Integrated Diagnosis and Control for Hybrid Dynamic Systems

In this paper we present an approach for combined fault diagnosis and reconfigurable control structure for a general class of hybrid systems. In this approach a plant is modeled using an extended version of bond graphs, namely hybrid bond graph, where discrete mode transitions are represented as binary switch junctions. A hybrid observer has been developed that uses this model to track the system behavior within and across modes. Two complementary approaches are used for fault detection and isolation. The first diagnoser is based on hybrid models and uses the hybrid observer, qualitative reasoning techniques, and real-time parameter estimation. The other diagnoser is based an abstracted discrete event model of the system that shows the causal and temporal relation between failure modes and corresponding abstract observations. To accommodate detected failure a new controller can be selected for a previously developed controller library based on the current condition. Control reconfiguration can be also achieved through online control techniques.