FENICE
Federated Edge iNtelligence for Industrial Critical infrastructurEs
FENICE - Federated Edge iNtelligence for Industrial Critical infrastructurEs
Project co-funded by the European Union under the ERDF Regional Programme Campania 2021-2027 (PR Campania FESR 2021-2027), Priority 1 bis “Digital, clean and biotechnologies: contribution to the STEP platform”, Specific Objective 1.6, Action 1.6.1 “Supporting the development and manufacturing of critical technologies”. Public call “Aid for the development and manufacturing of critical technologies – STEP, 2nd edition”.Beneficiary: Frontiere S.r.l.CUP: B67H26001730004Project location: Naples, Centro Direzionale, Via G. Porzio 4Duration: 1 June 2026 – 30 May 2028 (24 months)Eligible investment: € 1.211.054,00Grant awarded: € 823.329,03
The Project
Critical infrastructures such as distribution networks, industrial plants and utilities generate large volumes of sensor data every day. These data are often noisy, incomplete or collected at sites with unreliable connectivity. FENICE develops an artificial intelligence software platform that brings analytics directly to the field, on edge devices installed at the plants. The goal is to anticipate rare, critical events such as failures, anomalies and degradation before they turn into problems.The platform uses federated learning techniques, so models improve by learning from data across multiple sites without centralising it, strengthening confidentiality, security and operational continuity.
Objectives
- Build a pipeline that makes sensor data reliable, separating noise and instrument disturbances from meaningful signals.
- Develop multi-horizon predictive models that give early warning of rare events while keeping false alarms low.
- Enable distributed learning across multiple sites, with robustness and privacy safeguards.
- Ensure cybersecurity and regulatory compliance by design (Cyber Resilience Act, AI Act, GDPR).
- Industrialise the solution with an installable edge runtime and an MLOps platform to manage the model lifecycle.
Expected results
- A software suite deployable on industrial gateways, able to operate even with intermittent connectivity.
- AI modules for forecasting and anomaly detection, validated at a real pilot site.
- A platform to train, release and monitor models in a reproducible and auditable way.
- Moving the technology from demonstrator level to a configuration ready for industrial use and replicable at new sites.
Activities
The project is organised into eight work packages. It starts with requirements and architecture definition, continues with industrial research on data quality, rare-event forecasting, distributed intelligence and cybersecurity, and ends with the experimental development of the edge runtime, the MLOps platform and field testing and validation.