Towards predictive engineering-type simulations of upward flame spread in SBI scenarios

Abstract

Large eddy simulations of upward flame spread using FireFOAM are presented. Aiming at advancing predictive fire modelling, the approach considers the use of dynamic models, with limited use of model constants, for turbulence, combustion, and radiation. Modelling of convective heat transfer is based on Newton’s law of cooling considering simplified correlations for natural convection. The thermal decomposition of the solid material is represented through a 1D pyrolysis model with optimized model-effective material properties. For validation purposes, medium-scale Single Burning Item (SBI) experiments are used, involving both inert materials (calcium silicate) and flammable walls involving both charring (MDF and plywood) and non-charring (PMMA) materials. Separate validations for the gas and solid phase are also presented. A detailed comparison between the CFD predictions and experimental data is performed, focusing on global parameters (i.e., HRR, mass loss rate, heat feedback) and local quantities (i.e., total heat fluxes). The modelling approach performs very well, with predictions being fairly grid-insensitive, showing relative differences in the predicted HRR of up to 47% between the simulations and the experiments. Convection contributes up to 30% of the total wall heat feedback, highlighting the importance of accurately modelling convection alongside radiation in early flame spread.