Mixed approach for fault diagnosis and fault location of hybrid systems

Abstract Today, the fault diagnosis remains a very important task due to its essential role in improving the dependability and also in minimizing the harmful consequences which can be catastrophic for equipment and human safety. This article proposes a diagnostic approach for hybrid systems through the use of Bond Graph and observers well known in continuous field, coupled to timed automata well used in the field of discrete events systems. Indeed, for a complex system, the Bond Graph allows modeling of multi-energy physical processes. The observer-based approach is used to generate dynamic models (in normal and in failing mode). The fault detection algorithm consists mainly in comparing the real measures of the system with measures estimated by an observer. Thus, the fault isolation procedure consists in analyzing the fault signature matrix. Timed automata allow to take into account the dynamic evolution of the system and the failures propagation. To refine the failure location, a diagnoser based on timed automaton is built. The proposed approach is validated by simulation tests in a hydraulic system with two tanks.