| The implementation of performance-based codes and risk-informed regulation faces a number of barriers in the field of fire protection. In this dissertation some of those barriers will be discussed with the purpose of presenting computational methods designed to reduce the impact of the barriers in the decision-making process. The computational methods are presented in the framework of a comprehensive probabilistic fire risk assessment methodology.; In the probabilistic fire risk assessment methodology, the impact of a fire in a specific location is estimated in terms of a number of indices based on the time of hazard development versus the time to hazard mitigation. Comparing the risk index estimation of each enclosure in a building or plant, a ranking of locations can be generated and different fire protection alternatives can be evaluated.; The fire risk assessment approach is based on probabilistic fire models developed from traditional deterministic calculations. In order to develop the probabilistic tools, a complete fire two-zone model, called FireMD, the traditional DETACT model, and a target heating model are described. Next, the concepts of model and parameter uncertainty are introduced in the fire analysis in order to estimate probability distributions of hazardous conditions generated by a fire accident. Specifically, a probabilistic two-zone fire model, a probabilistic detector activation model, a probabilistic suppression model, and a probabilistic target-heating model form the foundation of the proposed fire risk assessment methodology. The main assumptions of the models, as well as the methods to estimated uncertain input parameters and the algorithms used to solve them are discussed and characterized. |
