Fault Prognosis with Supervisory Control of Discrete-Event Systems under Non-deterministic Observations

Abstract

Fault prediction seeks to predict the future occurrences of faults in order to implement effective control measures. In the context of non-deterministic observations, this article investigates the fault prognosis of discrete-event systems using supervisory control. The concept of active prognosability is formalized to characterize a discrete-event system’s ability to predict and prevent a fault under non-deterministic observations. A transformed automaton is created from the initial system to test the property by converting non-deterministic observations into deterministic observations. Based on this, a verifier is built, and the necessary and sufficient conditions to ensure such property are deduced. Furthermore, an algorithm for verifying active prognosability is proposed, the complexity of which is polynomial in the number of system states and events. Also, the conditions for the controllable sub-language’s existence are established, and the matching supervisor is synthesized, allowing the system to run faultlessly and properly. Our results expand on previous research with regard to safe controllability through diagnosis and prognosis.

DOI: 10.61416/ceai.v26i2.8984