Prediction of smoke properties and obscuration in compartment fires

This study describes the simulation and experimental investigation of a heptane pool fire, burning within a small compartment, in which interaction between a number of key physical processes is amplified. In particular, the configuration emphasises the coupling of buoyancy induced ventilation, smoke production and radiation heat transfer to the liquid fuel surface, from the luminous flame zone, from the smoke filled ceiling layer and from the confining walls. This study contrasts with those customarily performed for the purpose of model validation in compartment fires, which employ gas burners and so simplify much of the interaction. Initial experiments were carried out using a 0.23m diameter circular pan burning fixed amounts of heptane. Subsequently, a constant supply was used with a smaller circu- lar pan of 0.17m in diameter, in order to introduce experimental longevity under safe, controllable conditions whilst establishing a quasi steady-state system. Issues of non- stationarity in relation to heat-feedback to the fuel surface - an important pool fire mech- anism - are discussed. In addition to probe measurements of velocity and thermocouple temperature, the smoke yield was determined using a light extinction technique employing a 670nm wavelength diode laser and photo-diode detector, housed within a novel fully-traversible water- cooled probe. Data from these experiments illustrate the importance of accounting for room ventilation in terms of overall production of smoke and sound a cautionary note to the labelling of soot by a convenient marker such as temperature. Numerical simulation of the compartment fire is performed using the field model SOFIE, incorporating a simple evaporation model, which relates the mass-loss-rate of fuel to the net heat flux to the fuel surface and heat of gasification. This relationship assumes that heat losses to the pan, re-radiation by the fuel surface and other enthalpy loss terms, are small. Simulations of compartment fire scenarios using this model to calculate the rate of heat release are reported. Further comparisons are made between the industry stan- dard 'Eddy-Breakup' combustion model and the 'Laminar Flamelet' model. In general both the eddy-breakup model and laminar flamelet model tend to underpredict the yields of CO, whilst the eddy-breakup model over-predicts temperature and thus soot. The laminar flamelet approach shows more promise and shows particularly good agreement with the experimental measurements reported here under well ventilated conditions. SORE, the predictive tool employed in this research, has proved invaluable in discern- ing the reason for apparent ambiguities in the experimental measurements of soot con- centration. The results suggest that an alternative simplified zone model approach would overpredict visibility in smoke in terms of concentration, but underpredict in terms of layer depth, due to its inability to capture the important shape of the hot upper layer, which varies significantly from the homogenous, laterally uniform distribution which is assumed. The incorporation of a simple evaporation model which relies on accu- rate prediction of heat transfer in ultimately determining the heat release rate has been shown to be in very good agreement with the experiments. Despite the irregularity in predicted distribution of mass loss rate across the fuel surface - caused mainly due to the 'ray effect' of the radiation model - the main trend of lower heat transfer at the centre of the burner is demonstrated, in agreement with the experiments performed. This phe- nomenon is captured despite the lack of description of fuel vapour radiation blockage above the fuel surface, suggesting that this process may be disregarded. The heat flux distribution which is found here is in contrast to research conducted by other workers for similar sized pans in an open environment, which show a higher measured heat transfer at the centre of the burner. It has been shown that significant improvements could be made in experimental design of compartment fire experiments if CFD prediction is considered for the determination of suitable measurement locations in regions with lower local spatial variations.