Machine Learning Based Optical Fiber Fault Detection and Localization Using OTDR Signals

Year : 2026 | Volume : 16 | Issue : 02 | Page : 41 52
By

Shweta Zambare,

Shivani Shendage,

  1. Student, Department of Electronics and Telecommunication Engineering, SKN Sinhgad College of Engineering, Pandharpur, Maharashtra, India
  2. Student, Department of Electronics and Telecommunication Engineering, SKN Sinhgad College of Engineering, Pandharpur, Maharashtra, India

Abstract

Optical fiber networks provide high bandwidth, low transmission loss, and immunity to electromagnetic interference, making them fundamental to modern communication infrastructure. However, fiber faults caused by cable cuts, excessive bending, connector degradation, splicing defects, and physical disturbances can significantly affect network reliability. Optical time-domain reflectometry (OTDR) is widely used for identifying and locating such faults, but conventional OTDR interpretation often depends on threshold-based signal processing and manual analysis. This study presents a machine-learning-based framework for automatic optical fiber fault detection and localization using OTDR signals. A parameterized dataset containing 1,200 illustrative OTDR traces is considered, including normal fiber, fiber cut, bend, faulty connector, and splice-loss conditions. Signal features such as reflection amplitude, attenuation slope, event width, peak intensity, local variance, and event distance are extracted from the OTDR traces. Random Forest, Support Vector Machine, Artificial Neural Network, and XGBoost models are employed for fault classification, while regression models are used to estimate fault location. The proposed methodology combines signal preprocessing, feature extraction, machine-learning classification, and distance estimation into a unified diagnostic framework. Illustrative results indicate that XGBoost provides the highest classification accuracy of 96.8% and an F1-score of 96.5%, while the localization model achieves an illustrative mean absolute localization error of 42.6 m. The proposed approach demonstrates the potential of machine learning for automated OTDR-based fault diagnosis in embedded and intelligent optical-network monitoring systems. All numerical results are illustrative and must be validated using actual OTDR measurements before publication.

 

 

Keywords: OTDR, optical fiber, machine learning, fault detection, fault localization, random forest, SVM, ANN, XGBoost, optical communication

[This article belongs to Trends in Opto-electro & Optical Communication ]

How to cite this article: Shweta Zambare, Shivani Shendage. Machine Learning Based Optical Fiber Fault Detection and Localization Using OTDR Signals. Trends in Opto-electro & Optical Communication. 2026; 16(02):41-52.
How to cite this URL: Shweta Zambare, Shivani Shendage. Machine Learning Based Optical Fiber Fault Detection and Localization Using OTDR Signals. Trends in Opto-electro & Optical Communication. 2026; 16(02):41-52. Available from: https://journals.stmjournals.com/toeoc/article=2026/view=260573

References

  1. Godase V, Pawar P, Nagane S, Kumbhar S. Automatic railway horn system using Node MCU. J Control Instrum. 2024;15(1):11-19.
  2. Godase V, Godase J. Diet prediction and feature importance of gut microbiome using machine learning. Evol Electr Electron Eng. 2024;5(2):214-219.
  3. Jamadade VK, Ghodke MG, Katakdhond SS, Godase V. A comprehensive review on scalable Arduino radar platform for real-time object detection and mapping. J Microprocess Microcontroller Res. 2025;2(2):1-2.
  4. Godase VV. A comprehensive study of revolutionizing EV charging with solar-powered wireless solutions. Adv Res Power Electron Devices. 2025;2(1):23-37.
  5. Salunkhe A, Pawar V, Pise P, Mule S, Survase A, Godase V, et al. A review on real-time RFID-based smart attendance systems for efficient record management. Adv Res Analog Digit Commun. 2025;2(2):32-46.
  6. Dhope V, Chavan A, Hadmode N, Godase V. Smart plant monitoring system. Int J Creat Res Thoughts. 2024 May;12(5):b844-b849. doi:10.5281/zenodo.11213525.
  7. Abdelli K, Grieser H, Tropschug C, Pachnicke S. Optical fiber fault detection and localization in a noisy OTDR trace based on denoising convolutional autoencoder and bidirectional long short-term memory. J Lightwave Technol. 2022;40(8):2254-2264. doi:10.1109/JLT.2021.3138268.
  8. Wang D, Zhang C, Chen W, Yang H, Zhang M, Lau APT. A review of machine learning-based failure management in optical networks. Sci China Inf Sci. 2022;65(11):211302. doi:10.1007/s11432-022-3557-9.
  9. Singh S, Raj G, Abdohwr R, Dhamija K, Kumar A, Chohan JS. Machine learning-driven fault detection in long-haul optical fiber networks. 2025 Optical Communication, Photonics, Telecommunications, and Intelligent Machine Applications (OPTIMA), Tashkent, Uzbekistan. 2025. p.360-4. doi:10.1109/OPTIMA67660.2025.11380344.
  10. Srinivasan SR, Rehaan AM, Anto JJ, Srikamu C, Sujanth Narayan KG. AI-based OTDR multi-fault detection system for optical fiber networks. Eng Res Express. 2026;8(3):035527. doi:10.1088/2631-8695/aea3bd.
  11. Soothar KK, Chen Y, Memon KA, Magsi AH, Khan A, Qureshi KK. Enhancing fault detection and localization in passive optical networks through advanced deep learning and explainability techniques. Arab J Sci Eng. 2025;50(22):19025-19042. doi:10.1007/s13369-025-10370-z.
Regular Issue Subscription Review Article
Volume 16
Issue 02
Received 21/09/2026
Accepted 22/09/2026
Published 23/09/2026
Publication Time 2 Days

Login

My IP

PlumX Metrics

Support