Residential occupancies are the most common type of buildings where compartment fires occur. In order to provide fire protection in these occupancies and thus to significantly reduce the number of future victims, the prediction of the fire effects during the first minutes becomes crucial. Particularly, forecasting fire dynamics (i.e. fire growth, smoke dispersion, etc.) may send early warnings to occupants and may provide important information for the fire service rescue team. Nevertheless, validation analysis should be previously performed to assess the predictive capabilities of different modelling tools when determining the harmful fire effects in residential compartments.

The present paper describes three fire experiments undertaken in naturally-ventilated residential compartments. The aim of current work is to assess the fire effects prediction performance by means of different fire models compared to experimental data. The Q ¿ curves, which are derived from the temperatures measured inside the compartments, are employed as input data for the fire models used: an analytical model, a two-zone model (CFAST) and a field model (FDS). Recommendations on models choice are provided according to the results accuracy and the computational time required for the modelling tools employed. Preliminary outcomes reveal similar experimental fire behaviour in terms of temperatures, heat release rates and smoke layer heights among the three fire scenarios analysed. In addition, gas temperatures were correctly predicted with FDS; whereas the analytical model and CFAST were the most appropriate methods to forecast the smoke layer heights according to the experimental data collected.