The Application of a Genetic Algorithm to Estimate Material Properties for Fire Modeling from Bench-Scale Fire Test Data
A methodology based on an automated optimization technique that uses a genetic algorithm (GA) is developed to estimate the material properties needed for CFD-based fire growth modeling from bench-scale fire test data. The proposed methodology involves simulating a bench-scale fire test with a theoretical model, and using a GA to locate a set of model parameters (material properties) that provide optimal agreement between the model predictions and the experimental data. Specifically, a genetic algorithm based on the processes of natural selection and mutation is developed and integrated with the NIST FDS v4.0 pyrolysis model for thick solid fuels. The combined genetic algorithm/pyrolysis model is used with Cone Calorimeter data for surface temperature and mass loss rate histories to estimate the material properties of two charring materials (redwood and red oak) and one thermoplastic material (polypropylene). This is done by finding the parameter sets that provide near-optimal agreement between the model predictions and experimental data given the constraints imposed by the underlying physical model and the accuracy with which the boundary and initial conditions can be specified. The methodology is demonstrated here with the FDS pyrolysis model and Cone Calorimeter data, but it is general and can be used with several existing fire tests and almost any pyrolysis model. Although the proposed methodology is intended for use in CFD-based prediction of large-scale fire development, such calculations are not performed here and are recommended for future work.