Decentralized diagnosis and diagnosability of a class of hybrid dynamic systems

In this paper, an approach for fault diagnosis of hybrid dynamic systems (HDS), in particular discretely controlled continuous system, is proposed. The goal is to construct a decentralized diagnosis structure, able to diagnose parametric and discrete faults. This approach considers the system as composed of a set of interacted hybrid components (HCs). Each HC is composed of a discrete component (Dc), e.g. on/off switches, with the continuous components (Ccs), e.g. capacitors, whose continuous dynamic behaviour is influenced by the Dc discrete states. A local hybrid diagnosis module, called diagnoser, is associated to each HC in order to diagnose the faults occurring in this HC. In order to take into account the interactions between the different HCs, local diagnosis decisions are merged using a coordinator. The latter issues a final decision about the origin of the fault and identifies its parameters. The advantage of the proposed approach is that local hybrid diagnosers as well as the coordinator are built using local models.