Numerical Simulations of Small-scale and Full-scale Fire Experiments

Abstract

A fundamental part of fire safety engineering is dedicated to the application of numerical fire models. Accurate predictions of real-life fires are needed in scenarios related to fire growth, smoke propagation, occupant egress, and structural integrity. In the context of building safety, fire modelling tools can be used to predict the response of materials to fire situations, and are increasingly prevalent in performance based design. In the present work, heat transfer and fire simulations are created with the objective to predict the resultant fire effects of different experiments. The simulations range in complexity from algebraic finite difference models to computational fluid dynamics (CFD) calculations. For each set of simulations, numerical predictions are compared with experimental data, whenever available. FireFOAM, an open source computational fluid dynamics solver, is selected as the modelling tool of choice. In the present study, four sets of simulations are conducted based upon experimental work. Firstly, a small scale test apparatus, the cone calorimeter, is investigated. Predictions from both a finite difference model and a CFD model compare favourably to the experimental results, and it is confirmed that a 1D finite difference model is not appropriate for the experimental configuration. Secondly, a full-scale fire experiment is investigated. The CFD simulations are extended to include the effects of turbulence and combustion. Large Eddy Simulation (LES) is selected for the turbulence modelling with a one equation eddy-viscosity model. Infinitely fast chemistry is assumed, and the eddy dissipation concept (EDC) is employed where combustion is controlled by the rate of turbulent mixing. Thirdly, a two-step reaction mechanism is implemented to account for compartment fires with under-ventilated combustion and more complex fuels. Chemistry based upon Arrhenius rate constants is assumed, and the Partially Stirred Reactor (PaSR) approach is employed. Good agreement is found for species and temperature predictions, with over-prediction of carbon dioxide concentrations due to modelling the reaction rates too fast. Finally, a preliminary CFD study is carried out for a multi-compartment fire where a wall section separates two compartments. Heat transfer is found to be over-predicted through the non-degrading wall section. To enhance the capabilities of the simulations, pyrolysis is recommended to be implemented to enable modelling of representative wall sections and realistic fuel loads.