New developments in data regression methods for the characterization of thermal stratification due to fire

• Data regression methods for analyzing temperature stratification.
• Thermal stratifications with constant temperature gradient.
• Application for fire in closed and forced ventilated compartment.
• Methodology for identifying the most suitable data regression method.

The study deals with the data regression methods used for validating zone code modeling in comparison with experimental fire tests. These methods aim at determining both the temperatures of upper and lower zones and the interface location calculated from experimental vertical temperature profiles. A new data regression method is proposed to improve the determination of upper and lower temperatures and position of the interface. This new method merges both the thermal stratifications showing constant temperature in the upper layer as often encountered for well ventilated fire (approach of Quintiere et al. [4]) and those showing constant gradient as often observed for fire in closed and forced ventilated compartment (approach of Audouin et al. [8] and [9]). These two latter existing methods and the new former one are then applied to large scale fire tests experiments performed in IRSN facilities. The major outcomes indicate that the most appropriate regression method depends on the basic shapes of the experimental vertical temperature profiles. The interface height is the most sensitive variable to the choice of the regression methods. Indeed, discrepancy higher than 100% can be found concerning this last variable. In contrast, the temperature of lower zone appears to be the less sensible variable. The discrepancy on the upper temperature is found also significant in some cases. From these investigations, guidelines are proposed to improve the data regression process needed for validating zone codes and are supported by a thorough analysis of these three data regression methods using experimental data from large-scale fire tests.