Fire Safety Science Digital Archive

The behaviour of glazed windows were investigated in a series of flashover fire experiments conducted in a full-scale multi-storey building. Glass cracking and dislodgement processes were monitored during the experiment. A simple transient heat transfer model was established to estimate window glass temperature and the time of fracture under enclosure fire conditions. The model is incorporated into a fire growth model NRCC-VUT and an external fire spread model ExSpread to predict the effect of fire growth on window glass breaking and the time for external fire spread to occur via external windows respectively. The fire growth model was compared against the result of one experiment in which the glass window was allowed to disintegrate naturally. In this test, the period from initial cracking to complete dislodgment took 160 s. Hence, although it may be possible to predict the onset of cracking, the time for complete dislodgment to occur is not insignificant and not readily predictable. Accordingly, the refinement in the use of a heat transfer model for predicting glass breakage may not necessarily improve the predictions of room fire behaviour over the relatively simple room temperature approach for glass dislodgment. The failure of the external glass window located at the level above the fire enclosure were recorded in all four tests and were simulated with the external flame spread model. It was observed that following the initial dislodgment, there was a further delay of 1-4 minutes before complete dislodgment took place. A relatively high averaged glass failure temperature of 500 K was used in the model to allow for window dislodgment to occur. The model predicts the time of initial dislodgment reasonably well but conservatively assumes that it corresponded to complete dislodgment for the purpose of predicting time for flame spread via radiation from an external plume.