Cornelius Albrecht, Dietmar Hosser - iBMB Fire Protection Division, University of Braunschweig
The process of designing for life safety is usually performed using parameters which are known to be subjected to high uncertainties. Hence, the choice of the appropriate deterministic parameters is usually very conservative to envelope those uncertainties. Until recently, the probabilistic assessment of life safety using CFD and microscopic evacuation models only seemed to be theoretically possible as a Monte Carlo simulation of CFD-based fire design is unfeasible due to the high computational cost. In this paper we present a fast and efficient adaptive response surface approach to perform probabilistic life safety analysis using state-of-the-art fire engineering tools, namely the CFD fire simulation code FDS and the built-in FDS+evac microscopic evacuation software, to assess the reliability of life safety during a hostile fire. This approach also allows for sensitivity analysis to identify the critical input parameters. As installed protection barriers (i.e. sprinklers etc.) are still prone to possible failure, a subsequent system analysis can yield information about the total fire protection system reliability considering the possible malfunction of the barriers. Additionally, the identification of the most critical scenarios is possible. The methodology will be exemplary applied to a small size assembly building for multiple scenarios and barriers, and the results will be discussed in detail. Probabilistic assessment of the life safety using state-of-the-art engineering tools delivers valuable information about the behavior of a fire protection system, influence of the various possible barriers and it allows for a quantitative comparison of multiple strategies in order to identify the most cost-effective solution.