Fire Modeling: Where Are We? Where Are We Going?

Abstract

Current trends in computational fluid dynamics (CFD) modeling of fire are discussed, in particular the widening communication gap between model developers, model users, and the larger fire protection engineering community in regard to combustion and turbulence. The paper suggests steps that can be taken by all of these groups to improve the current situation and move forward to develop better fire models. Last year, I visited the Large Fire Lab at NIST to observe an experiment recreating the initial stages of the fire that killed 100 people in a Rhode Island nightclub in February, 2003 (1). One of the aims of the experiment was to test our fire model, the Fire Dynamics Simulator (FDS), which was being used in the NIST investigation of the tragedy. I asked one of the technicians who was setting up the experiment, Lauren DeLauter, whether or not he expected to see flames emerge from a door at the rear of the test compartment. He said no, because the polyurethane foam lining the walls would burn rapidly, the compartment would become under-ventilated, and the unburned fuel gases would cool before reaching the exit. Because the mock-up was not designed to duplicate the geometry of the entire club, but rather the area in the immediate vicinity of the stage, his assessment did not contradict what was observed in the Rhode Island fire. Indeed, he was correct, whereas our model was wrong, in that it predicted flames emerging from the exit. Although it appears to many that numerical simulation is the cutting edge of fire protection engineering, many non-modelers are surprised to learn that our ability to reproduce fire phenomena via computer simulation lags our empirical understanding by about 10 years. Indeed, current generation zone and field (computational fluid dynamics or CFD) models address transport phenomena reasonably well, making them useful for many engineering applications. For example, FDS did successfully replicate many of the phenomena associated with the Rhode Island nightclub fire. However, it has not yet reached the point of reliably predicting, for large scale applications, such important phenomena as flame spread, extinction, suppression, and CO and smoke production, all of which demand more detailed chemistry and physics than are currently incorporated in the model. Hindering our efforts to move forward is the increasing level of miscommunication between modelers and experimentalists, scientists and engineers, and even professors and students. For various reasons, those who actually write the computer programs too often cannot or will not explain which algorithms work, which do not, and which have no effect at all on the results. This communication gap leads to unwarranted claims by end users who believe the models are predicting more than they actually are. Moving ahead will require that we bridge these gaps, so that all sectors of the fire and combustion