Numerical modelling of pool fires and flame spread
A historical analysis by Planas-Cuchi ‘et al’. showed that pool fires are one of the most frequent fire accidents in family homes, in the processing industries and in the transport of hazardous substances. Experimental and theoretical studies of the turbulence structure, fluid mechanics and heat transfer in pool fires, therefore, are of great importance for fire engineers to the understanding of the inherent mechanisms in pool fires. To this end, the first objective of this study is to investigate pool fires and improve their modelling by means of the computational fluid dynamics (CFD) approach, in which coupled descriptions of the controlling mechanisms of heat transfer, turbulence, combustion, and soot production are included. In order to improve the accuracy and applicability of existing turbulence and gas radiation models, advanced models including a four-equation turbulence model, a statistical narrow band (SNB) gas radiation model and a correlated-k (CK) gas radiation model are developed and implemented into the CFD code simulation of fires in enclosures (SOFIE) as the first step of this PhD study. The modified code is applied to three pool fire scenarios, i.e. methane, methanol and ethanol pool fires. Simulation results are numerically analysed, and quantitatively compared between the predictions obtained with different models, as well as with experimental data. Results confirm the improvements in accuracy from advanced models-in terms of temperature predictions, up to 59% relative difference for the four-equation turbulence model and 8.8% for the SNB models are found, though more CPU time is required - the four-equation model requires about 10% more than the two-equation models investigated and the SNB model requires 4.8 times of the traditional weighted-sum-of-grey-gas (WSGG) model. As shown in the Methane fire simulations, the CK model yields results very close to the SNB model, while 2.7 times more time consuming, and thus the CK model is not further studied in this work. After the analysis of pool fires mechanisms and validation of the incorporated models, this research is focused on the numerical and experimental investigation of upward flame spread over solid fuel surfaces. A non-charring pyrolysis model has been developed in SOFIE and is used in this study. A formulation of Quintiere is implemented and employed for predicting the flame spread rate. The pyrolysis model coupled with the filed model is employed to simulate an upward flame spread experiment conducted as part of this work by the author’s research group. In the experiment, temperature, gas velocity, and radiative heat fluxes are measured respectively with thermocouples, particle-image-velocimetry (PIV) and Gunner's radiometers. CFD predictions for surface heat fluxes, the gas velocity, surface temperatures, the flame spread rate, the mass burning rate and the heat release rate are evaluated. The simulation results are generally in good agreement with the experiment. The last objective is to assess the effect of gas radiation models on upward flame spread simulations. Further predictions are attempted for two, more complex flame spread configurations, one involving a 5mX0.6m polymethylmethacrylate (PMMA) wall while the other representing flame spread along the corner. Quantitative and qualitative comparisons are made between the predictions and experimental data. Significant improvement is made by the SNB model over the WSGG model. Although the WSGG approach yields relatively closer results to the SNB approach for the simulations of the steady problem - pool fires, it produces larger discrepancies for those of the transient problem - flame spread, in which the accurate predictions of combined surface heat fluxes are of particular importance.