On-line robust graphical diagnoser for hybrid dynamical systems

This paper proposes a generic tool named Hybrid Bond Graph (HBG) driven by an automaton. Along with the conventional Bond Graph (BG) properties, the proposed approach offers some very valuable advantages in modelling the Hybrid Dynamical Systems (HDS). It allows having one global model of all the distinct dynamics called dynamical operating modes generated due to the hybrid aspect of the system. Compared with other classical approaches such as the hybrid automata, the Operating Mode Management (OMM) is much simpler and more effective thanks to the separation between the continuous dynamics associated with the BG continuous state and the discrete state governed by a separate classical automaton. From the diagnosis point of view, the approach allows to obtain one global graphical robust diagnoser simply derived from the model using the Linear Fractional Transformation (LFT). Using the BG causal properties along with the correspondences between the system components and the BG elements, the Event-Driven Hybrid Bond Graph (EDHBG) also helps to associate between the faults and their suspected components. The innovative interest in this work is that it allows for a non-expert user to perform the modelling, the on-line robust diagnosis and the OMM, without expressing the algebraic equations of the system and regardless of its hybrid aspect. A pedagogic experimental set-up serving as a comprehensive example is implemented on 20Sim1 to demonstrate the potential of this technique.