k and dissipation of it named ε [1].

The NO_x postprocessor uses a detailed chemical reaction approach the so called partially stirred reactor model implemented in OpenFOAM for calculation of chemical reaction [9]. It takes chemical reaction kinetics, for instance out of GRI 3.0., as well as turbulent mixing effects into account [2]. The solver uses a conservation equation for each species and chemical reaction is implemented as a source term in every species equation. This means for chemical reaction in every mesh cell the whole reaction mechanism consisting of hundreds of Arrhenius equations has to be solved.

The model validation was done by means of an industrial test where a burner was operated in a test chamber. Measurements of wall temperatures by thermocouples and a detailed exhaust gas analysis were performed. Out of the measurements of 20 thermocouples at different positions on the furnace wall a temperature profile was generated to implement it as boundary condition for the flamelet calculation. This lab test was modelled using the developed simulation toolbox. A mesh of the whole test geometry was generated using cfMesh [10]. Out of mass flow and temperature measurements boundary conditions for simulation were calculated. The first validation for the model was a comparison between exhaust gas temperature in the lab test and simulation. These data were then used to start the NO_x post processing. Therefor the developed transient postprocessor was used to calculate the formation of NO_x and various other species which are not predicted well by the flamelet method. At that point it is only necessary to calculate the chemical reaction source terms without solving flow and temperature fields which are now constant in this calculation and already known from the flamelet calculation. This way the simulation is much faster and calculation of NO_x formation now is possible in affordable time using the detailed chemical reaction mechanism with the benefit of a much more accurate description of NO_x formation. For the whole geometry a mesh independency study was performed on the one hand for the flamelet model and on the other hand for the postprocessing. The mesh was generated with a cell size which delivers mesh independent results. For the test chamber with a height of 2.3, a width of 1.7 and a depth 2.5 meters the mesh consists of 5.26 million polyhedral cells.

A second validation was made on the Sandia Flame D [1]. Therefor a mesh out of literature was used which was already analysed regarding mesh independency [5]. The benefit of validation on Sandia Flame D is that temperature, velocity, turbulence and species decomposition is also known in the flame. This means it is feasible to analyse the whole flame shape and its position dependent characteristics and not only the properties at the outlet. For reliable results it is important to be able to predict the temperature maximum very well. Comparison of species decomposition shows very good agreement in terms of oxidizer consumption and CO₂ formation [1]. Using the newly developed postprocessor a very accurate description of NO_x formation is possible which fits well to literature data [1].

References

[1] Barlow R, Frank J. Piloted CH4/Air Flames C, D, E, and F – Release 2.1. Available at: http://www.sandia.gov/TNF/DataArch/FlameD/SandiaPilotDoc21.pdf; 2007 [accessed 07.04.2017].

[2] Smith GP, Golden DM, Frenklach M, Moriarty N, Eiteneer B, Goldenberg M, Bowman T, Hanson RK, Song S, Gardiner WC, Lissianski, Jr., Vitali V., Qin Z. GRI Mech 3.0. Available at: http://www.me.berkeley.edu/gri_mech/; 1999 [accessed 11.04.2017].

[3] Cuoci A, Frassoldati A, Faravelli T, Ranzi E. Formation of soot and nitrogen oxides in unsteady counterflow diffusion flames. Combustion and Flame 2009;156:2010–22, doi:10.1016/j.combustflame.2009.06.023.

[4] Cuoci A, Frassoldati A, Faravelli T, Ranzi E. Kinetic Modeling of Soot Formation in Turbulent Nonpremixed Flames. Environmental Engineering Science 2008;25:1407–22, doi:10.1089/ees.2007.0193.

[5] Spijker C. Unsteady Laminar Flamelet Modellierung zur Beschreibung von Mündungsmischbrennern. Masterthesis, Montanuniversitaet. Leoben; 2010.

[6] Thynell ST. Discrete-ordinates method in radiative heat transfer. International Journal of Engineering Science 1998;36:1651–75, doi:10.1016/S0020-7225(98)00052-4.

[7] Barlow RS, Karpetis AN, Frank JH, Chen J-Y. Scalar profiles and NO formation in laminar opposed-flow partially premixed methane/air flames. Combustion and Flame 2001;127:2102–18, doi:10.1016/S0010-2180(01)00313-3.

[8] Pfeiler C, Spijker CJ, Raupenstrauch H. CFD als Werkzeug in der Industrieofentechnik. Berg Huettenmaenn Monatsh 2011;156:347–52, doi:10.1007/s00501-011-0018-z.

[9] OpenFOAM. OpenFOAM 2.4.0. Available at: https://openfoam.org/; 2017.

[10] cfMesh. Professional Meshing Solutions - cfmesh.com. Available at: http://cfmesh.com/; 2017 [accessed 11.04.2017].