abstract
- © 2017This paper is about fault detection for hybrid dynamical systems modeled with Partially Observed Timed Hybrid Petri Nets (POTHPNs). The marking of some continuous places and the firing of some discrete transitions are assumed to be measured on-line. Abrupt faults are considered as unexpected firings of some discrete silent transitions. From the collected measurements, a fault detection approach is proposed that combines residual design based on continuous time observers with a discrete diagnosis method based on the computation of all discrete trajectories that are consistent with the measurements. The method is suitable for the class of hybrid systems that concerns continuous processes driven by discrete controllers.