Railway systems are known as Safety Critical Systems (SCSs). In this kind of system, safety measures derived from the dysfunctional analysis used to be expressed in an informal way. This latter has several gaps in the context of the one going numeric transition: in the early phase of SCSs design, there is a need to link these safety measures to main safety goals. A first step provides a knowledge structure, where the considered knowledge is composed by a set of data and a set of engineering rules. These rules, including safety measures, correspond to a knowhow built through information sharing between actors during previous industrial system life-cycle. From this structured knowledge, models using main concepts can be designed. As concepts come from ontology, the system models are naturally high-level ones and directly linked to the source needs. Indeed, source needs are expressed on the basis of the structuring concepts of the ontology. Obviously, obtained models are abstractions of the real systems. Model based system engineering (MBSE) allows a systematic reasoning and tooled conformance checking and it is possible to assign a meaning to measured data during the whole life cycle of the railway system. A fundamental assumption is the validity of models used during this life cycle. As an abstraction is a partial point of view, the relevance of this partiality must be monitored during the system life cycle in order to avoid ambiguous interpretations. In this paper, the semantic interoperability is tackled to avoid ambiguities and to ensure the railway digital continuity.