Pandora : a logic for the qualitative analysis of temporal fault trees

Fault Tree Analysis (FTA) is a valuable systems analysis technique widely used in safety analysis and reliability engineering, but it is not without its faults; in particular, it struggles to analyse systems in which time plays an important role, because fault trees use only Boolean logic and so there is no simple way of representing time or sequences of events in a fault tree. Although there have been attempts to extend FTA to enable analysis of such systems, most have focused on probabilistic analysis and there remains a need for a technique that allows logical analysis of dynamic systems. Pandora is a technique that aims to provide a solution to this problem. It is based around three logical gates capable of representing sequences: the Priority-AND (PAND) gate, the Simultaneous-AND gate (SAND), and the Priority-OR gate (POR). These three "temporal" gates are more expressive, allowing analysts to model sequences as part of a fault tree and thus enabling fault trees to analyse more complex dynamic systems. In addition, Pandora provides a set of logical rules that can be used to reduce fault trees incorporating the three new gates in much the same way that existing Boolean laws can be used to reduce ordinary fault trees. This makes it possible to perform logical analysis of fault trees using Pandora, the results of which provide the analyst with information about the weak points of the system by showing what combinations or sequences of event can cause the system to fail. This thesis presents the evolution of Pandora thus far, explaining the background that led to its inception and the choices made during its development as well as detailed explanations of how Pandora is applied. Pandora has been created with possible automation in mind, so there is also a description of some preliminary algorithms that support Pandora-based FTA. Pandora is thenapplied to a case study to demonstrate how it can function in practice. Finally, the success of Pandora is evaluated by contrasting it with other temporal FTA approaches as well as standard non-dynamic analysis and from this conclusions about the potential benefits of using Pandoraare drawn