Classification of In-Service Optical Time Domain Reflectometer Trace Changes Using Supervised Learning

Optical fibers form the backbone of global telecommunications, and ensuring their reliability is crucial for network operators. Conventional Optical Time Domain Reflectometer (OTDR) methods focus on discrete events, but inservice measurements can detect distributed effects like fiber loss, Raman amplification, and Stimulated Raman Scattering (SRS) effects, which traditional methods struggle to handle. In this paper, we show how supervised learning models can effectively classify changes observed in in-service OTDR traces caused by fiber loss, Raman gain and channel loading. Multilayer Perceptron (MLP) outperformed Random Forest (RF) and Convolutional Neural Network (CNN) with 0.891 accuracy. ©2024 The Author(s)