Conjunctive and Disjunctive Architectures for Decentralized Prognosis of Failures in Discrete-Event Systems

Kumar and Takai have proposed an architecture for the decentralized prognosis of discrete-event systems (DES), where several local prognosers cooperate to predict failures in a DES. In this paper, we first present the proposition (Kumar and Takai, "Decentralized prognosis of failures in discrete event systems," IEEE Trans. Autom. Control, vol. 55, no. 1, pp. 48-59, Jan. 2010) as a disjunctive architecture, and then we develop a conjunctive architecture which is dual and complementary to the disjunctive one. We also propose a mixed architecture that combines and generalizes the disjunctive and conjunctive architectures. We finally show that our work can be easily extended to predict a failure at least steps before its occurrence, for a given k 1 >=. Note to Practitioners-This study is in the scope of DES, whose behaviors are defined by sequences of events. DES may be representations of multimodes continuous systems, where we are interested by the sequences of modes, and not by the continuous behavior in each mode. For example, during a flight, an aircraft goes through a sequence of flight modes, such as heading, altitude changes, and speed changes. DES may also be artificial systems whose dynamics is intrinsically defined by sequences of events. For example, a telecommunication system where " receive a message" and " send a message" are examples of events. DES are encountered in several areas, such as: avionics, air traffic, road traffic, wireless sensor networks, web services, communications protocols. In the design of DES, we are usually confronted with the risk that a DES violates required properties; such violation will be referred to as fault. In this paper, we study prognosis, which consists in predicting faults. The interest of prognosis is that we are informed of a fault before its occurrence, and hence we can react to avoid the fault or at least reduce its effects. In this paper, we study decentralized prognosis, where several so-called local prognosers cooperate to make an effective prediction. A justification of using decentralized prognosis is that concrete DES are usually decentralized by the fact that they consist of several modules. Another reason is that the decentralized approach can help increase the clarity and reduce the complexity of the prognosis system. The presented study is promising, but prognosis of DES is quite new and necessitates more developments before being commonly used in industry.