COGNITWIN · Project

Self-Learning Digital Twins That Predict and Prevent Factory Problems Before They Happen

manufacturingPilotedTRL 7

Imagine your factory equipment could learn from its own history, spot trouble brewing, and fix itself before anything breaks. That's what COGNITWIN built — a digital copy of your plant that watches sensor data in real time, learns what "normal" looks like, and warns you when something is drifting toward failure. Think of it like a seasoned operator who never sleeps, never forgets, and gets smarter every day. The system was tested on real steel, non-ferrous metals, and engineering production lines across Europe.

By the numbers

TRL 5 → TRL 7

Technology readiness advancement from lab validation to operational prototype

Industrial partners providing real operational test environments

Consortium partners across 7 European countries

Demonstrated deliverables including complete cognitive twins for 3 industrial domains

69%

Industry partner ratio in the consortium

The business problem

What needed solving

Manufacturing plants lose significant revenue to unplanned downtime, quality defects, and reactive maintenance. Traditional control systems respond to problems after they happen — they cannot learn from historical patterns or predict failures before they occur. Plant operators rely on experience and manual monitoring, which does not scale and walks out the door when experienced staff retire.

The solution

What was built

The project delivered a complete cognitive digital twin platform with two main toolboxes: a Sensor and Data Interoperability Toolbox for connecting plant sensors and data sources, and a Hybrid AI Cognitive Twin Toolbox for self-learning predictive models. Both toolboxes went through three development cycles. Complete working cognitive twins were deployed at steel production, non-ferrous metals, and engineering pilot sites.

Audience

Who needs this

Steel mills and foundries with aging equipment and high downtime costsNon-ferrous metals processors (aluminum, copper) seeking to reduce scrap ratesLarge process manufacturers (chemicals, cement, pulp & paper) running continuous operationsIndustrial automation companies wanting to add AI-based predictive capabilities to their offeringsPlant engineering firms looking for digital twin solutions to offer their clients

Business applications

Who can put this to work

Steel and metals production

enterprise

Target: Steel mills and non-ferrous metal producers

If you are a steel producer dealing with unplanned downtime and quality variations — this project developed cognitive digital twins tested directly on steel and non-ferrous metal pilot lines. The system uses sensor networks and AI to predict equipment failures and process deviations before they happen. It was demonstrated with 6 industrial partners across 7 countries, moving from lab validation to operational prototype.

Process manufacturing

enterprise

Target: Chemical, pulp & paper, or cement plants

If you are a process manufacturer struggling with unpredictable operations and rising maintenance costs — this project built a platform that connects your existing sensors to self-learning models. The cognitive twin adapts to changes in your process without manual recalibration. With 11 industry partners involved in development, the toolbox was designed for real industrial conditions, not just lab settings.

Industrial equipment and automation

mid-size

Target: Automation vendors and system integrators

If you are an automation company looking to add predictive intelligence to your product offering — COGNITWIN produced a sensor and data interoperability toolbox plus a hybrid AI cognitive twin toolbox, both iterated through three versions. These toolboxes can be integrated into existing control systems to add self-learning and predictive maintenance capabilities for your customers.

Frequently asked

Quick answers

What would it cost to implement this in our plant?

The project data does not include specific licensing or implementation costs. Since this was an Innovation Action with 16 partners, costs would depend on plant size and complexity. Contact the consortium lead SINTEF AS in Norway to discuss pricing for the toolboxes.

Can this work at full industrial scale, not just a pilot?

The project moved technology from TRL 5 (validated in controlled environment) to TRL 7 (prototype demonstrated in operational environment). Complete cognitive twins were delivered for steel pilots, non-ferrous pilots, and an engineering pilot — all in real industrial settings with 6 industrial partners.

Who owns the IP and can we license it?

IP is shared among 16 consortium partners led by SINTEF AS. With 11 industry partners and 3 SMEs in the consortium, there are likely multiple licensing paths. SINTEF AS as coordinator would be the first point of contact for licensing discussions.

How does this integrate with our existing control systems?

COGNITWIN specifically designed a Platform, Sensor and Data Interoperability Toolbox — iterated through initial, updated, and final versions — to connect with existing plant infrastructure. The system adds a cognitive layer on top of existing process control systems rather than replacing them.

What industries has this actually been tested in?

The project delivered complete cognitive twins for three industrial domains: steel production, non-ferrous metals processing, and engineering manufacturing. These were real operational demonstrations, not simulations, with 6 industry partners providing the test environments.

Is this ready to deploy or still experimental?

The project ended in February 2023 at TRL 7, meaning operational prototypes were demonstrated in real environments. The toolboxes went through three development cycles (initial, updated, final). It is past the experimental stage but may need customization for your specific plant.

Consortium

Who built it

COGNITWIN has a strongly industry-driven consortium with 11 out of 16 partners (69%) coming from industry, supported by 4 research organizations and 1 university across 7 countries (Germany, Spain, Finland, France, Norway, Serbia, Turkey). The project is led by SINTEF AS, one of Europe's largest independent research organizations based in Norway. With 6 industrial partners providing real test environments and 3 SMEs in the mix, this consortium was built to move technology out of the lab and into factories. The geographic spread across Western, Northern, and Southeastern Europe suggests the solution was tested under varied industrial conditions.

SINTEF ASCoordinator · NO
CYBERNETICA ASparticipant · NO
OULUN YLIOPISTOparticipant · FI
DEUTSCHES FORSCHUNGSZENTRUM FUR KUNSTLICHE INTELLIGENZ GMBHparticipant · DE
SUMITOMO SHI FW ENERGIA OYparticipant · FI
SIDENOR ACEROS ESPECIALES SLparticipant · ES
ELKEM ASAparticipant · NO
SIDENOR INVESTIGACION Y DESARROLLOSAthirdparty · ES
PRIVREDNO DRUSTVO ZA PRUZANJE USLUGA ISTRAZIVANJE I RAZVOJ NISSATECH INNOVATION CENTRE DOOparticipant · RS

How to reach the team

SINTEF AS (Norway) — one of Europe's largest independent research institutes. Search for COGNITWIN project coordinator at SINTEF for direct contact.

Order a report on this consortium See SINTEF AS profile →

Next steps

Talk to the team behind this work.

Want to explore whether COGNITWIN's cognitive twin toolbox fits your production environment? SciTransfer can arrange an introduction to the right consortium partner for your industry.

Order a report on this topic More in Manufacturing & Industry 4.0