This thesis covers a variety of analytical approaches that validate the use of the EvacuatioNZ model on high-rise building analysis. Through performing a number of sensitivity analyses, several model deficiencies as well as functional limitations were improved upon and part of the model developments are continued based on the previous research done by two Master's students at the University of Canterbury. In this thesis, data from three evacuations were considered for different validating aspects. These evacuations were, a hypothetical 21-storey hotel building located in the United States of America, which was previously simulated using Simulex and EXIT89; a trial evacuation that was carried out in a 13-storey office building located in Canada; and a fire drill conducted at a 21-storey office building located in Australia. Overall, the results indicated that the EvacuatioNZ is able to produce reasonable predictions of the total evacuation time regardless of the number of floors involved. The component testing also showed satisfactory outcomes regarding the involvement of disabled occupants, complexity of node configurations, and different pre-movement time distributions. However, the current model still has a number of limitations that need to be verified and tested. These include the preferred route function and the connection problem for long stairs. Further research should also be carried out on the use of the Evacuation model on other types of building structures so as to increase the confidence level of utilizing the EvacuatioNZ model for general applications.

L’objectif de cette thèse est de proposer un modèle comportemental de l’être humain pour la simulation d’évacuation de bâtiment en cas d’incendie et de l’intégrer dans un outil de simulation d’évacuation de bâtiment. Le modèle proposé représente une approche individu-centré du comportement et répond aux axiomes de la Rationalité Limitée énoncés par Herbert Simon grâce à une conception hiérarchique des moyens cognitifs des agents simulés. L’implémentation du modèle du comportement présente la particularité d’être intégralement réalisée en GPU (via OpenGL 2.0). Ainsi la fréquence du moteur de comportement est très proche de celle du simulateur et permet une adaptation quasi temps réel des comportements des agents à un changement de perception de leur environnement. Le logiciel d’évacuation développé permet :– l’importation de scénario d’incendie de bâtiment simulé grâce au logiciel FDS,– la configuration des archétypes de comportement des agents évacuant, notamment :– la description des hypothèses sur le monde (connaissance individuelle),– la configuration des comportements des agents (moyens cognitifs individuels).– le positionnement des agents dans le bâtiment,– la simulation de l’évacuation,– l’enregistrement et le play-back d’un scénario d’évacuation.Les contraintes (feu, fumée, obstacles, autres agents) captées par un agent sont interprétées par ce dernier en fonction de son archétype de comportement afin de déterminer si sa stratégie d’évacuation doit être remise en cause. / The main objective of this thesis is to propose a behavioral model of the human being in presence of several constraints and to integrate it into a simulation tool for building egress. The proposed model represents an individual-based approach of behavior modelisation and implement axioms of the bounded rationality set by Herbert Simon providing two key features :– individual prioritization and parametisation of cognitivemeans,– individual perception and knowledgemanagement.The proposed software has the particularity to run entirely on GPU via OpenGL 2.0. Thus the frequency of the behavior engine is very near to that of the simulator and allow adaptation of near real-time behavior of agents in a changing perception of their environment.