The purpose of this research is to develop a dataset for some of the most important fire characteristics, namely CO2 yield, CO yield, HCN yield, soot yield and heat of combustion for probabilistic analysis and modelling.

Raw data in time series are required to mechanically reduce experimental data into yields (kg/kg) and effective heats of combustion (MJ/kg), which are expressions for the amount of products generated per unit mass of fuel. Mass loss rate thresholds were applied to all tests to define the beginning and end of tests. These species yields and heat of combustions were then grouped by material compositions and fitted with distribution functions to produce distributions curves.

As fire species productions and heat of combustions are dependent on the fire conditions as it develops, different yields are expected at different fire stages. These have been identified as the growth (G), transition (T), and smouldering (S) stages in this research. These values are also compared against, and are generally in agreement with, other research data. Nonetheless, some discrepancies have occurred and require further information to ascertain the material characteristics and combustion conditions.

In conclusion, design recommendations for these fire characteristics have been made for several material groupings and verified against other research results. Certain physical and chemical limitations exist for combustions and have not been reflected in the fitted distribution, including stoichiometric yields and unlimited air yields. As such, species yields and heat of combustions beyond these values should not be considered in fire engineering design and analysis.

Research results on HCN including all required data parameters for yield conversions were difficult to obtain and require further research efforts. Tube furnace results were initially investigated. Unfortunately, without a continuous mass record, has proved to be challenging in producing reliable mass loss rate profiles for yield conversions. A semi-automated data reduction application UCFIRE was also used. However, certain technical difficulties were encountered and require modifications to broaden its applicability.