Abstract

The challenge of a posteriori (i.e. after the event) modelling of a well characterized full-scale fire test using computational fluid dynamics is illustrated in this work. Test One of The Dalmarnock Fire Tests was conducted in a 3.5 m by 4.75 m by 2.5 m concrete enclosure with a real residential fuel load. It provides measured data at the highest spatial resolution available from a fire experiment to date. Numerical simulations of the growth phase have been conducted with the numerical code Fire Dynamics Simulator (FDSv4) while having full access to all the measurements. This includes description of the fuel load, compartment layout, temperature and heat flux data, and camera footage. No previous fire simulation had this large amount of data available for comparison. Simulations were compared against average and local measurements. The heat release rate is reconstructed from additional laboratory tests and upper and lower bounds for the fire growth are found. Within these bounds, the evolution of the average hot layer temperatures in time could be reproduced with 10-50% error. Worse agreement with local measurements of gas temperatures was achieved (between 20% and 200% error). Wall temperatures were predicted with less than 20% error and wall incident heat fluxes within 50-150% error. Largest discrepancy is seen close to the flames. This level of agreement with the measured data was achieved after Exploratory simulations were conducted and several uncertain parameters were adjusted and readjusted while having full access to all the measurements. (C) 2010 Elsevier Ltd. All rights reserved.