Simply supplying or operating highly productive and reliable machines, systems, or components will no longer be sufficient as a distinguishing feature and basis for business success in the future. In addition to traditional hardware-related skills, factory operators and their equipment suppliers must quickly learn and master comprehensive skills to make effective use of new methods and tools such as artificial intelligence (AI) and machine learning (ML), digital twins, and data ecosystems/data spaces. This will not be successful on its own: only in cooperation with like-minded partners across the entire value creation network, where everyone plays to their strengths, can AI models create added value for all involved partners.
Data ecosystems and data spaces form the basis for the beneficial use of industrial AI, especially for small and medium-sized manufacturing companies. They standardize data from different sources, offer easy access to the data ecosystem via standardized connectors, and thus enable reusable AI applications. In the following paragraphs, further key aspects and advantages are listed.
Data ecosystems facilitate easy data accessibility and quality, even across company borders, e.g., by providing access to machines process, quality, and/or maintenance data from various sources. Standardized formats and metadata defined by the data ecosystems facilitate the targeted use of runtime data. Open standards, which are already beginning to become widespread in the manufacturing industry, then facilitate the combination of data from different systems, e.g., ERP, MES, SCADA, or PLM systems.
Companies are often reluctant to exchange data across company borders because they fear to reveal valuable knowledge that has been built up over many years. Data ecosystems offer maximum protection for precisely this purpose through defined governance and data security: defined roles, access and usage controls, trusted data sources, and automatable data contracts (data access agreements) facilitate and support data exchange within the network. This enables ecosystem partners, such as suppliers, customers and service providers, to integrate data from entire value chains and exploit all potential benefits that they would never be able to leverage on their own.
This results in significant advantages when using artificial intelligence and machine learning:
- Existing, prepared data sets can be used for multiple AI applications, e.g., predictive maintenance, process optimization, quality assurance, etc.
- The AI models have higher model quality and are trustworthy because consistent data improve model accuracy, leading to stability and comprehensibility of AI applications.
- Sharing data in trusted data ecosystems avoids costly isolated solutions and enables the gradual and therefore risk-reduced deployment of AI, from pilot projects to serial production. Open standards lead to less data preparation per project and thus to faster prototypes and their iterative optimization.
- Partnership-based value creation arises because companies in a value chain use data cooperatively, e.g., jointly trained models with suppliers, thereby strengthening their competitiveness together. Ultimately, every participant in the data ecosystem benefits.
Typical usage scenarios are
- Predictive maintenance, whereby AI detects at an early stage when machines or components are at risk of failure, when maintenance can be optimally scheduled in the complex production plan without jeopardizing deadlines, or simply reduces downtime.
- Process optimization, for example, in the fine-tuning of process parameters such as temperature, speed and pressure to reduce waste.
- Quality predictions through early detection of deviations or reduction of scrap and rework.
- Energy and resource efficiency with AI models identifying waste and improving consumption curves.
- Digital twins of individual plants or lines, in which runtime data from machines and plants is combined with data from physical simulations to enrich the models or simply generate more data for machine learning purposes.
This allows AI applications to be implemented faster, more cost-effectively and with less risk, resulting in reduced downtime, better quality, more efficient production and ultimately improved competitiveness.
Data space labs act as AI incubators
Many factory operators and their equipment suppliers are small and medium-sized enterprises, most of which do not have a large IT department, only little floor space for prototypes, hardly any free capacity for data collection from running machines or hardly any personnel capacity for research projects.
The work described above to utilize artificial intelligence and machine learning in the data ecosystem is something that hardly any company can accomplish on its own: the actual knowledge gap should be closed in collaboration with trusted partners. The secure exchange of data promotes cooperation and innovation within the ecosystem, enabling the economic implementation of new business models that were previously unprofitable.
This is precisely why Fraunhofer IOSB is preparing its well-known and proven research factories and AI real-world laboratories in Karlsruhe [1] and Lemgo [2] to become data space labs: Here, manufacturing companies will find everything they need to quickly and easily develop and test data space technologies.
A data space lab must offer all the elements necessary to quickly set up a data space environment, obtain heterogeneous data from various sources, standardize and process it, and generally make the added value for companies from the data ecosystem clear and understandable. These elements include
- Options for instrumenting machines, systems and production processes in cases where the sensor technology intended or already in place by companies is insufficient for data acquisition.
- Processing of production data in the ‘edge-cloud continuum’, i.e., on-site computing capacity and connection to a powerful and secure cloud infrastructure, e.g. to train and create AI- and ML-models of components and machines to demonstrate predictions regarding quality, availability or output.
- MX-Port configuration tools for quickly connecting assets to the data space via the various versions of MX-Port [3] (currently Hercules – EDC, Leo – Asset Admin. Shell AAS, and Orion – OPC UA).
- Expertise in IT security management (IEC 62443) and IT security (network, communication protocols, IT/OT-coupling, access/usage control, etc.).
- Systematic development of AI applications and their operation in industrial data spaces [4] in accordance with the methodology of AI engineering [5].
Using these elements, data space labs offer different formats, ranging from bilateral support for individual companies to hackathons, e.g., for collaborative testing of connectors, to governance workshops, in which participants discuss and agree on rules, roles, and responsibilities in a data room. Standardization initiatives can also be initiated by the data space labs.
Author:
Von Dr. Olaf Sauer, Fraunhofer-Institut für Optronik, Systemtechnik und Bildauswertung, Karlsruhe (v.i.S.d.P.)
Sources
[1] https://www.reallabore-innovationsportal.de/karlsruher-forschungsfabrik-fuer-ki-integrierte-produktion, letzter Aufruf am 26.01.2026
[2] https://smartfactory-owl.de/ki-reallabor/, letzter Aufruf am 26.01.2026
[3] https://factory-x.org/wp-content/uploads/MX-Port-Concept-V1.10.pdf, letzter Aufruf am 26.01.2026
[4] Usländer, T. (2025). KI-Engineering in industriellen Datenräumen. In: Hoffmann, C.H., Hersberger, S. (eds) Wie die Künstliche Intelligenz die Wirtschaft verändert. Springer, Wiesbaden. https://doi.org/10.1007/978-3-658-46839-2_2
[5] Usländer, Th., Schulz, D. (Hrsg.): KI-Engineering in der Produktion. Whitepaper der Fraunhofer-Institute IOSB und IAIS. Stuttgart: Fraunhofer Verlag, 2023. https://doi.org/10.24406/publica-1685.
Links
https://industry-insights.podigee.io/32-sauer-fraunhofer-iosb-manufacturingx-industrie-forschung
