Manufacturing Data Spaces for Artificial Intelligence: the EC perspective
It is a common sense statement that the quality of advanced AI-based decisions heavily depends on the quality of the Data we introduce into the system: Garbage in, Garbage out. Common European Data Spaces are currently discussed, developed and deployed in several initiatives at European level with the aim to break the silos and fill the gaps that currently prevent a pan-EU “free flow of data”. In the Manufacturing sector, four initiatives are to be considered in order to determine a state of play regarding this strategic matter: The EC directives and communications; the BDVA (Big Data Value Association) working groups; the IDSA (International Data Spaces Association) publications and papers; the Connected Factories (EFFRA European Factories of the Future Research Association) and Digital Manufacturing Platforms ecosystem. In this first blog regarding Manufacturing Data Spaces for AI, we will address the EC viewpoint through its communications and recommendations.
The 2018 EU Data Package included a specific staff working document on how to implement data sharing spaces in the Private Sector . Three main Business Models have been identified for the Private sector to fully benefit from the Data Economy and the Data Revolution: the Open Data model, the Data Monetisation model and the Trusted Ecosystem model.
In the first model, private sector companies could contribute to the Open Data movement by disclosing and publishing DataSets which do not include privacy or confidentiality issues. The typical case includes out-of-production data maybe aggregated, anonymised or pseudonymised, put at disposal of open innovation “livinglab” ecosystems for advanced experimentations. More recently, this has become an important topic for the so-called Teaching / Learning / Didactic Factories where companies and students can find a hands-on “test before invest” facility to innovate and experiment.
In the second model, private sector companies are able to valorise the data they produce, to provide an internal or externalised Data Marketplace and to start and develop new servitisation business models. Manufacturing enterprises, e.g. in the sector of Machine Tools and Robots, publish and valorise high value Datasets which could be used by service providers to develop and test their advanced service value propositions, e.g. in diagnosis and maintenance of complex equipment. The monetisation of such a business could be implemented in different ways like credits or virtual coins and not just by immediate e-commerce and payment functions.
In the third model, private sector value chain is organised in hierarchical (e.g. tier 1, tier 2, tier n) or non hierarchical (e.g. SMEs ecosystems) trusted networks implementing in a more flexible and configurable way what the EDI did many years ago. Hierarchical trusted network usually follow the business and governance model of a large company (e.g. a car manufacturer, a ship building, a food producer) which dictates the way the network should behave in order to contribute to the trusted cluster. Along the time, such rigid models and chains have been gradually transformed in more open networks where entry barriers have been substantially lowered and internal competition rules blurred. Non-hierarchical trusted networks are usually implemented by local regional districts of SMEs usually sharing in real time their productive capacity and allow dynamic, on-the-fly matching between demand and offer of manufacturing services (e.g. MaaS Manufacturing as a Service).
The 2020 EU Data Strategy finally defined four main pillars in order to implement Data Economy: A cross-sectoral governance framework for data access and reuse (now implemented by the Data Governance Act ), a set of Technology Enablers implementing Personal and Industrial Data Platforms, a pool of new professions and competencies to introduce Data Economy in the enterprise, an ecosystem of common EU Data Spaces in several crucial economic sectors, such as Manufacturing
The STAR ICT-38-2020 project is one of the H2020 projects aiming at applying the EU Data Strategy principles to the field of “Data for AI in the Manufacturing sector”, especially focussing on technological enablers and pillars for Data Quality, Cyber Security, Explainable and Trustworthy AI. Moreover, STAR technology assets enable the three Business Models of the 2018 Data Package through its Virtual DIHs (Open Data), Assets Catalogue (Data Marketplace) and its three industrial pilots (Trusted Networks) in the domains of Human-Cobot Collaboration for Robust Quality Inspections, Human Centred AI for Agile Manufacturing 4.0 and Human Behaviour Prediction and Safe Zone Detection for Routing.