Work

AI-Based Performance Classification of Multi-Level System-of-Systems via Hypergraph Modeling

Year: 2024

Type: article

Abstract: Accurately classifying the performance of System-of-Systems (SoS) is crucial for maintaining system reliability and resilience. Without clear classifications, engineers may struggle to assess the severity of malfunctions and implement timely corrective actions. To this end, this paper proposes an evaluation framework to assess the performance of SoS, so as to classify them. The objectives of the framework are to inform users of the SoS status and to determine the sufficient or necessary reconfigurations to recover from SoS degradation. A hybrid graph- and hypergraph-based model is proposed for multi-level SoS, structured around management decomposition. The proposed model describes the multi-way organization and communication among diverse component systems (CSs). Subsequently, a performance evaluation framework is proposed to assess the performance degradation of SoS under faults, perturbations and communication issues. One of the key challenges in classifying SoS performance lies in handling raw data collected from diverse CSs across multiple levels. The varying magnitudes and dimensions of this data can lead to low generalization and potential overfitting during the classification process. Various performance features are extracted from different aspects, such as the response of physical component systems (PCSs), the task-oriented attributes of managerial component systems (MCSs), the inter/intra-layer communication among CSs of multiple levels and the top-down management efficiency. These features provide a comprehensive view of the system’s performance. Using the proposed feature construction and AI-based classifiers, the SoS performance is classified into five status types. The classification results indicate the degree of malfunction and the urgency or necessity of a resilient response, which help engineers to identify the level of degradation in the SoS and decide on appropriate actions to restore its performance. Finally, a simulation of a three-level SoS with three MCSs and eight PCSs is conducted to validate the effectiveness of the feature construction for SoS performance. Various algorithms are utilized in the simulation, demonstrating high accuracy in both the training and test sets.

Source: IEEE Access

DOI: 10.1109/access.2024.3521986

Authors: Jun Jiang, Abdeslem Smahi, Yiwen Chen, Othman Lakhal, Rochdi Merzouki

Institutions: Centre National de la Recherche Scientifique, Université de Lille, Centre de Recherche en Informatique, Signal et Automatique de Lille, Laboratoire d'Automatique, de Mécanique et d'Informatique Industrielles et Humaines, Université Polytechnique Hauts-de-France

Language: English

Topic: Systems Engineering Methodologies and Applications

Subfield: Control and Systems Engineering

Field: Engineering

Domain: Physical Sciences

SDG: Industry, innovation and infrastructure

Open Access status: gold

Estimated APC paid: $1,995