Probabilistic Analysis of Building Fire Severity Based on Fire Load Density Models

Abstract

This paper presents the compilation of 35 fire load density surveys for office and residential buildings. Limited data can be obtained for contemporary buildings. Using recent fire load survey results and the Bayesian inference theory, probabilistic fire load density models are proposed for office buildings (up to 120 m2 floor area) and for residential buildings (up to 35 m2 floor area). These models predict that the mean fire load density decreases when the floor area becomes larger. Meanwhile, given floor area, the models indicate that the fire load density follows the lognormal distribution. Statistics of fire load density are then calculated and explicitly presented for a direct reference in fire safety design. Afterwards, these models are incorporated in the parametric temperature model presented in the Eurocode to predict the maximum temperature within fire compartments. Parameter uncertainties are also considered, including the variations in opening factor and thermal absorptivity of enclosure that are observed from 57 experiments. Based on the proposed fire load density models, building fire severity is estimated using empirical equation in the Eurocode considering the influence of these parameters. Monte Carlo simulations are performed to obtain the statistical values of the maximum temperature and fire severity. The results in this work can be used in a performance-based fire safety design.