Firstly this report has examined the characteristics of pre-flashover fire experiments for centrally located 55 kW, 110 kW, 160 kW fires, and a 110 kW corner fire, all of which were conducted in a two-compartment structure 7.2 m long × 2.4 m wide × 2.4 m high. Secondly, the accuracy of preliminary SMARTFIRE simulations for these fires was determined.
10.1 Conclusions from the Pre-flashover Fire Experiments
The following conclusions were reached from the pre-flashover fire experiments conducted at McLeans Island:
Overall:
• Temperature profiles in the two-compartment structure comprised of a hot upper layer and a cool lower layer, with a temperature gradient between the layers.
• Temperature profiles in the fire room varied significantly from the adjacent room.
• Lower layer temperatures in the fire room were significantly higher than the adjacent room.
• The 110 kW corner fire produced higher temperatures and higher CO2 concentrations in the upper layer than the equivalently sized 110 kW central fire (approximately double the concentration of CO2).
• The 110 kW corner fire produced a temperature profile in the fire room that varied significantly from the centrally located fires.
• Temperature variations occurred in areas of large temperature gradients and non-uniform mixtures between the hot fire gases and ambient air.
• Radiation from the fire increased the floor surface thermocouples temperature. • Ceiling surface thermocouples illustrated how heat is conducted into the
Fire Room:
• Temperature profiles on trees 1 and 2 for the centrally located fires produced well-defined upper and lower layers of constant temperature, with a temperature gradient between the layers starting at 1350 mm.
• Temperature fluctuations increased markedly at the temperature gradient between the upper and lower layers.
• Temperatures in the upper layer for the centrally located fires reached 130oC for the 55 kW fire, 200oC for the 110 kW fire, and 250oC for the 160 kW fire. • The 110 kW corner fire produced the highest temperature nearer the ceiling
reaching 335oC.
• The upper layer temperature profile of the 110 kW corner fire did not display a constant temperature profile as seen with the central fires.
• Corner temperatures at the back of the fire room illustrated slightly higher temperatures than the front corner thermocouples, with the difference in most cases being 5oC overall.
Adjacent Room:
• Temperature profiles for all trees showed a well-defined constant temperature lower layer near the ambient temperature, with an upper layer displaying a temperature gradient up to the ceiling. For all fires, temperatures increased for the upper layer, starting at 1850 mm.
• Temperature variations increased considerably above 1850 mm, and continued to remain high beyond this height.
• Temperatures in the upper layer for the centrally located fires reached 110oC for the 55 kW fire, 160oC for 110 kW fire, and 200oC for the 160 kW fire. • The 110 kW corner fire produced the highest temperature in the upper layer,
reaching up to 225oC at tree 6.
• The temperatures of the adjacent room corner thermocouples were lower than the fire room corner thermocouples. Overall the drop was about 25 – 30oC for the 55 kW fire, 30 – 35oC for the 110 kW fire, 60 – 70oC for the 160 kW fire,
10.2 Conclusions from the SMARTFIRE Simulations
• Predicted temperature profiles suggested the formation of upper and lower layers associated with pre-flashover fires.
• The temperature profile for all simulations in the upper layer in the fire room displayed a temperature gradient rather than a constant temperature profile. • The upper layer height in the fire room was predicted to be lower than the
adjacent room upper layer height.
• Simulations, including the simulation with the six-flux radiation sub-model, produced lower temperatures in the upper layer than the simulations without the six-flux radiation sub-model.
• Upper layer temperature profiles were more horizontal for the simulations which included the six-flux radiation sub-model.
• Close to the ceiling the 110 kW corner fire simulations produced the highest temperatures, which rapidly decreased with descending height.
• By specifying the fire as a volumetric block, SMARTFIRE was unable to simulate flame layback as a result of incoming air, with little or no layback of the fire’s plume.
• Simulations where the six-flux radiation sub-model was incorporated required more computational time than simulations without the six-flux radiation model.
10.3 Conclusions from the Comparisons
Fire Room:
• All simulations predicted constant lower layer temperature profiles, but underestimated lower layer temperatures.
• All simulations for the centrally located fires predicted a temperature gradient in the upper layer for trees 1 and 2, rather than the constant temperature profile seen experimentally.
• Temperatures close to the ceiling were often overestimated by both simulations, with the simulations predicting a temperature profile of a ceiling
• Simulations for the 110 kW corner fire produced corresponded the least with experimental temperatures.
Adjacent Room:
• Both simulations under predicted the lower layer temperature slightly.
• The simulated upper layer temperature profiles generally corresponded well with experimental results.
Doorway:
• Both simulations predicted the lower layer temperature well.
• Upper layer temperature profiles were generally simulated with reasonable accuracy.
10.4 Further Research
Experiments at McLeans Island:
1. Aspirated thermocouples need to be used to determine the magnitude of the effects of radiation on the thermocouples.
2. The location of the fire should be varied throughout the fire compartment, eg against the back wall. Also, experiments need to be carried out, whereby for each experiment, the fire’s position is moved in increments away from the wall or corner.
3. Experiments should be conducted with the inclusion of a doorway and a soffit on the adjacent compartment.
SMARTFIRE simulations:
1. An investigation needs to be carried out to determine the best settings for the simulations.
2. The sensitivity of the solution’s dependence on grid size needs to be investigated.