Leveraging Optical Communication Fiber and AI for Distributed Water Pipe Leak Detection

Water distribution networks (WDNs) are essential infrastructure for providing fresh water to communities, but detecting leaks for WDNs is challenging and costly. In this article, we propose a novel solution that combines an optical network and WDN for distributed water pipe leak detection. Our approach involves using a standard outdoor fiber-optic cable for distributed vibration measurement along a 40-meter water pipe. To accurately identify and locate leaks, we introduce a leak identification algorithm based on 3Dconvolutional neural networks (3D-CNNs) that consider the temporal, spectral, and spatial information. Additionally, we propose a leak quantification method that can help prioritize repairs based on the severity of the leak. We evaluate our scheme for different conditions and find that it can detect leak flow rates as low as 0.027 L/s with a location accuracy of within 3 meters and a quantification accuracy of over 85%. Our proposed method offers a cost-effective and valueadded solution for designing optical networks and WDNs in new development areas.