The Modeling Of Homogeneous Reaction And Application Of Methane Premixed Combustion Based On Adaptive Chemistry

Numerical simulation of reactive flows in industrial scale combustors constitutes, presently, a research topic and has started a few decades ago. While in the past, reactive flows research focused mainly on a description of aerodynamics phenomena making recourse to extremely simple chemical mechanisms, the recent increase in computer capacities opened the possibility towards a more detailed modelling of chemical reactions in a combustor, including hydrocarbons and nitrogen chemistry. In fact, in some cases, the global chemistry models cannot even predict accurately the main features of the flames. Moreover, it is very important in many cases to model the chemistry in considerable detail, in order to accurately predict minor by-products such as pollutants and toxic emissions. However, the numerical solution of the conservation equations including detailed kinetics is very computational demanding or even prohibitive.Adaptive Chemistry appears to be a most promising approach. As described there, through this approach different reduced models are used in different regions of the flow, according to the actual chemistry reactions occurring dominantly in each region. This approach originates from the observation that inside most combustors, the dominant chemistry is different in different regions of the flame. In most existing simulations, however, only one chemistry model is used to calculate the chemistry everywhere, resulting in large amounts of CPU time being spent in computing chemistry that is negligible for large regions of the flow-field. The idea behind AdapChem is to calculate only the dominant chemistry in any particular region of the flow.The following investigations are made in this dissertation:Homogeneous mercury and tin speciation in combustion-generated flue gases were modeled by detailed kinetic model. These speciation models includes the oxidation and chlorination of key flue-gas components, as well as new reactions with reaction rate constants calculated directly from transition state theory(TST), neither from experimental data nor estimated, which was commonly used by other investigators before. The results from the model kinetic calculation are in reasonable agreement with experimental data. A new optimization-based approach to kinetic model reduction is introduced. The reaction-elimination problem is formulated as a linear integer program is the smallest possible reduced model consistent with the user-set tolerances. Sensitivity analysis is used to reveal the main pathway of reaction process. The first set of results is obtained by reducing H2/O2 combustion mechanism comprising 20 reactions. The reduced models have fewer reactions, and vary in size of 6, 7, 14, 16 under the four points t1-t4. The second set of results presented here is obtained by reducing GRI-Mech 3.0, a methane combustion mechanism comprising 325 reactions and 53 species. The full mechanism is first integrated using an initial temperature of 1500 K and air-fuel feed mixture at atmospheric pressure with equivalence ratioφ= 0.5. The optimally reduced models have significantly fewer reactions, and vary in size between 51 and 135. The computational accuracy of the reduced models is compared to that of the full mechanism.Three atmospheric pressure, over-ventilated, axisymmetric, coflowing, laminar flames are simulated using AdapChem. The computational results obtained for the three laminar flames are discussed, showing that adaptive chemistry can simulate properly all the main features of the chemical structures of the flame with high efficiency. AdapChem approach provides a method for avoiding this loss of efficiency while retaining chemical accuracy where it is required. However, it is necessary to add radiation model in order to develop AdapChem for application to real combustion system. Discrete Ordinates Method for radiation model was used in this paper. When radiation is included, the temperature value decreases 20-100 K. The predicted results approach more closely the experimental ones when radiation is accounted for.