Probabilistic simulation of glass fracture and fallout in fire

Window openings act as ventilation openings for the fire after the glass closing the window opening has been broken and fallen out. Because ventilation is one of most important factors influencing the fire severity, the question that when a window breaks during a fire to such extent that it forms a ventilation opening is one the most important problems of fire sciences. Due to the importance of the understanding of the performance of window glass in fire, it has been addressed in many outstanding theoretical and experimental studies. The models and programs developed predict the occurrence of the first cracking of the window pane exposed to fire and often a very good agreement with experiments can be obtained. The models do not, however, answer the question of when a window turns into a ventilation opening. They can give a very conservative lower bound estimate on the opening formation time, but experiments show that catastrophic window failure leading to glass fallout and creation of ventilation openings takes place at much higher temperatures and later times than the occurrence of the first crack in the window pane. Keski-Rahkonen was the first author to emphasise the variabilities involved in the problem of evaluation of response of window glass to fire: both the fires we have to consider in fire safety engineering and the glass response properties vary within a broad range. This report presents a probabilistic approach to evaluate the conditions when a window pane exposed to fire heating fails in such extent that it forms a ventilation opening. The variability in the glass response in included explicitly. The variability in the fire characteristics is not addressed directly, but in an indirect manner so that the output of the probabilistic glass failure model, i.e., is used as an input to the Probabilistic Fire Simulator (PFS) of VTT. The decoupling of the variability in the response of glass and the fire description reflects the pragmatic approach taken in this study: the objective of the work has been to establish well-founded guidelines and tools to fire safety engineers to treat glass breaking in fire, not to provide an all-inclusive calculation theory of the problem. In brief, the approach comprises two parts: the first one is calculation of the time and gas and glass temperature at the first occurrence of a crack in the window pane by using Monte Carlo simulation with the BREAK1 program. The second part involves assessment of subsequent crackings by using a more simple thermal response model of the glass, the isothermal lumped-heat capacity model. With this heating model we determine the glass fallout following the suggestion of Pagni that glass fallout results from multiple crackings. Thus, in our model, the glass is deemed fall out and form a ventilation opening when sufficiently many calculated crackings have taken place. The key factor, i.e., how many calculated cracking events constitute sufficiently many for the glass to fallout is obtained from experimental data found in the literature.