Reliable fault prediction of decentralized stochastic discrete‐event systems under dynamic observations

Abstract Most studies on fault prediction of decentralized discrete‐event systems (DESs) are based on the assumptions that the observability of events remains constant and any local information is never lost. Nevertheless, in numerous practical applications, the observability of events is closely associated with the changing states of systems, and some information may not be available in time. In this paper, the reliable fault prediction issue for decentralized stochastic DESs (SDESs) under dynamic observations is investigated in the face of the changing observability of events and the potential unavailability of some local information. The main contributions are as follows. First, the notion of ‐reliable ‐step stochastic copredictability (or ‐copredictability) of SDESs is formalized, in which an SDES with ‐copredictability distributed at local sites is required that all faults can be predicted in the sense of probability at least steps in advance under dynamic observations, even if local sites are invalid. Second, an automaton named reliable ‐coverifier is formally constructed, and a necessary and sufficient condition for testing the ‐copredictability is presented based on the constructive reliable ‐coverifier. Finally, a polynomial‐time algorithm is proposed to verify the ‐copredictability of SDESs under dynamic observations.