Gated Recurrent Unit based Autoencoder for Optical Link Fault Diagnosis in Passive Optical Networks

Abstract

We propose a deep learning approach based on an autoencoder for identifying and localizing fiber faults in passive optical networks. The experimental results show that the proposed method detects faults with 97% accuracy, pinpoints them with an RMSE of 0.18 m and outperforms conventional techniques.

Faulty branch identification in passive optical networks using machine learning

“…In this respect, Usman et al [17] proposed a data-driven approach for fault monitoring in PON systems by combining the use of fiber sensors and ML whereby the fiber Bragg grating sensors having different characteristic grating for each branch are adopted to acquire the monitoring data, and the ML technique is used for fault identification. We presented a gated recurrent unit based autoencoder model that automatically identifies the type of the optical fiber fault in PONs and fully characterizes it without requiring either the intervention of trained personnel or the installation of additional equipment on the network infrastructure [18]. However, such model could not discriminate the faults occurring in similar length branches in PONs.…”

“…However, this approach necessitates varying branch lengths, which limits its applicability to real-world installed networks. Recently, machine learning (ML)-based approaches have demonstrated great promise for improving fault monitoring in PON systems by extracting insights from OTDR monitoring data without the intervention of trained personnel or the installation of additional network infrastructure equipment [4]. In this paper, we propose and discuss different ML approaches for faulty branch or ONU identification in PON systems with similar or close branch lengths by leveraging monitoring data obtained from reflectors installed at each branch's end.…”