Search

This paper addresses the problem of cyber-attack detection in a discrete event system by proposing a novel model. The model utilizes graph convolutional networks to extract spatial features from event sequences. Subsequently, it employs gated recurrent units to re-extract spatio-temporal features from these spatial features. The obtained spatio-temporal features are then fed into an attention model. This approach enables the model to learn the importance of different event sequences, ensuring that it is sufficiently general for identifying cyber-attacks, obviating the need to specify attack types. Compared with traditional methods that rely on synchronous product computations to synthesize diagnosers, our deep learning-based model circumvents state explosion problems. Our method facilitates real-time and efficient cyber-attack detection, eliminating the necessity to specifically identify system states or distinguish attack types, thereby significantly simplifying the diagnostic process. Additionally, we set an adjustable probability threshold to determine whether an event sequence has been compromised, allowing for customization to meet diverse requirements. Experimental results demonstrate that the proposed method performs well in cyber-attack detection, achieving over 99.9 % accuracy at a 1 % threshold and a weighted F1-score of 0.8126, validating its superior performance. [ABSTRACT FROM AUTHOR]

This paper considers the verification of decentralized fault pattern diagnosability for discrete event systems, where the pattern is modeled as a finite automaton whose accepted language is the objective to be diagnosed. We introduce a notion of codiagnosability to formalize the decentralized fault pattern diagnosability, which requires the pattern to be detected by one of the external local observers within a bounded delay. To this end, a structure, namely a verifier, is proposed to verify the codiagnosability of the system and the fault pattern. By studying an indeterminate cycle of the verifier, sufficient and necessary conditions are provided to test the codiagnosability. It is shown that the proposed method requires polynomial time at most. In addition, we present an approach to extend the proposed verifier structure so that it can be applied to centralized cases. [ABSTRACT FROM AUTHOR]

This paper investigated the verification of detectability for discrete event systems based on a class of partially observed unbounded Petri nets. In an unbounded net system, all transitions and partial places are assumed to be unobservable. The system administrator can only observe a few observable places, i.e., the number of tokens at these places can be observed, allowing for the estimation of current and subsequent states. The concepts of quasi-observable transitions, truly unobservable transitions, and partial markings are used to construct a basis coverability graph. According to this graph, four sufficient and necessary conditions of detectability are proposed. Correspondingly, a specific example is proposed to prove that the detectability can be verified in the unbounded net system. Furthermore, based on the conclusion of detectability, the system's ability to detect critical states was explored by using the basis coverability graph, called C-detectability. Two real-world examples are proposed to show that the detectability of discrete event systems has not only pioneered new research methods, but also demonstrated that the real conditions faced by this method are more general, and it has overcome the limitations of relying only on the ideal conditions of bounded systems for verification. [ABSTRACT FROM AUTHOR]

This paper concentrates on robust deadlock control problems for an automated manufacturing system with uncontrollable, unobservable events and resource failures, which is modeled by a subclass of Petri nets. A recovery subnet is added to the holder of each unreliable resource to model resource failure and recovery. In order to prevent a strict minimal siphon in a Petri net model from being emptied, an extended constraint set is constructed based on the complementary set of a strict minimal siphon to obtain a robust liveness constraint. Due to the uncontrollability and unobservability of the transitions, an automated manufacturing system cannot enforce inadmissible robust liveness constraints. It is necessary to decide the admissibility of robust liveness constraints and transform the inadmissible constraints into admissible ones. To this end, a monitor is designed for each admissible robust liveness constraint such that it can be enforced. Finally, an iterative algorithm is developed to perform the above steps and a robust liveness controlled system is obtained. The feasibility of the control strategy is demonstrated by examples. [ABSTRACT FROM AUTHOR]

Recently, researchers have proposed a novel and computationally efficient method to design optimal control places and an iteration approach that computes the reachability graph once to obtain a maximally permissive, if any, liveness-enforcing supervisor of flexible manufacturing systems (FMS). The approach solves the set of integer linear inequalities to compute control places. If, given a Petri net model, no solution exists, the optimal control place does not exist for the Petri net model. We discover that a solution always exists for systems of simple sequential processes with resources (S3PR), but not for the case of FMS modelled by generalised Petri nets (GPN). We propose a theory to prove that there are no good states that will be forbidden by the control policy for S3PR, in which live and dead states cannot have the same weighted sum of tokens in the complimentary set of a siphon. For a system of simple sequential processes with general resource requirements (S3PGR2) modelled by GPN, we find the reason why the integer linear programming (ILP) may not have solutions, which is consistent with the fact that optimal supervisor synthesis for GPN remains unknown. We show that live and dead states may have the same weighted sum of tokens in the complimentary set of a siphon in a GPN. These theoretical results are verified by case studies. [ABSTRACT FROM AUTHOR]

Structural analysis is one of the most important and efficient methods to investigate the behaviour of Petri nets. Liveness is a significant behavioural property of Petri nets. Siphons, as structural objects of a Petri net, are closely related to its liveness. Many deadlock control policies for flexible manufacturing systems (FMS) modelled by Petri nets are implemented via siphon control. Most of the existing methods design liveness-enforcing supervisors by adding control places for siphons based on their controllability conditions. To compute a liveness-enforcing supervisor with as much as permissive behaviour, it is both theoretically and practically significant to find an exact controllability condition for siphons. However, the existing conditions, max, max′, and max″-controllability of siphons are all overly restrictive and generally sufficient only. This paper develops a new condition called max*-controllability of the siphons in generalised systems of simple sequential processes with resources (GS3PR), which are a net subclass that can model many real-world automated manufacturing systems. We show that a GS3PR is live if all its strict minimal siphons (SMS) are max*-controlled. Compared with the existing conditions, i.e., max-, max′-, and max″-controllability of siphons, max*-controllability of the SMS is not only sufficient but also necessary. An example is used to illustrate the proposed method. [ABSTRACT FROM AUTHOR]