A Scale Separation Method for Pollutant Prediction in Turbulent Flames Using Transported Scalars With Flamelet Generated Manifold (FGM) Method

Abstract

In this work, a scale separation method has been proposed and implemented in the framework of Flamelet Generated Manifold (FGM) model. In this approach, first a list of slow evolving species like NO, N2O etc., are identified. Then, a separate transport equation for each of these species (called FGM scalars) is solved in addition to the mixture fraction and progress variable equations. The forward and reverse reaction rates of these slow forming species are computed in two-dimensional FGM flamelets and pre-tabulated as a function of progress variable, mixture fraction and their respective variances. At run time, the pre-tabulated probability density function (PDF) averaged production rates of these FGM scalars are used, while their tabulated reverse rates are modified with a linear scaling based on the ratio of tabulated values of the FGM scalar and the prevailing values of the FGM scalars from three dimensional CFD solution. This mechanism allows the reverse rates to provide continuous feedback and respond to the slow evolution of scalar. Other than the list of selected scalars, all other species and temperature are still computed as a function of the main progress variable and mixture fraction. Since, a small set of scalars can be used to track key species, this methodology remains computationally efficient. The current approach has been implemented into commercial CFD solver, ANSYS Fluent, and has been validated for two lab scale turbulent flames, the first one is Sandia Flame D, while the second one is a lifted turbulent methane flame in vitiated co-flow. In the current work, two additional FGM scalar transport equations are solved for CO and NO and comparisons have been made against the tabulated values as well as the experimental data. It has been seen that the scale separation methodology of these scalars leads ∼10–15% improvements in the CO mass fraction, while it reduces the peak NO formation up to 4 times leading to better agreement with experimental data compared to tabulated values. The quality of predictions from the current method is also evaluated against finite rate chemistry-based model as well as reduced order NO model. It is found that the current model has consistent results, and is an improvement over current reduced order modeling approach.