A Black-Box Identification Method for Automated Discrete-Event Systems

Abstract

This paper deals with the identification of discrete event manufacturing systems that are automated using a programmable logic controller (PLC). The behavior of the closed loop system (PLC and Plant) is observed during its operation and is represented by a single long sequence of observed input/output (I/O) signals vectors. The proposed method follows a black-box and passive identification approach that allows addressing large and complex industrial DES and yields compact and expressive interpreted Petri net (IPN) models. It consists of two complementary stages; the first one obtains, from the I/O sequence, the reactive part of the model composed by observable places and transitions. The I/O sequence is also mapped into a sequence of the created transitions, from which the second stage builds the non observable part of the model including places that ensure the reproduction of the observed input output sequence. This method, based on polynomial-time algorithms on the size of the input data, has been implemented as a software tool that generates and draws the IPN model; it has been tested with input/output sequences obtained from real systems in operation. The tool is described and its application is illustrated through a case study. Note to practitioners-Automated modeling of controlled discrete manufacturing systems can be achieved by efficient identification algorithms that cope with large and complex plants performing concurrent and repetitive tasks a priori unknown. The black-box identification procedure processes an input/output sequence recorded during the system functioning for a long period of time, and then yields a comprehensive model of the closed-loop controlled system; this model approximates closely the actual behavior of the compound system controllerplant. A tool based on identification algorithms constitutes an excellent resource for computer-aided reverse engineering of controlled manufacturing systems. The method proposed herein allows processing sequences composed by thousands of I/O vectors in few seconds.

An Improved Industrial Control System Device Logs Processing Method for Process-Based Anomaly Detection

“…One may argue that ICS should be simply re-programmed to record events. However, it increases the initial cost of implementation limits its adaptability of many theoretical approaches in industrial environment [9]. On the other hand relying solely on expert's knowledge is prone to error.…”

“…[5]. While, existing methods [7], [9] for the conversion of ICS device logs (I/O signal vectors) into events are limited to discrete event systems as discussed earlier.…”

“…Device logs contain record of device status with timestamps [8], as shown in Table. I. While, existing ICS device logs processing methods such as discussed in [7], [9] are limited to discrete event systems (DES) i.e., inputs and outputs of a system are logical (0/1 or on/off). However, most of the real-time automated systems are hybrid.…”

Fault diagnosis based on identified discrete-event models

“…In the case of DES, the data observed during the system functioning are sequences of binary events, and identified models are abstract machines, like Petri nets or finite automata. Several works in the literature propose identification methods based on automata or Petri nets for different purposes [18,4,13,12,14,2,15]. The majority of these works address the problem of identifying Petri net models that are not suitable for fault diagnosis.…”

A relative identification method for reactive systems

“…Roth et al (2010) and Schneider et al (2012), for instance, propose methods to build automata that are used later on for diagnosis purposes. Meda-Campaña and López-Mellado (2005), Giua and Seatzu (2005), Dotoli et al (2011), Estrada-Vargas et al (2015), Saives et al (2015), to name a few, focus on identification based on Petri nets. The latter two references place their works in the context of reverse engineering, i.e.…”