Machine Learning Agents Leveraging Digital Twins for Failure Prediction in Optical Networks - (2025) - IEEE ICMLCN Paper

Aug 2, 2026

Upload by Sage

This study proposes an advanced framework for predicting optical amplifier failures in optical networks by integrating Digital Twins (DT) and Machine Learning (ML). Using the GNPy open-source framework, Digital Twins replicate amplifier behavior under diverse conditions and generate telemetry capturing both short-term dynamics and long-term performance trends. The simulated telemetry trains an LSTM model that achieves 98% classification accuracy for amplifier fault levels, identifying soft failures and estimating severity. The controlled DT environment supports dataset creation that is hard to obtain from live networks, enabling earlier detection, reduced service disruptions, and improved operational continuity for optical communications.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages7

Reading Time18 min

AccessFree to download

Abstract

Index Terms

Introduction

Motivation for proactive maintenance

Failure management with ML

Dataset imbalance challenge

Digital Twins as training-data alternative

POLITECNICO DI TORINO Repository ISTITUZIONALE Machine Learning Agents Leveraging Digital Twins for Failure Prediction in Optical Networks Original Machine Learning Agents Leveraging Digital Twins for Failure Prediction in Optical Networks / Mohamed, Mashboob Cheruvakkadu; Masood, Muhammad Umar; Ambrosone, Renato; Malik, Gulmina; D'Ingillo, Rocco; Straullu, Stefano; Bhyri, Sai Kishore; Galimberti, Gabriele Maria; Pedro, João; Napoli, Antonio; Wakim, Walid; Curri, Vittorio. - (2025), pp. 1-6. (Intervento presentato al convegno 2025 IEEE International Conference on Machine Learning for Communication and Networking (ICMLCN) tenutosi a Barcelona (Spa) nel 26-29 May 2025) [10 . 1109/icmlcn64995 .2025. 11140450] . Availability: This version is available at: 11583/3002800 since: 2025-09-04T14:16:52Z Publisher: IEEE Published DOI:10.1109/icmlcn64995.2025.11140450 Terms of use: This article is made available under terms and conditions as specified in the corresponding bibliographic description in the repository Publisher copyright IEEE postprint/Author's Accepted Manuscript ©2025 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collecting works, for resale or lists, or reuse of any copyrighted component of this work in other works. (Article begins on next page) 04 October 2025 Machine Learning Agents Leveraging Digital Twins for Failure Prediction in Optical Networks Mashboob Cheruvakkadu Mohamed Politecnico di Torino, Italy mashboob.cheruvakkadu@polito.it Muhammad Umar Masood Renato Ambrosone Gulmina Malik Politecnico di Torino, Italy Politecnico di Torino, Italy Politecnico di Torino, Italy muhammad.masood@polito.it renato.ambrosone@polito.it gulmina.malik@polito.it Rocco D’Ingillo Politecnico di Torino, Italy rocco.dingillo@polito.it Stefano Straullu Links Foundation, Torino, Italy [stefano.straullu@linksfoundation.com](stefano.straullu@linksfoundation.com) Sai Kishore Bhyri In􀀂nera India sbhyri@in􀀂nera.com Gabriele Maria Galimberti In􀀂nera USA ggalimberti@in􀀂nera.com Jo˜ao Pedro In􀀂nera Unipessoal Lda, Portugal jpedro@in􀀂nera.com Antonio Napoli In􀀂nera Germany anapoli@in􀀂nera.com Walid Wakim In􀀂nera USA wwakim@in􀀂nera.com Vittorio Curri Politecnico di Torino, Italy vittorio.curri@polito.it Abstract—This study proposes an advanced framework for predicting optical ampli􀀂er failures in optical networks by integrating Digital Twins (DT) and Machine Learning (ML). Utilizing the GNPy open-source framework, DTs replicate ampli􀀂er behavior under various conditions, resembling faults and captures the network conditions and performance metrics of the optical networks. The telemetry data generated from these simulations represents both short-term dynamics and long-term trends in ampli􀀂er performance, enabling the training of a Long Short-Term Memory (LSTM) model. The ML model demonstrates an ‘ampli􀀂er fault level’ classi􀀂cation accuracy of 98 %, effectively identifying soft failures and assessing fault severity. By leveraging the ability to model complex fault scenarios in a controlled environment, the framework provides a comprehensive solution for generating datasets that are otherwise dif􀀂cult to obtain from live networks. This approach enables early detection and intervention, minimizes service disruptions, and enhances network reliability. The integration of DTs and LSTM-based ML offers a scalable and data-driven solution for improving the resilience, ef􀀂ciency, and operational continuity of modern optical communication systems. Index Terms—Machine Learning, Digital Twin, Optical Ampli􀀂ers Fault Detection, Soft failures, Proactive Failure Management, Network Resilience. I. INTRODUCTION Today’s optical networks are constrained by rigid operational approaches, which can

FAQs

What is the proposed framework for failure prediction in optical networks?

The framework integrates Digital Twins (DT) and Machine Learning (ML) to predict optical amplifier failures using simulation-generated telemetry data.

How are training data and amplifier behavior represented?

Digital Twins built with the GNPy framework replicate amplifier behavior under various conditions and produce telemetry capturing short-term dynamics and long-term trends.

What ML model is trained and how accurate is it?

An LSTM model is trained on the telemetry data and reports 98% classification accuracy for amplifier fault level, enabling detection of soft failures and fault severity assessment.

Aug 2, 2026

Upload by Sage