Leak Detection with Economics-Driven Convolutional Neural Networks applied to German cities and benchmarking with previous studies

This project aims to employ cutting-edge deep learning models to address the critical issue of water leakage in water distribution networks (WDNs). Leakage in WDNs leads to significant water wastage, infrastructure damage, service disruption, and even contamination. The proposed approach leverages Convolutional Neural Networks (CNNs) trained explicitly for optimizing leak detection. Using synthetic datasets that simulate leaks of varying sizes, times, and locations, it is intended to develop a model that excels at economic score-based leak detection. This project’s primary focus is to provide efficient and accurate AI-based solutions for promptly identifying and pinpointing leaks in WDNs across German cities; since Canadian cities are experiencing a high level of water leakages, in the future steps, applying the developed model to Canadian cities. The expected benefit for participating institutions is to enhance the sustainability and reliability of water distribution systems, reduce water loss, and contribute to a more resilient and cost-effective water infrastructure management.

Faculty Supervisor:

Rebecca Dziedzic

Student:

Partner:

Technische Universität Berlin

Discipline:

Engineering

Sector:

Water; Sustainability & the Environment; Artificial Intelligence

University:

Concordia University

Program: