AI-Driven Fault Prediction and Restoration Leveraging Real-Time SOP Monitoring

This paper presents a real-time, artificial intelli- gence (AI)-powered framework for proactive fault detection and dynamic restoration in optical transport networks, leveraging continuous state-of-polarization (SOP) monitoring. By correlating live SOP telemetry from dense wavelength division multiplexing (DWDM) transceivers with machine learning-based analytics, the system accurately detects anomalous polarization patterns arising from external disturbances, commonly preceding fiber damage. Upon identifying critical events, the platform autonomously alerts the open software defined network (SDN) controller, which executes immediate traffic rerouting to ensure uninterrupted service. The solution is experimentally validated on a four- node optical ring, demonstrating precise event classification, low inference latency, and seamless network recovery. This work highlights the transformative potential of SOP-based sensing combined with intelligent control to realize self-healing, resilient optical infrastructures