Graph Neural Networks for AI-Driven Fault Diagnosis in Embedded Systems

Abstract

Embedded systems are critical components in modern computing, powering applications in industries such as automotive, healthcare, aerospace, and industrial automation. However, diagnosing faults in these systems remains a challenging task due to their constrained resources, real-time processing requirements, and complex dependencies. Traditional fault diagnosis methods often struggle with scalability and adaptability in dynamic environments. This research explores the application of Graph Neural Networks (GNNs) for AI-driven fault diagnosis in embedded systems. By modeling system components and their interactions as a graph structure, GNNs can effectively capture intricate dependencies, enabling robust fault localization and predictive maintenance. The proposed approach integrates message passing neural networks to propagate fault-related information across system nodes, improving diagnosis accuracy and reducing downtime. Additionally, we leverage self-supervised learning to enhance generalization capabilities with limited labeled data. Experimental results on benchmark datasets and real-world embedded platforms demonstrate that GNN-based fault diagnosis outperforms traditional machine learning methods in terms of accuracy, interpretability, and computational efficiency. This study highlights the potential of GNNs in enhancing the resilience and reliability of embedded computing environments, paving the way for AI-driven, self-healing systems.