| Abstract: | This article presents a calculation-based methodology to determine the dominant event class in each of the phases of disasters being analysed, and to address the question of whether different disasters have similarities at crucial times in each phase of the disaster. Our approach is based on event network analysis. Disasters can be modelled using block diagrams and multiphase process trees. We propose trees in this article can be used as a tool for modelling phases of a disaster. The starting point for developing these models was fault tree analysis used for modelling the reliability structure of complex systems. This study demonstrates the possibility of using dual fault trees to describe the process as opposed to the structure. In our analyses, we examined four major disasters of production platforms that occurred in the last 50 years: Ixtoc I, Piper Alpha, Petrobras 36 and Deep Water Horizon. The course of each of these disasters has been described, the basic events of these disasters have been isolated, and assigned to event classes. The hierarchical importance of events was determined using the Birnbaum reliability measure, Birnbaum structural measure, Fussell-Vesely measure, criticality measure and improvement potential. For each phase of the analysed disasters, event importance is ranked, and the most important events that contributed to the phase are identified. General principles on the analysed disasters and the methodology used are also discussed. |
