Event-Based Automaton Model for identification of discrete-event systems for fault detection

Fault diagnosis is a crucial task to guarantee reliability, and reduce losses and production cost in industrial systems. In the traditional techniques for designing a fault diagnoser, it is necessary to obtain the complete model of the system, including its post-fault behavior. However, in general, industrial systems are large and composed of several subsystems which makes their modeling a laborious and time-consuming task. In addition, only predefined faults can be detected using the traditional fault diagnosis approach. In order to circumvent these problems, black-box identification techniques have been proposed in the literature to obtain an automaton that models the fault-free behavior of the system from the observation of the input and output signals of the system controller. Then, this model is used in a fault detection algorithm, and, after detection, the fault is isolated offline based on a comparison between the identified model and the sequence of observed signals. In all these approaches, it is assumed that the system has the same initial status of inputs and outputs of the controller. In practice, however, the system may start the execution of tasks with different input and output controller signal values. In this work, we present a new identification model, called Event-Based Automaton Model (EBAM). Differently from the other models proposed in the literature, the EBAM can be used to represent the fault-free system behavior when the tasks executed by the system start with different controller input and output values. A practical example, consisting of a plant simulated by using a 3D simulation software controlled by a Programmable Logic Controller, is used to illustrate the identification method and to show the efficiency of the fault detection algorithm based on the identified EBAM.