Model uncertainty in a parametric fire curve approach: A stochastic correction factor for the compartment fire load density

Abstract

A commonly used approach to represent the thermal load in a compartment fire is the Eurocode Parametric Fire Curve (EPFC), which specifies gas temperatures (or rather adiabatic surface temperatures). Recognizing the significant deviations between real fires and the EPFC framework, the concept of model uncertainty is explored. This study does not aim to assess or improve the EPFC, but introduces a model uncertainty, allowing for reliability-based structural fire engineering (SFE). It presents a stochastic correction factor for the fire load density, based on the maximum temperature in steel sections. The focus is on the fire load density, but in general other parameters can be jointly taken into account as well. This correction factor considers protected and unprotected sections, incorporating variations in section factor and protection thickness. The findings reveal that the fire load density within the EPFC framework can be modified to better represent the severity of fire experiments. This approach ensures physical consistency of the obtained compartment gas temperatures, as opposed to alternative approaches for addressing the EPFC model uncertainty. While promising results are evident in this proof of concept, exploration for other types of structural elements and evaluation for structural systems is necessary before integration into design practices.