Functional-architectural diagnosability analysis of embedded architecture

Diagnosability analysis of functions offers now a serious complement to knowledge-based methods of diagnosis, such as FMEA (Failure Mode and Effects Analysis) and fault tree analysis. State of the art of diagnosability analysis focus on what we call "functional diagnosability", where the hardware architecture of the system and its constraints are not directly considered. This paper contributes to the analysis of the functions-architecture interaction impact on the diagnosability of an embedded system, especially automotive systems. The approach we developed can be integrated into the design cycle. It has two important phases; first, the diagnosability analysis of discrete event systems, then the verification of a property set that we have defined and called the “diagnosability functional-architectural properties”. Properties verification is done in two stages: check the description of the architecture, described in AADL, and check the functions-architecture interaction, modeled in SystemC-Simulink. The validation process is applied on a real automotive experimental embedded platform based on several Electronic Control Units. Finally, we have developed through this paper a novel methodology for the analysis of diagnosability that takes into account the constraints of the hardware architecture of the system.