Heat fluxes and flame heights in façades from fires in enclosures of varying geometry

Abstract Fire spread in high rise buildings from floor to floor occurs if flames emerge and extend on the facade of the building to cause ignition in floors above the floor of fire origin. Even though considerable effort has been exerted to address this issue, proposed relations for flame heights and heat fluxes are incomplete and contradictory because the relevant physics have been poorly clarified. By performing numerous experiments in small scale enclosures having various door-like openings and fire locations, the physics and new relations are underpinned for flames on facade emerging from (under-ventilated) ventilation controlled fires at the floor of fire origin. To limit the variables and uncertainties, propane and methane gas burners created controlled (theoretical) heat release rates at the source. Gas temperatures inside the enclosure and at the opening, heat fluxes on the facade wall, flame contours (by a CCD camera) and heat release rates (by oxygen calorimetry) inside and outside the enclosure have been measured. The gas temperatures inside the enclosure were uniform for aspect ratio (length to width) of the enclosure varying from one to three to one. Previous relations for the air inflow and heat release rate inside the enclosure were verified. These flames are highly radiative because soot can be formed at high temperatures inside the enclosure before the combustion gases and the unburned fuel exit the enclosure. Remarkably the efficiency of combustion is one for well over-ventilated and very under-ventilated fires by it dropped to 80% for burning conditions around stoichiometric. The flame height and heat fluxes have been well correlated by identifying new length scales related to the effective area of the outflow and the length after which the flow turns from horizontal to vertical due to buoyancy. The results can be used for engineering calculations for real fires and for validation of new large eddy scale simulation models.