Assessment of Emissions Prediction Capability of RANS based PDF models for Lean Premixed Combustion of Methane

摘要

The high computational cost of Large Eddy Simulation (LES) makes Reynolds Averaged Navier-Stokes (RANS) methods the current standard for turbulent combustion modeling. Empirical models for turbulence, turbulence-combustion interaction and chemical kinetics are, however, a major source of uncertainty in RANS based combustion simulation. While Probability Density Function (PDF) based models overcome some of these issues, most commercial codes do not take full advantage of these models. In this study, lean premixed combustion of methane in a bluff-body combustor is simulated using two different reduced chemical mechanisms (ARM9 and ARM19) combined with the composition PDF transport combustion model in the commercial code FLUENT. Two different turbulence models, namely the RNG k-ε model and the Reynolds Stress Model (RSM) are used and the results of the simulations are compared to experimental data. For all the models tested, the prediction of temperature and major species (CH{sub}4, O{sub}2, CO{sub}2, CO, H{sub}2, and H{sub}2O) was good when compared to experiments. While all of the model predictions for the intermediate species OH showed an order magnitude difference (compared to the experiments) close to the bluff body surface; downstream axial locations showed good quantitative and qualitative agreement with the experiments. In a trend similar to the previous study (Nanduri et al., 2007) using the Eddy Dissipation Concept (EDC) model, predicted values for NO emission radial profiles showed an average difference of ±5 ppm when compared to experimental values. The results were also compared to the results of a velocity-composition joint PDF model developed by researchers at the University of Pittsburgh. In terms of emissions (NO and CO) predictions the relatively expensive composition PDF model in FLUENT did not give significant improvement when compared to the computationally cheaper EDC models. However, the velocity-composition joint PFD model used by researchers at the University of Pittsburgh did show significant improvement over EDC models in the prediction of NO. Both of the PDF models resulted in better qualitative and quantitative agreement in H{sub}2 prediction, thus showing the promise of PDF based models in simulating lean premixed combustion of fuel blends like hydrogen enriched natural gas.