Hybrid Bond Graphs for diagnosis of three cells converter

Bond Graph as a unified and multidisciplinary tool, stay very interesting regarding the interaction between different continuous modes. These industrial systems, having more than one continuous state, need hybrid tehniques for their model based diagnosis. Traditional Bond Graph algorithms for continuous diagnosis use a single model that does not accommodate the transition between modes. The physical system is converter containing three pairs of switches (cells), leading to eight modes. The aim of the paper is to show how the hybrid bond graph (HBG) can be used not only for dynamic modelling of all modes but also the fault Detection and isolation. Based on structural and causal properties of the HBG, the fault indicator named Global Analytical Redundancy Relation (GARR) is generated systematically for all modes. The Fault Signature Matrix is then deduced from candidate GARRs for decision procedure step. The first part of this work considered to obtain a hybrid model, who allows to present different modes in one structure. The changes in the causalities only on level of the switching junctions when the mode changes occur was observed. Our work is motivated by the requirements of the fault accommodation tasks, where diagnosis is performed online for the hybrid system. This model of the residuals and the related fault signature matrix can be implemented in real time applications to provide a decision support system including both the fault detection and isolation capabilities.