AI-Powered Fault Diagnosis in Network Infrastructure: State-of-the-Art and Emerging Trends

Abstract

The rapid expansion of IoT, cloud services, and next-generation communication networks has introduced unprecedented complexity and heterogeneity into modern infrastructures, making effective fault diagnosis increasingly critical. Conventional approaches such as centralized monitoring, threshold-based alerts, and periodic inspections are often insufficient for detecting subtle anomalies, intermittent failures, and large-scale disruptions. Artificial Intelligence (AI) offers advanced, data-driven methods that enable proactive, adaptive, and predictive fault diagnosis, ensuring greater resilience and efficiency. Machine learning techniques, including unsupervised, supervised, and semi-supervised approaches, have been successfully applied to anomaly detection, fault classification, and predictive maintenance, providing improved accuracy and scalability. Deep learning architectures, such as Recurrent Neural Networks (RNNs), Convolutional Neural Networks (CNNs), and Long Short-Term Memory (LSTM) models, further enhance the recognition of complex temporal and spatial patterns in high-dimensional data. Reinforcement learning contributes additional benefits by enabling adaptive resource allocation and optimized fault recovery strategies in dynamic environments. Emerging trends highlight hybrid AI models, federated learning, edge intelligence, explainable AI, and digital twin-based monitoring, all of which contribute to robust, transparent, and intelligent infrastructures.