Computational methods for diagnosability verification of hybrid systems

Modern industrial systems are real time controlled and supervised by means of automatic computer-based control systems, combining discrete and continuous behaviors, and are best modeled as hybrid systems (HS). In this paper, two methods for the verification of the diagnosability of hybrid systems are proposed. The first method is based on the construction of a diagnoser automaton, that can also be straightforwardly used for online diagnosis, and the second method is based on a verifier automaton that, although cannot be used for online diagnosis, can be constructed in polynomial time, leading to a smaller computational complexity for the verification of the diagnosability of HS than the method using diagnoser automata. The main idea of the second method is to build a verifier of the underlying discrete-event system (DES), taking into account the distinguishability of the system modes based on the continuous state models of the HS.