Abstract
Probabilistic fire simulation has been gaining interest in performance-based design approach. Very few evaluations on modelling performance can lead to unnecessary simulations which limit the understanding of fire engineering practitioners to select suitable model for a specific fire phenomenon. This study presents comparison of models' capabilities to recreate kitchen fire experiment through probabilistic assessments when there are various possibilities of fire spread can occur within the compartment. Such situation must consider the possible heat release rate curve from previously burning kitchen item, as this will effect on the possibility that other item igniting, and so enhance the intensity of fire. Determination of accumulated heat release rate data is assessed through the use of B-RISK and simplified model. Simulation input employs design fire for single furniture that was characterized through combination of peak growth and exponential decay functions using statistical analysis of fire severity characteristics from collated experimental data. Probabilistic correlation discloses that both models are able to produce results within an acceptable range of fire practice (> 20%), but B-RISK has better representation of possible cumulative heat release rate region than simplified model when more than 58% similarity of fire spread predictions among multiple items is achieved over different boundary ranges.