Fundamental understanding and modelling hypotheses assessments for Moderate or Intense Low-oxygen Dilution (MILD) combustion using Direct Numerical Simulations

Abstract

Moderate or Intense Low oxygen Dilution (MILD) combustion is a reliable method for achieving low emissions and high thermal efficiency. Despite its advantages, modelling MILD combustion is challenging and remains an open topic. To date, relatively limited effort has been directed to the modelling of MILD combustion in comparison to conventional combustion methods. Thus, a thorough physical understanding of MILD combustion and its differences compared to conventional combustion modes remain necessary. Moreover, an assessment of different conventional combustion modelling hypotheses is required to improve and develop MILD combustion modelling. The present thesis aims to address these aspects by carrying out a three-dimensional Direct numerical simulation (DNS) of homogenous and inhomogeneous mixture MILD combustion of methane. Conventional turbulent premixed and stratified flames are also considered in this thesis to provide a comparison with the corresponding MILD combustion cases. The simulations have been performed under different dilution levels and turbulence intensities using a skeletal methane/air chemical mechanism consisting of 16 species and 25 reactions. The fundamental building blocks of premixed flame modelling, such as Damköhler’s hypotheses, have been assessed for homogenous mixture MILD combustion using different reaction progress variable definitions. It has been found that Damköhler’s hypotheses hold in order of magnitude sense in homogeneous mixture MILD combustion only for reaction progress variable definition, which has a Lewis number close to unity (e.g. CH4). This has been partly attributed to the non-negligible role of the local differential diffusion, which affects the validity of Damköhler’s hypotheses even for statistically planar turbulent premixed flames with effective Lewis number close to unity. The differences in statistical behaviours of the surface density function (SDF) and scalar dissipation rate (SDR) between conventional premixed flames and MILD combustion have been investigated. The influence of the strain rates induced by fluid motion and flame propagation on SDF evolution has been investigated, and the findings suggest that the curvature dependence of displacement speed and the scalar gradient alignment with local principal strain rate eigendirections need to be addressed in the FSD/SDR based modelling methodology for MILD combustion. Furthermore, the SDR statistics in MILD combustion show qualitative similarities to passive scalar mixing, suggesting that a liner relaxation SDR closure has the potential to be applicable to MILD combustion. The reaction VIII rate predictions according to statistical (e.g., presumed probability density function), phenomenological (e.g., eddy-break up (EBU), eddy dissipation concept (EDC) and SDR-based approaches), and flame surface description (e.g., Flame Surface Density) based closures are compared based on a priori DNS analysis in the context of Reynolds Averaged Navier-Stokes (RANS) and large eddy simulations (LES). The usual presumed probability density function (PDF) approach using the beta function predicts the PDF of the reaction progress variable in MILD combustion. Moreover, a linear relaxation-based algebraic closure for the Favreaveraged SDR has been found to capture the behaviour of the Favre-averaged SDR. Furthermore, the EBU, SDR and FSD-based mean reaction rate closures do not adequately predict the mean/filtered reaction rate of the reaction progress variable for the parameter range considered. The variants of the EDC closure, which have been found to be more successful in predicting the mean/filtered reaction rate of the reaction progress variable compared to other modelling methodologies for the range of turbulence intensities and dilution levels, have been identified based on a priori DNS analysis.