Modelling of premixed combustion in a gas turbine

Abstract

Three steady state combustion models, two turbulence models and a model for tK'6 prediction of NO., were implemented and investigated on a simple backward facing step experiment as well as an experimental lean prevaporised premixed (LPP) combustor. The three combustion models included the simple Eddy Break-up model as well as a presumed probability density function (pdf) model and a form of the BML crossing frequency flamelet model. These models were adapted to consider a variable mixture fraction to account for a non-homogeneous fuel air mixture. The two turbulence models used were the k-e and second moment models. Despite being unable to capture the flame front spreading in the case of the backward facing step, these predictions provided insight into the performance and implementation of the models. All three of the combustion models, after appropriate tuning, worked well for the LPP test combustor. This illustrates that such time averaged models are useful for flows which do not contain large transient coherent structures, such as that of the LPP test combustor and most practical engine combustors designed today. The second moment closure turbulence model was found to have the greatest impact on the flame front through the flow field predictions rather than through counter gradient diffusion. The Eddy Break-up and BML crossing frequency models both performed very well, qualitatively predicting the correct trends. The additional consideration of flame front straining in the BML crossing frequency model did not appear to significantly influence the flame front. This is because the type of model adopted to predict this effect had a relatively uniform influence everywhere in the flow. The presumed pdf model also performed well and was additionally found to self ignite without the existence of hot products when the inlet temperature was high enough. The NO., model faired well for a simple experimental geometry. However, it considerably over predicted the NO., formed within the LIPP test combustor, which was most probably due to poor boundary conditions. Despite this overprediction, the results give insight into how to improve the NQ, emissions for the experimental combustor.