Design and Evaluation of a Real-Time AI-Driven Autonomous Network Assistant (AI-ANA) For Management and Fault Diagnosis in Iran Telecommunications Company Large-Scale Networks

Managing complex multi-layer networks through traditional fault diagnosis methods has become increasingly challenging due to large scale and intricate internal dependencies. This paper presents the design and evaluation of an autonomous network assistant, named AI-ANA, which employs artificial intelligence to deliver an integrated solution for the intelligent management of such networks. The proposed framework, by combining machine learning—specifically leveraging statistical anomaly detection (Z-score) for real-time fault flagging and LSTM for predictive loop modeling—and natural language processing, provides capabilities including real-time monitoring, intelligent anomaly detection, root-cause analysis across different network layers, and automated configuration management for remediation. The system architecture is based on a modular and event-driven approach, enabling concurrent processing of diverse network data streams. Evaluation of the system was conducted in a precise simulated environment with realistic failure scenarios across L۲, MPLS, and VPLS domains. The assessment results indicate a significant improvement in key performance indicators compared to conventional methods, specifically showing a marked reduction in fault detection and recovery time, along with a high Automation Success Rate. These accomplishments have led to a considerable decrease in the operational burden on network management teams. This research represents an effective step toward realizing the vision of self-healing and self-managing networks and demonstrates the high potential of deeply integrating artificial intelligence into next-generation network management systems.

کلیدواژه ها:

Autonomous Network Assistant ، AI-Driven Network Management ، Multi-Layer Networks ، Fault Diagnosis ، Z-score