Feasibility study of correlating mass quantity output and fuel parameter input of different simulations using Fire Dynamics Simulator

Fire accidents often spur massive attention due to the associated potential for enormous consequences for lives and property. To facilitate prevention and mitigation of fire accidents, knowledge as to the dynamics associated with fire development and spread of smoke and hot gases needs to be ascertained. To this end, forensic investigators, engineers and scientists apply software for fire simulation. Such software embodies mathematical models of the intricate physical processes associated with fire. Using such simulations, engineers and others can compute, for instance, the duration of which conditions in an egress route are tenable. Tenable conditions may often be related to concentrations of soot (which obscure visibility) and carbon monoxide (CO, which can be asphyxiating), and can rarely be calculated more precise than by use of simulations. A drawback of such simulation tools are the computational costs of conducting simulations of a fire scenario. This can be highly time consuming and costly in projects. Additionally, it is often necessary to conduct several simulations, where simulation input is revised between simulations, further increasing time consumption. This is necessary, due to the uncertainty often associated with input parameters applied for the simulation. For instance, there is seldomly pristine knowledge as to what commodities (or fuels) may be involved in a building fire, or what energy levels and species yields combustion may lead to. The purpose of this study is to investigate the feasibility of correlating simulation fuel input data to mass quantity output data (such as concentrations of soot and CO). This between two simulations applying different fuel parameter input. As such, the correlations may be used to estimate the mass quantity output of one simulation based on the mass quantity results of a base simulation. Provided reasonable estimations can be produced from this procedure, the computational cost of an analysis involving simulations may be reduced. The correlations are deduced from functions paramount to the mass transport equations of a much-used simulation software, Fire Dynamics Simulator (FDS). Two functions of correlation, or correlation factors, have been proposed and developed. Correlation factor 1 considers the species mass source term of the FDS mass transport equation. Correlation factor 2 is based on the method FDS applies to collect information as to the species mass fractions of the smoke. To investigate the feasibility of correlation, 93 simulations have been conducted in 19 series. Each series as comprised of 4-5 simulations. Individual series applied simulations of different fire scenarios. Between every simulation in each series, only fuel input parameters were altered. One of the simulations in each series was used as a base simulation. The mass quantity output of the base simulation was multiplied with the correlation factors, to ascertain estimations of mass quantity results of all other simulations in the series. The estimations were further compared to the actual simulations, to assess the performance of the correlation factors. This, for instance, in relation to potential tendencies for over- and underestimations associated with the characteristics of the fire scenario. The mass quantities considered and estimated were soot densities and mass fractions of CO. These quantities were statistically measured (mean and maximum values) in stationary volumes in different locations within the simulated enclosures. The findings of this study suggest that reasonable estimations of mass quantity output may be produced by applying the correlation factors and a base simulation. Out a total of 744 mass quantity estimations, 82,2% and 94,8% deviated from the simulated mass quantities within the percentile intervals of ∓5% and ∓10% respectively. Approximately 0,5% deviated above 20%, with a maximum deviation of approximately 22%. One of the findings regarding tendencies of over- and underestimations, is that high values of heat release rate (HRR) and/or measurements made in the vicinity of the fire origin, rendered simulation output of different simulations in a series more similar than estimated by the correlation factors. This led to underand overestimations for low and high values of the correlation factors respectively. Adversely, for low values of HRR and/or for measurements made somewhat far away from the fire origin, low and high values of the correlation factors led to over- and underestimations respectively. Further, it was found that correlation factor 1, which considers the species mass production rates, provided more precise estimations than correlation factor 2, for the conducted simulations. The differences were, however, rather small. These, and other findings, mainly apply to the simulations conducted in this study. However, it is likely that many of the identified tendencies are applicable to similar scenarios. As this is challenging to precisely determine, the correlation factors should be used cautiously. The most important limitation of the correlation factors has generally been the difference of fuel parameter input between the base simulation and the simulation of which estimations were made. Large differences, indicated by the difference between the value of the correlation factor and unity, generally coincided with increased potential for estimation discrepancies. In this study, many of the correlation factor-values varied between 0,14 and 1,7. Nonetheless, limitations considered, the correlation factors can likely be used to ascertain increased knowledge as to the possible consequences of altering simulation fuel parameter input of a base simulation, and this in a relatively short amount of time