Flame properties and fire spread on cylindrical fuel beds

This thesis investigates the properties of flame and the propagation of fire on cylindrical fuel beds composed of several types of wood. The majority of experiments were conducted on Tasmanian Oak, although several other fuel types were considered as well. Empirical and physical models of flame propagation are presented. The shape of a flame envelope around a fully ignited horizontal wooden brand was studied experimentally. The change in the flame shape during the period of burning, the time over which burning occurred and the burn efficiency were modelled empirically. Correlations relating these quantities to the geometry of the fuel bed and the rate of the fuel mass loss were performed. These are compared with results previously reported in the literature. A further set of experiments was then performed on flame propagation along the cylindrical brand, for several wood types. The speed of the flame front, the flame height and width, the rate of fuel consumption and the efficiency were all measured, and their behaviour was correlated to the angle of the fuel bed to the horizontal, and to the diameter of the fuel cylinder and the initial moisture content. Some additional experiments were also performed on parallel horizontal cylindrical fuel brands, and the separation between them was found to have a significant effect on flame height. Some mechanisms for this effect are proposed. The emissivity of small flames was measured for different fuel types, and for different geometrical arrangements of the fuel load. Experiments were performed using Tasmanian oak, nogal, ramin, white berch, pine, balsa, sapelli and pine needles as fuels. A mathematical model is presented for describing flame propagation along a horizontal fuel bed. This results in a system of two non-linear coupled partial differential equations involving temperature and available fuel, as functions of time and the distance along the fuel bed. A detailed comparison with the experimental observations is undertaken. Some of the experimentally derived rules are used in the model, and some model parameters have been determined from the experimental observations. The model is capable of predicting the observed behaviour.