Joint-scalar transported PDF modelling of soot in a turbulent non-premixed natural gas flame

Abstract

The focus of the present work is on the prediction of soot in the turbulent Delft III/Adelaide natural gas flame at a Reynolds number of 9700. A parabolic flow solver with the SSG (Speziale, Sarkar and Gatski) Reynolds stress transport model for turbulence is coupled to a joint-scalar transported PDF (probability density function) approach allowing exact treatment of the interactions of turbulence with the solid and gas phase chemistries. Scalar mixing is treated via the modified Curl's coalescence/dispersion model and two different closures for the scalar dissipation rate are explored. The gas phase chemistry is represented by a systematically reduced mechanism featuring 144 reactions, 15 solved and 14 steady-state species. The dynamics of soot particles, including coagulation and aggregation in the coalescent and fractal aggregate limits, is treated either via a simplified two-equation model or via the MOMIC (method of moments with interpolative closure). The inclusion of soot surface reactions based on a second ring PAH (polycyclic aromatic hydrocarbon) analogy is also investigated. Soot oxidation via O, OH and O(Formula presented.) is taken into account and the sensitivity to the applied rates is investigated. An updated acetylene-based soot nucleation rate is formulated based on consistency with detailed chemistry up to pyrene and combined with a sectional model to compute soot particle size distributions in the NIST well-stirred/plug flow reactor configuration. The derived rate is subsequently used in turbulent flame calculations and subjected to a sensitivity analysis. Radiative emission from soot and gas phase species is accounted for using the RADCAL method and by the inclusion of enthalpy as a solved scalar. Computed soot levels reproduce experimental data comparatively well and approximately match absolute values. The axial location of peak soot in the Delft III/Adelaide flame is consistent with previous large eddy simulations and possible causes for the discrepancy with experimental data are analysed.