Monitoring of a class of timed discrete events systems

This paper extends the notion of residuals for fault detection, well known in the continuous system to the timed discrete events systems. The aim is to design the fault indicators, two problems are considered. Firstly, the dynamics of clocks which monitor the system must be sensible to the behavior change resulting from a fault. Secondly, we must specify the inequality clocks constraints for which the system is normal as long as they are satisfied. We introduce the notion of acceptable system behavior modeled by stopwatch automata. This behavior is supervised by two clocks for each interruptible task of the system. The time sub-spaces in the stopwatch automaton locations delimit exactly the range of acceptable behavior. They are synthesized using an algorithm based on the reachability analysis techniques of stopwatch automata. One of the main results of designing this time space is the detection the system faults as early as possible.