Quantitative fire risk assessment by combining deterministic fire models with automatic event tree analysis

Abstract

Risk assessment is an integral part of the decision making process within the built environment especially with the adoption of performance-based regulations in place of prescriptive-based regulations. This dissertation examines risk assessment methods, identifying problems within the current methods available. Specifically, this dissertation defines gaps in current quantitative risk assessment where there is a need for event tree analysis, a component of quantitative risk assessment, to be improved. The dissertation also examines agent-based modelling and some of its applications across different industries. Agent based modelling has lead to a development of a novel methodology to automate the process of producing event trees for fire risk assessment. The fire risk methodology provides the risk curve for a set of scenarios by developing a software package combining the use of a) deterministic models i.e. fire zone models, b) probabilistic models i.e. Monte Carlo model, and c) an agent-based model including uncertainty analysis.

The present fire risk methodology has been applied to two case studies. The first case study involves the application of the methodology to assess the benefit of installing two different options of fire safety systems in a warehouse. The second case study is the application of the methodology to a two-storey dwelling house where its flexibility and its ability to perform uncertainty analysis is further examined. These applications of the methodology to the case studies show the methodology flexibility to be applied within different fire risk assessment domains. Finally, recommendations are made to further develop the methodology to include components such as structural risk assessment module and evacuation assessment module.

Date of Award	Mar 2011
Original language	English
Supervisor	Michael Delichatsios (Supervisor) & Jianping Zhang (Supervisor)