Automatic fault tree building and importance measures ranking aggregation for decision making in train design

The increasing complexity of physical systems involves an increasing effort for reliability modeling and analysis. Yet, reliability analyses are asked to be more quick and confident, so that they can timely support system design, as well as to give answers about the relationship among system and its components failures. The issue is arisen during my stage in Alstom Ferroviaria SPA, where reliability analyses on trains’ traction systems are needed to respect the reliability parameters fixed by low and customers: lean system modeling would allow the designer to improve system reliability during design phase, whereas focused analyses on system components lead to a reduction in needed improvements and component costs. In this context, a novel approach for system modelling has been proposed, together with an innovative method for decision making in reliability design. The first part of this thesis proposes a method to automatically build Fault Trees (FT), which combines a hierarchical representation of the analyzed system, with a functional analysis of its components. The method is based on a Goal Tree/Success Tree-Master Logic Diagram (GTST-MLD) approach, and is applied to a simple electrical circuit to show its features of modularity and iterativity. The second part of this thesis proposes an innovative method for comparing the impact of components failures on the system unavailability, joining the different information obtained from different measurements. In particular, the method processes the Importance Measure (IMs) rankings of the components, building an absolute ranking aggregating the existing ones. Six procedures for ranking aggregation are presented, together with a procedure for make a ranking coherent with the Condorcet Criterion (CC) about majority principle, and then the results are compared on a case study present in literature.