Modular fault diagnosis using temporized analysis for a class of discrete event systems

Classical approaches for fault diagnosis in discrete event systems can build centralized diagnosers that are able to detect a fault under diagnosability conditions. However, these conditions do not always apply in real applications. In order to decrease the complexity inherent to monolithic approaches and with the aim to provide the human operator with all the possible available information, even when diagnosability conditions do not hold, this paper proposes a special modular decomposition of the monolithic system with controlled modules. Coordination rules based on temporized analysis are provided in order to exploit the information from the local diagnosers without diagnosability conditions.