Digital Twin-Enabled Fast Fiber Loss Anomaly Detection in Multi-Band Optical Transmission Systems via SRS Intensity Variation

In deployed optical networks, fiber loss anomaly cause transmission quality degradation and service interruption, presenting significant challenges to the availability and reliability of networks. This issue is particularly prominent in high-capacity and multi-band systems, where the volume of data affected by fiber loss anomaly significantly exceeds that in traditional C-band systems. In this paper, we propose a simple and rapid method for detecting fiber loss anomaly by leveraging the stimulated Raman scattering (SRS) effect. A digital twin (DT) of multi-band transmission system is constructed for training data generation through simulating various scenarios of fiber loss anomaly, considering anomalies that occur at random locations along the fiber and introduce varying levels of attenuation. Then, an DT-assisted neural network (NN) is trained to capture the inherent relationship between the received power of two edge channels (i.e., the lowest and highest frequency channels) and loss anomaly characteristics (i.e., anomaly location and additional attenuation). The real-time dropped power of these two edge channels can be captured by widely deployed optical channel monitors (OCMs) and fed into the NN for fiber loss anomaly detection. We constructed a C+L-band simulation setup to evaluate detection accuracy, accounting for OCM power measurement deviations, NN generalization, and SRS intensity dependent on both anomaly location and attenuation. Additionally, a C+L-band experimental system was built to validate the proposed method, and the results showed an average location error of 0.68 km and an attenuation error of only 0.05 dB.