| Sommario: | In the paper, we will discuss the main features of Generalized Continuous Time Bayesian Networks (GCTBN) as a dependability formalism, by resorting to a specific case study adapted from the literature, and by discussing modeling choices, analysis results and advantages with respect to other formalisms. We show that the introduction of parts evolving in continuous time together with static parts into the same model is a desirable feature in dependability applications, making possible the modeling of significant dynamic dependencies, without the need of time discretization. From the modeling point of view, in addition to the modeling of functional dependencies, spare dependencies and priority in failures, GTCBNs allow the introduction of general probabilistic dependencies and of conditional dependencies in state change rates of system components. From the analysis point of view, any task ascribable to a posterior probability computation can be implemented, among which the computation of system unreliability, importance measures, system monitoring, prediction and diagnosis. We will show some examples of the above tasks on the case study, by highlighting the practical usefulness of such analyses. We then claim that GCTBN can be a suitable formalism for dependability applications, and future works will concentrates on the modeling of more general dependencies in the framework, as well as on the definition of flexible inference algorithms in addition to existing ones. |
