Performance-Enhanced Explainable AI-Assisted Fault Detection in Optical Fiber Networks using OTDR Analysis
C Jenila, Deva Varun Kumar D, Naveen Kumar Reddy B, C Gowtham, Dani Akash T, Hemanth M · 2025
This research presents an Explainable AI (XAI)-assisted machine learning approach for real-time fault detection in optical fiber networks using Optical Time-Domain Reflectometry (OTDR) data. Optical fibers, critical for modern communication, are prone to faults from physical disturbances like digging, shaking, and environmental conditions. The proposed method employs machine learning models—Subspace KNN, Logistic Regression Kernel, and Wide Neural Network—to classify fault events with high accuracy while minimizing false positives. Experimental results demonstrate superior performance, with Subspace KNN achieving 100% accuracy, Logistic Regression Kernel 99.7%, and Wide Neural Network 99.9%. XAI techniques such as SHAP and LIME enhance model interpretability, increasing operator trust and transparency. This research contributes to SDG 9 (Industry, Innovation, and Infrastructure) by improving communication network reliability and SDG 11 (Sustainable Cities and Communities) by enabling real-time fault monitoring in urban and remote areas. The proposed framework enhances automated fault management, ensuring efficient and interpretable fault detection in optical fiber networks.