Deep Learning-Based Fault Detection and Localization in Hollow-Core Fiber Communication Systems

Abstract

Hollow-core fiber communication systems are emerging as a transformative infrastructure for low-latency, high-capacity transport, particularly in data center interconnects, metropolitan backbones, and latency-sensitive computational fabrics. Unlike conventional solid-core silica fibers, hollow-core fibers confine light predominantly in air or vacuum, reducing nonlinear impairment and propagation delay while introducing new coupling, microbending, surface-mode, and manufacturing variability challenges. These characteristics alter the fault landscape: degradation may be more localized, more spatially intermittent, and less observable through conventional performance indicators. This paper examines deep learning-based fault detection and localization for hollow-core fiber communication systems from a system-level perspective. It analyzes how deep learning architectures can fuse optical time-domain reflectometry, spectral monitoring, coherent receiver telemetry, environmental sensing, and network control-plane data to identify and localize incipient failures. The discussion emphasizes architectural trade-offs involving latency, capacity, resilience, interpretability, cost, and operational complexity. It also considers governance, policy, sustainability, and fairness implications, especially where advanced diagnostic intelligence becomes concentrated among a small number of infrastructure operators. Through cross-domain comparisons with solid-core optical networks, data center fabrics, and critical infrastructure monitoring, the paper argues that hollow-core fiber fault analytics should be designed as a socio-technical capability rather than a narrow signal-processing task. The conclusion outlines deployment pathways, open challenges, and forward-looking perspectives for resilient, accountable, and equitable hollow-core fiber infrastructures.