Issues in modeling, supervision, and fault detection for automated RTP systems

Event detectability is the corner stone of constructing reliable diagnosers for Petri net (PN) models. In some PN models, the transition firing sequences are not detectable based on their outputs and structural information only. This paper introduces a novel diagnoser to overcome such problems. The developed diagnoser depends not only on the output and the structural information of the PN models, but also on its associated marking vector. The main advantage of the proposed diagnoser is to evaluate the faulty states without checking event detectability and coverability trees of the discrete event systems. Issues for rapid thermal processing (RTP) modeling and supervision are also addressed in this paper. Early fault detection will minimize cost and processing time of the RTP systems, accordingly, this paper also develops a unified PN-based frame work for fault detection and isolation in these systems.