Development of fire simulation models for radiative heat transfer and probabilistic risk assessment: Dissertation

An essential part of fire risk assessment is the analysis of fire hazards and fire propagation. In this work, models and tools for two different aspects of numerical fire simulation have been developed. The primary objectives have been firstly to investigate the possibility of exploiting state-of-the-art fire models within probabilistic fire risk assessments and secondly to develop a computationally efficient solver of thermal radiation for the Fire Dynamics Simulator (FDS) code. In the first part of the work, an engineering tool for probabilistic fire risk assessment has been developed. The tool can be used to perform Monte Carlo simulations of fires and is called the Probabilistic Fire Simulator (PFS). In Monte Carlo simulation, the simulations are repeated multiple times, covering the whole range of variability of the input parameters and thus resulting in a distribution of results covering what can be expected in reality. In practical applications, advanced simulation techniques based on computational fluid dynamics (CFD) are needed because the simulations cover large and complicated geometries and must address the question of fire spreading. Due to the high computational cost associated with CFD-based fire simulation, specialized algorithms are needed to allow the use of CFD in Monte Carlo simulation. By the use of the Two-Model Monte Carlo (TMMC) technique, developed in this work, the computational cost can be reduced significantly by combining the results of two different models. In TMMC, the results of fast but approximate models are improved by using the results of more accurate, but computationally more demanding, models. The developed technique has been verified and validated by using different combinations of fire models, ranging from analytical formulas to CFD. In the second part of the work, a numerical solver for thermal radiation has been developed for the Fire Dynamics Simulator code. The solver can be used to compute the transfer of thermal radiation in a mixture of combustion gases, soot particles and liquid droplets. The radiative properties of the gas-soot mixture are computed using a RadCal narrow-band model and spectrally averaged. The three-dimensional field of radiation intensity is solved using a finite volume method for radiation. By the use of an explicit marching scheme, efficient use of look-up tables and relaxation of the temporal accuracy, the computational cost of the radiation solution is reduced below 30% of the total CPU time in engineering applications. If necessary, the accuracy of the solution can be improved by dividing the infrared spectrum into discrete bands corresponding to the emission bands of water and carbon dioxide, and by increasing the number of angular divisions and the temporal frequency. A new model has been developed for the absorption and scattering by liquid droplets. The radiative properties of droplets are computed using a Mie-theory and averaged locally over the spectrum and presumed droplet size distribution. To simplify the scattering computations, the single-droplet phase function is approximated as a sum of forward and isotropic components. The radiation solver has been verified by comparing the results against analytical solutions and validated by comparisons against experimental data from pool fires and experiments of radiation attenuation by water sprays at two different length scales.