Compartment fire risk analysis by advanced Monte Carlo simulation

Abstract

Quantitative fire risk analysis aims at providing an assessment of fire safety on a scientific basis and taking relevant uncertainties into account in a rational quantitative manner. Under a probabilistic approach, performance measures are formulated as multi-dimensional probability integrals, whose efficient computation is pivotal for practical implementation. Direct Monte Carlo method is a well-known technique, but it is not efficient for investigating rare failure events which are commonly encountered in engineering applications. A recently developed stochastic simulation approach called Subset Simulation is presented for quantitative fire risk analysis with a focus on the critical temperature in a compartment fire event. In the method, random samples leading to progressive failure are generated efficiently and they are used for computing probabilistic performance measures by statistical averaging. The random samples can also be used for probabilistic failure analysis, which yields information conditional on the occurrence of the failure event. A global approach is adopted for incorporating the uncertainties in the functionality of active fire measures into the fire risk analysis, where the failure probabilities can be obtained by a single simulation run rather than by multiple runs exhausting the possibilities in the associated event tree.