Large Eddy Simulation of multi-injector flame blow-off sensitivities to inlet biases

Abstract

Reactant biases of mass flow rate or stochiometry can result from design trade-offs in industrial implementations of multi-injector, lean-premixed flames. Rules for maximising the lean-extinction limit require additional insight from experiments and/or computations as global scalings may not necessarily apply. Models, however, need extensive validation as the timescale separation between chemistry and turbulence decreases towards the lean limit, and a larger range of thermochemical states may be present. This leads to difficulties in parametrising them accurately. In this work, large eddy simulation (LES) is used to model blow-off in a linear array of lean, swirling, methane-air flames at atmospheric conditions. The LES methodology is assessed with regard to reproducing partial blow-off due to reactant equivalence ratio ($\varphi$) and flow rate ($\dot{m}$) biases. It is found that the blow-off transients at ideal (no bias) and biased conditions are similar with regard to the large-scale effects. Progress variable based flamelet generated manifolds (FGM), as well as transported species, are employed and contrasted. Both methods could reproduce the highly transient nature of blow-off, though the flamelet strategy underpredicts blow-off for some conditions. Using flame-resolved simulations, it is shown that the combustion regime near and during blow-off allows applying flamelet methods. However, the scatter of thermochemical states appears to require more than strain and enthalpy as manifold parameters.