ABSTRACT
ABSTRACT Probabilistic structural fire design enables the assessment of loss of life or property that may occur from fire events. The effects of fire hazards on buildings are interpreted in terms of levels of damage in order to predict probabilities of failure or losses. As there is limited data on monitored response of buildings to real structural fires, current failure predictions are based on numerical simulations of structural behaviour. However, numerical simulations are model dependent, which implies that the same structure could have different responses if modelled under different configurations. These differences could be significant and may affect failure estimations, and hence the probability of failure. Uncertainty in structural modelling is classified as epistemic uncertainty. The response of a structural element is different if modelled as an isolated member (2D) or as part of a structural system (3D). For the probabilistic assessment of a given fire hazard, the variation in structural response due to different styles of modelling (2D or 3D) produces different annual probabilities of structural response. The quantification of this epistemic uncertainty is important because it provides information about the risk involved in the design solution due to structural modelling. This is demonstrated with the help of a case study of a composite steel beam, modelled both as an isolated element and as a part of a 3D structure, exposed to a suite of fire profiles. The annual probabilities of failure for both structural types are evaluated and compared to highlight the effects of structural modelling on failure probability.
