IMPROVE · Project

AI-Powered Predictive Monitoring That Catches Factory Breakdowns Before They Happen

manufacturingPilotedTRL 6

Imagine your factory machines could tell you they're about to break down — days before it actually happens. IMPROVE built a system that watches sensor data from production lines, learns what "normal" looks like, and flags anything unusual before it turns into costly downtime. Think of it like a fitness tracker for industrial equipment: it learns your machine's healthy patterns and alerts you when something's off. The system was tested with real prototypes across three different industries.

By the numbers

consortium partners across 6 countries

industry partners in the consortium

64%

industry ratio in the consortium

industrial prototypes in different industries

demonstrators used for validation

EUR 4,148,554

EU funding for development

total project deliverables

The business problem

What needed solving

Manufacturers lose significant money and time when machines break down unexpectedly or when switching between product runs takes too long. Traditional monitoring relies on scheduled maintenance or waiting until something fails, while manual modeling of machine behavior is expensive and slow. European factories need smarter ways to predict problems, optimize production, and reduce energy consumption — without hiring armies of data scientists.

The solution

What was built

The project built a virtual factory system with machine learning algorithms that automatically learn normal equipment behavior from sensor data, detect anomalies before breakdowns occur, and provide optimization recommendations through a Decision Support System and custom Human Machine Interface. These algorithms were validated on lab demonstrators and then deployed as 3 industrial prototypes across different industries, supported by 17 total deliverables.

Audience

Who needs this

Automotive and discrete parts manufacturers with high-cost unplanned downtimePlastics and continuous-process manufacturers struggling with long ramp-up phasesMachine builders wanting to add predictive maintenance to their product offeringMid-size manufacturers looking to reduce energy consumption in productionProduction managers dealing with complex supply chain optimization

Business applications

Who can put this to work

Automotive Manufacturing

enterprise

Target: Car parts manufacturers with complex production lines

If you are an automotive parts manufacturer dealing with unexpected machine breakdowns that halt your production line — this project developed a predictive monitoring system that learns normal machine behavior from sensor data and flags anomalies before they cause downtime. It was validated on lab demonstrators and transferred to 3 industrial prototypes across different sectors. With 14 consortium partners including 9 from industry, the solution was built for real factory conditions.

Plastics & Packaging

mid-size

Target: Mid-size plastics processors running continuous production

If you are a plastics processor struggling with long ramp-up phases when switching between product batches — this project built machine learning models that optimize production parameters automatically. Instead of manual trial-and-error, the system uses data-driven models combined with expert knowledge to reduce changeover time. The Decision Support System provides operators with clear optimization suggestions through a purpose-built Human Machine Interface.

Industrial Equipment & Machinery

any

Target: Machine builders looking to add smart monitoring to their products

If you are a machine builder wanting to offer predictive maintenance as a value-added service to your customers — this project created algorithms that automatically learn equipment behavior from sensor observations without requiring manual modeling. The technology was proven across 4 demonstrators and 3 industrial prototypes, making it adaptable to various machine types and production environments.

Frequently asked

Quick answers

What would it cost to implement this kind of predictive monitoring system?

The project itself received EUR 4,148,554 in EU funding across 14 partners over 3 years to develop the technology. Implementation costs for an individual factory would depend on existing sensor infrastructure and production complexity. Contact the consortium for licensing or deployment pricing.

Has this been tested at industrial scale or only in the lab?

Both. The project first validated algorithms on lab demonstrators, then transferred the implementation to 3 prototypes in different industries. With 9 industry partners in the consortium, real production environments were part of the testing from the start.

What about IP and licensing — can we use this technology?

The project involved 14 partners across 6 countries with a mix of universities and industry. IP arrangements would need to be discussed with the consortium coordinator (Technische Hochschule Ostwestfalen-Lippe, Germany). Based on available project data, specific licensing terms are not publicly documented.

How does this integrate with our existing factory systems and sensors?

The system was designed to learn from sensor observations already present in production environments, rather than requiring entirely new hardware. The Human Machine Interface was purpose-built to present complex data in an accessible way for operators. Integration specifics would depend on your current setup.

What specific production problems does this solve?

The project targeted four main challenges: reducing ramp-up phases when starting new production, optimizing production plants to increase cost-efficiency, reducing time to production with condition monitoring, and optimizing supply chains. Energy consumption reduction in manufacturing was also a key outcome.

How long before we would see results after implementation?

The project ran from 2015 to 2018, producing 17 deliverables including validated prototypes. Since the technology learns from sensor observations, initial results depend on how quickly the system can observe enough normal and abnormal patterns in your specific production environment.

Consortium

Who built it

The IMPROVE consortium is heavily industry-oriented, with 9 out of 14 partners (64%) coming from industry — a strong signal that this was built for real-world use, not just academic research. The partnership spans 6 countries (Austria, Germany, Denmark, Italy, Poland, Turkey), giving it broad European manufacturing coverage. With 4 universities and 1 research organization providing the scientific backbone, and 2 SMEs adding agility, the consortium is well-balanced for technology transfer. The coordinator is Technische Hochschule Ostwestfalen-Lippe in Germany, a university of applied sciences known for close ties to regional manufacturing industry.

TECHNISCHE HOCHSCHULE OSTWESTFALEN-LIPPECoordinator · DE
TRANSITION TECHNOLOGIES SAparticipant · PL
TECHNISCHE UNIVERSITAET MUENCHENparticipant · DE
EURICE EUROPEAN RESEARCH AND PROJECT OFFICE GMBHparticipant · DE
TRANSITION TECHNOLOGIES PSC SPOLKA AKCYJNAthirdparty · PL
ARCELIK A.S.participant · TR
OCME SRLparticipant · IT
UNIVERSITA DEGLI STUDI DI MODENA E REGGIO EMILIAparticipant · IT
MARMARA UNIVERSITYparticipant · TR

How to reach the team

Technische Hochschule Ostwestfalen-Lippe (Germany) — a university of applied sciences. Look for the project lead in their manufacturing or computer science department.

Order a report on this consortium See TECHNISCHE HOCHSCHULE OSTWESTFALEN-LIPPE profile →

Next steps

Talk to the team behind this work.

Want to explore how IMPROVE's predictive monitoring technology could reduce downtime in your production? SciTransfer can connect you with the right consortium partner for your specific industry and use case.

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