ABSTRACT
Fault prognosis has become a major scope for complex and interconnected systems. Such significant events as fault events can cause partial or total stop of attempted functionalities. Prevention failure events are an issue to preserve performance, availability and safety of both operators and equipment. The aim of prognosis is to prevent fault events before their occurrence. Fault/repair management refers to event control, and so it is relevant to the domain of Discrete Events Systems (DES), for which stochastic finite state automaton and Petri Nets (PN) have been used to prognosticate fault state. They are based on predictions of fault event at least m-steps in advance.
The proposal is based on the time notion, which is crucial for fault prognosis. Indeed one can give the remaining time before the occurrence of fault event. The goal is to prevent the occurrence of a fault event at τ – timeunits in advance. This approach is based on labeled and T-temporal Petri Net, which has the advantage of a formal character for the assessment of properties and of sufficiently generic in order to apprehend a high level of complexity.
