Fire Safety Science Digital Archive

Over the past few years a growing number of investigators have used parameter estimation techniques based on genetic algorithms to determine material properties for pyrolysis modeling from small-scale test data. Southwest Research Institute is exploring the use of these techniques to determine material properties needed for computer modeling to predict the behavior of structures in a fire. This paper focuses on the kinetic parameters that are needed to predict dehydration and thermal degradation of gypsum board using a set of Arrhenius equations. Specimens of a type C board were tested in a TGA apparatus in nitrogen at four heating rates; 5, 10, 20 and 60 °C/min. A more conventional approach was used instead of a genetic algorithm to determine the parameters from the TGA data. This approach is implemented on a spreadsheet and is much easier to use than optimization techniques based on genetic algorithms. In many cases it may provide a good, albeit less versatile, alternative. Initially, reactions that occur in temperature ranges that overlap were lumped into a single reaction. At the end of the paper an attempt is made to separate overlapping reactions and estimate the kinetic parameters for each of the overlapping reactions separately.