Machine Learning Based Fiber Monitoring for Passive Optical Network

Abstract

In this paper, we present a machine learning-based approach for monitoring a failure in an optical fiber link for a passive optical network utilizing an optical mirror. A trained support vector machine (SVM) model is intended to detect and identify optical links with problems in an optical distribution network. The method is based on the optical reflected signal acquired from the optical reflectors, which is then processed using the SVM algorithm in MATLAB. A simulated 1x4 Gigabit passive optical network with a monitoring power of -4.16 dBm across a distance of 20 kilometres provides the training and testing dataset for the proposed model. Optimal SVM parameters are selected through cross-validation, resulting in an optimized non-linear decision boundary. The suggested technique is evaluated on a range of datasets produced from optical reflectors and indicates that the model achieves a high detection accuracy (97%-99%) when compared to the observed optical reflected spectra from the mirrors using an optical spectrum analyzer. The proposed technique is significant in that it may identify a failure in an optical channel using monitoring data supplied by multiple reflectors rather than an expensive optical spectrum analyzer or optical testing channel modules.