A Novel Deep Learning Approach for Non-Invasive Condition Assessment of Broadband Pipes

As countries worldwide transition from outdated copper networks to high-speed fiber-optic broadband, the installation of new cabling remains a labor-intensive and injury-prone process. The telecommunications industry urgently requires an efficient and safer method for assessing the condition of underground pipe infrastructure. However, no commercially available solutions currently address this need. In this study, we propose a novel approach that enhances existing condition assessment methods using acoustic reflectometry sensing. The developed Pipe Condition Assessment System (PiCAS) framework integrates a hybrid deep learning model to classify and localize blockages based on captured acoustic data. System performance was evaluated using four blockage sizes, achieving an F1-score of 0.99 on the validation dataset, demonstrating high accuracy in experimental conditions.