Evaluation of the predictive capabilities of current computational methods for fire simulation using the HDR T51 and T52 tests with a focus on performance-based fire codes.

摘要

Regulatory groups from various disciplines have a great interest in moving from a prescriptive code system to a performance-based code system. For such a shift in regulatory behavior to occur, there will need to be analytical methods capable of analyzing structures from determining their level of fire safety. Current computational tools, however, have seen little application to full-scale fires in real life structures. Thus, their applicability to a performance-based code system is questionable.; This work uses three methods: hand calculations, a zone model code (CFAST), and a computational fluid dynamics code (FDS), to examine two fire tests from the HDR facility, a decommissioned reactor containment building. The two tests, T51.23 and T52.14, used different fuels, propane gas and a hydrocarbon solvent, and occurred in two quite different locations, low in the containment and just below the operating deck, respectively.; The two tests both pose different challenges for analysis. The T51.23 test contains a simple fuel, propane gas premixed with air, but a fairly complex geometry for the fire compartment with restricted ventilation. The T52.14 test uses a more complex fuel, a hydrocarbon solvent that burned in an underventilated manner, with a somewhat simpler geometry for the fire compartment. However, the T52.14 test contained a large atria-like space in close proximity to the fire room which poses an additional challenge. Together these two tests can examine the suitability of computational methods for calculating fires in real-life structures containing multiple levels with complex compartment interconnections.; For both computer codes and both tests several models were used to simulate the tests. The different models, which varied in complexity, were used to in order to examine the effects of modeling assumptions on the codes' predictive quality.; Results of the three methods of analysis are compared to the test data. Furthermore, the comparisons are made in terms of a performance-based code system and, thus, compare the analysis results versus the data in terms of desired analytical performance. Conclusions on the current state-of-the-art in fire prediction in terms of usefulness in a performance-based code regime are made.