Investigation of the structure of a networked system

The information coming from the network topology, which does not depend on the failure and repair probabilities of its elements, is useful in the preliminary RAMS (Reliability, Availability, Maintainability, Safety) analysis and in the Security assessment of critical infrastructures. We propose an approach to investigate the network structure by ranking its elements through Importance and Sensitivity measures. The System performances are referred to User nodes (Partial risks) and to the whole network (Global risk). The First order Differential Importance Measure and a so-called Hybrid measure allow the ranking of elements according to their Importance and to the Sensitivity of the model outputs. The ranking of Edges through the above measures super-imposes the ranking of the (unfaultable) User nodes. The original proposals of the paper concern the use of the Total order Differential Importance Measures in order to consider the "interactions" among the independent input variables of the model (Edges unavailability) and the investigation of the system structure through the estimation of measures in the whole range of values [0;1]. The proposed approach is applied to a simple case study; the simplicity emphasizes the obtained results, allows judging them quantitatively and qualitatively and is not a limitation for real case applications. - Highlights: ► The network performances are specified in terms of Partial and Global Risks. ► The network structure is investigated through Importance and Sensitivity measures. ► The measures are estimated in the whole range of the input variables [0,1]. ► The One At Time measures allows the ranking the unfaultable User nodes. ► The Partial risks correlations and the input variables interactions are considered.