Abnormal Event Monitoring of Drainage Pipelines Based on Distributed Optical Fiber Vibration Sensing

Urban drainage networks are critical infrastructure for modern cities, yet their complex and concealed nature makes real-time monitoring of abnormal events a persistent challenge. To address this, this work proposed a distributed optical fiber vibration sensing system based on phase-sensitive optical time-domain reflectometry (Φ-OTDR) for drainage pipeline monitoring. To accurately capture abnormal events from complex environmental noise, a two-tier intelligent processing architecture comprising localization and classification modules was developed. The localization module employed an enhanced You Only Look Once (YOLO) model integrating anchor-box optimization, a channel-attention module, and an additional small-target feature layer, achieving a mean average precision of 96.8% in laboratory tests. The classification module utilized a hybrid model combining convolutional neural network, bidirectional long short-term memory, and attention classifier, achieving an overall accuracy of 98.1% for five abnormal events. During two-week of continuous field monitoring, the system localized simulated events with errors below 2.60 m and distinguished internal leakage, knocking, and drilling from walking, running, and vehicle interference with precision and recall exceeding 99.6% for the tested classes. Notably, the system successfully detected and accurately localized an actual branch pipe leakage event, demonstrating its practical effectiveness under real-world operating conditions.