Resumen: In this talk we consider the problem of building Petri net based evaluation scenarios for fault tolerance automation systems and we illustrate a modelling process in which as much information as possible is extracted from a high level description of system entities and their relationships using the Class Diagram notation. Having fixed the domain to that of fault tolerance automation systems also allows to reuse a number of predefined models. The modelling approach is based on modularity and inheritance: modularity has been widely exploited in performance and dependability modelling, inheritance is instead pretty much a "still-to-investigate" topic for this field. In particular, we will describe the role of inheritance in Generalized Stochastic Petri Nets (GSPN) modelling, by considering a representation of the Fault, Error, and Failure (FEF) chain based on hierarchies of classes and corresponding hierarchies of GSPN models. The class of nets of reference is that of Generalized Stochastic Petri Net, and their colored extension Stochastic Well Formed Nets, so that the dependability of the system under study can be evaluated and validated.