An approach to model-based diagnosis of hybrid systems

The need for reliability and robustness in present day systems requires that they possess the capability for accommodating faults in the controlled plant. Fault accommodation requires tight integration of online fault detection, isolation, and identification with the system control loop. This paper develops an effective fault diagnosis scheme for plants that mainly exhibit continuous behavior, but include supervisory control making the overall dynamic behavior hybrid in nature. We use hybrid bond graph models to develop diagnosis algorithms that combine hybrid behavior tracking, mode estimation, and qualitative-quantitative reasoning techniques. The effectiveness of the approach is demonstrated with example scenarios generated from a three- and five-tank systems with control valves and sources that can be turned on and off by supervisory control.