Close
Over the years, we have delivered consulting services across water infrastructure optimisation and modelling. We decided to integrate and amplify our techniques with domain-specific AI agents, and to make them available to everyone through foreKEY: three modules, one ecosystem.
Physics first. AI second.
The algorithm is the easy part.
Over the years, we have processed thousands of time series from sensors across drinking water, sewer and treatment networks — and modelled over 4,500 km of Italian water distribution networks under the Next Generation EU programme. Two things became clear: field data is rarely usable in its raw form, and existing modelling software is powerful but inaccessible. foreKEY was built around both problems.
EPANET AI · DISTRIBUTION
Hydraulic modelling in natural language, built
on EPANET
SEWER AI · SEWER
Advanced anomaly detection and sewer signal reconstruction
DATA HARMONIZATION LAYER
Temporal alignment · Outlier detection · Gap filling · Unit conversion — the trusted data foundation for the entire suite
Your data, finally trustworthy.
Water networks generate heterogeneous measurements — different instruments, different sampling rates, different units, different column names. Before any analysis can begin, that data needs to be harmonised. REED is the horizontal layer that does exactly that: it takes raw data from any source and makes it comparable, clean, and continuous.
One AI agent calls backend tools as needed. Four functions, one pipeline.
Instruments sampling at different rates are synchronised onto a common time grid. Interpolation and realignment preserve the integrity of the original signal.
Bad values are identified and temporarily removed from analyses. Simple outliers detected by REED can be inherited by BLEAK for deeper processing.
Short interruptions — an instrument offline for half an hour, a transmission failure — are covered with elementary interpolations, ensuring continuity in the data flow.
Heterogeneous units of measurement and different naming conventions are normalised automatically. Any source in, one format out.
The frontend always accompanies the backend with statistical charts and insights in the time and frequency
domain, for rapid visual inspection of harmonised data quality.
The hydraulic simulator that speaks your language.
NEWT combines a solid computational backend built on EPANET — the most validated hydraulic calculation engine in the world — with a conversational interface that dramatically lowers the barrier to hydraulic modelling. The chat is not a feature: it is the main interface. The AI agent is not an add-on: it is the architecture of the system.
“Show me where pressure drops
below 2 bar.” “What happens if I
close this pipe?” Type what you
need to know — NEWT runs the
simulation and responds. No
menus, no wizards.
Upload your network shapefiles
and get a ready-to-use hydraulic
model. The GIS pipeline is
guided by the AI agent, with
explicit confirmation at every
significant step.
Every network modification
requires explicit confirmation.
No operation is ever executed
without consent. Control stays
with the people who know the
network.
Navigate upstream, downstream, by district — combining spatial and temporal filters in a single request.
Validated on operational networks with tens of thousands of pipes. Not on simplified academic datasets.
Sewer signal, as good as new.
Sewer networks are the most hostile environment for instrumentation in the water cycle: fouling, drift, rainfall events, backflows. Raw sewer data is not simply noisy — it is systematically distorted in ways that depend on sensor type, season, weather conditions, and network morphology. BLEAK inherits simple bad values from REED and builds a deep, domain-specific layer of intelligence on top.
Backend with an ensemble of models trained on sewer measurements, orchestrated by a complex AI agent.
Classifies every period of the timeseries into three categories: dry weather, rainfall event, measurement anomaly. Identifies the sensor type and the nature of the problem — progressive fouling, drift, failure — with a depth that REED, without sewer specific tools, cannot reach.
Once the signal is classified, the ensemble of models reconstructs anomalous periods adapting to context: dry weather vs. rainfall, sensor type, position in the network. The outcome is a continuous, reliable signal — as good as new — ready for hydraulic models, Early Warning Systems, or regulatory reporting.
Via Chiesanuova 127/A
35136 PADOVA
Viale Virgilio 150
74123 TARANTO
UNI EN ISO 9001:2015
UNI EN ISO 14001:2015
UNI EN ISO 45001:2018
©2022 FORE SRL | Privacy Policy | Cookie Policy | Cap.Soc 10.000,00 € | VAT: 05413560284