Nowadays, industrial systems are getting more and more complicated, integrating various technologies. Their designs involve many different engineering fields to maximize rentability and to offer the most up to date functionalities and services. The Model-Based System Engineering (MBSE) approach address specifically these issues by allowing a more global way of designing a complex system from many points of view. However, MBSE does not ensure dependability. That is the reason why, in this thesis, our work aims to connect the MBSE approach with dependability analysis. The SysML modeling language is used to reify the results of system engineering activities, to obtain a model of the system. This model is then computed to extract data that will help dependability analysis. These automatic processes of extraction and redaction are part of the MéDISIS methodology which is defined in this thesis. Two aspect of the MéDISIS methodology are discussed: the generation of functional FMEA and the use of the FIDES methodology in association with SysML to evaluate the failure rate of COTS. This work and the whole MéDISIS methodology are applied in the industrial context of the LEA project. This project, financed by MBDA, consists in designing an hypersonic vehicle.
