An identification model for fault detection of reinitializable Discrete-Event Systems

Abstract Recently, Discrete-Event System (DES) identification techniques with the aim of fault detection have been proposed in the literature. In simple words, identification of DES with the aim of fault detection consists in computing a model which is capable of simulating the observed fault-free behavior generated by the system. Then, a fault is detected when a discrepancy between the evolution of the system and the model estimate is observed. Although several identification models for DES have been proposed in the literature, from the authors’ knowledge, they are capable of correctly simulating the fault-free behavior of reinitializable systems, i.e., systems that always returns to its initial state after completing a task, only in some cases. In this paper, we propose a new identification model that simulates the fault-free behavior generated by reinitializable systems. In addition, a didactic mechatronic plant is used to illustrate the proposed method.