On the Effect of Dynamic Event Observations in Distributed Fault Prognosis of Discrete-Event Systems

Abstract

In the conventional framework for distributed fault prognosis of discrete-event systems (DESs), it is assumed that observable events are always observed [such case is called static event observations (SEOs)]. However, the assumption may not hold in many DESs such as sensor networks. This article introduces the concept of distributed fault prognosis with dynamic event observations (DEOs), in which observable events are not always observed. Communication models and extended models are constructed, based on which, for each local prognoser, an extended dynamic observation mask with two forms is constructed to capture its aggregate information. In order to verify prognosability subject to DEOs, one algorithm whose complexity is polynomial in the number of states but exponential in the number of local prognosers is presented. Furthermore, one significant condition for prognosability subject to DEOs is derived. Finally, the obtained results are applied to an Alipay online trading system and an Industry 4.0 manufacturing system.