AI-Detector

The problem

The management of water distribution networks has often lagged behind today’s available technologies. In many cases, knowledge of how an entire water system works rests with a handful of municipal technicians who, once retired, take years of experience and information with them that no one else has — a real risk for infrastructure that has not always received the attention it needs over time.

In this context, promptly detecting bursts and hidden leaks is crucial for safeguarding water resources. The problem is even more pressing in mountain water networks, characterised by high pressures: the higher the network pressure, the greater the amount of water lost in the event of a failure, with significant economic and environmental impacts.

How it works

AI-Detector was created to bridge this gap with a system capable of autonomously learning a water network’s behaviour from its physical characteristics — flow rates and pressures measured in the field — and interpreting its status in real time, with hourly resolution.

At the heart of the system is a predictive model that learns to accurately estimate the network’s expected behaviour. By constantly comparing this forecast with the real measured data, AI-Detector promptly identifies differences between what is expected and what actually happens — the typical signal of a burst or anomaly — classifying them and generating targeted alerts for utility managers.

The application is designed to integrate into the network’s daily management, offering a dynamic diagnostic tool rather than a single snapshot in time.

The impact

For water utility managers — municipalities and utilities — promptly detecting leaks means being able to plan targeted maintenance interventions, rather than only acting once the damage is already evident. This translates into a real reduction in water waste and operating costs.

AI-Detector was also created with the aim of supporting utilities in their transition towards more modern network management, at a time when historical knowledge tied to individual people risks being lost and water demand is under growing pressure from climate and socio-economic factors.

The project is rooted in the scientific research activity of the AIAQUA team, a spin-off of the Free University of Bozen-Bolzano, and in particular in studies on the most advanced methods for burst detection in water distribution networks.