AI-driven physical-layer fault monitoring and diagnosis in optical fiber communication networks

Abstract The rapid evolution of optical fiber communication networks has significantly increased the demand for intelligent and automated fault monitoring at the physical layer. Traditional methods often struggle to provide timely and accurate diagnosis under dynamic network conditions. To overcome these limitations, this study presents an artificial intelligence–based framework for reliable monitoring and diagnosis in optical communication networks. The proposed approach utilizes critical transmission quality indicators, namely optical signal-to-noise ratio (OSNR), Q-factor, and bit error rate (BER), to characterize network conditions. A deep neural network (DNN) classifier is designed to distinguish among normal operation, moderate faults, and severe fault scenarios. Experimental results demonstrate an overall training accuracy of 99.58 % and validation accuracy of 99.33 %. The training phase achieved precision, recall, and F1-scores of 0.9930–0.9977, 0.9943–0.9966, and 0.9936–0.9971, respectively, across the three classes. Similarly, validation precision, recall, and F1-scores ranged from 0.9902–0.9944, 0.9918–0.9956, and 0.9910–0.9950, respectively. These results confirm the robustness and generalization capability of the proposed framework. Furthermore, scatter plots demonstrate that OSNR and Q-factor decrease with increasing fault severity, whereas BER increases, confirming their effectiveness for fault diagnosis.