Observer-based diagnosis modeling using stochastic activity networks for the dependability assessment purpose

Nowadays, architectures with increasing complexity are designed to meet the productivity and safety requirements in industrial systems. These architectures include subsystems and procedures that allow fault detection, diagnosis and accommodation, completed by some system´s maintenance and reconfiguration policies. These components work together and interact with each other to improve the system´s dependability. Thus, it is more judicious to consider them explicitly in the system´s dependability analysis models. This paper proposes a modular and systematic approach to model diagnosis procedure based on Luenberger observer using stochastic activity networks (SANs). Combined with Monte Carlo simulation, this approach allows studying diagnosis performances and its impact on some dependability factors like the availability.