AI Sniffs out Leaks Beneath Europe’s Streets

AquaSentinel uses AI and hydraulic science to spot leaks with fewer sensors, promising cheaper, smarter water grids

13 Feb 2026

Europe’s water utilities are confronting aging infrastructure and mounting pressure to curb losses, and researchers say artificial intelligence could change how leaks are found. A system known as AquaSentinel, described in a November 2025 preprint on arXiv, proposes detecting underground water leaks with far fewer physical sensors than utilities have traditionally required. Its developers argue that software-driven analysis, layered onto existing hydraulic models, may offer a less hardware-intensive path to modernizing water networks.

The project was developed by an international research team that includes contributors from Delft University of Technology, Texas A&M University and the University of Missouri. According to the preprint, the system combines established hydraulic principles with machine-learning algorithms to identify subtle irregularities in water flow. In simulated tests covering 110 leak scenarios, the researchers reported a 100 percent detection rate.

Those findings, while preliminary, arrive at a moment of heightened urgency. Across Europe, utilities face tightening sustainability mandates, rising infrastructure costs and growing public scrutiny over water losses. At the same time, the industry has been investing heavily in digital platforms, predictive analytics and A.I.-enabled monitoring. A system that reduces the need for dense sensor networks, long considered the backbone of leak detection, could significantly lower capital costs and reshape project design.

Analysts said innovations like AquaSentinel may influence procurement strategies, shifting emphasis from large-scale hardware deployments to more software-centric intelligence. The approach aligns with the broader rise of digital twins and advanced diagnostics in infrastructure management, where limited physical inputs are paired with detailed simulations to anticipate failures before they escalate.

Still, the research remains based on simulations, and real-world conditions can introduce complications, including aging pipes, fluctuating demand and data-quality constraints. Questions around data governance and cybersecurity could also affect adoption. If future pilot studies confirm the early results, AquaSentinel may help redefine how utilities approach leak detection, with implications for cost, efficiency and water conservation across the continent.