Skip to main content
European Commission logo print header

Towards AI powered manufacturing services, processes, and products in an edge-to-cloud-knowlEdge continuum for humans [in-the-loop]

Periodic Reporting for period 1 - knowlEdge (Towards AI powered manufacturing services, processes, and products in an edge-to-cloud-knowlEdge continuum for humans [in-the-loop])

Reporting period: 2021-01-01 to 2022-06-30

AI is one of the biggest mega-trends towards the 4th industrial revolution. While these technologies promise business sustainability and product/process quality, it seems that the ever-changing market demands and the lack of skilled humans, in combination with the complexity of technologies, raise an urgent need for new suggestions.

To break the entry barriers for these technologies and unleash their potential, the knowlEdge project will develop a new generation of AI methods, systems and data management infrastructure. This framework will provide means for the secure management of distributed data and the computational infrastructure to execute the needed analytic algorithms and redistribute the knowledge towards a knowledge exchange society. To do so, knowlEdge proposes 6 major innovations in the areas of data management, data analytics and knowledge management: (i) A set of AI services that allow the usage of edge deployments as computational and live data infrastructure, an edge continuous learning execution pipeline; (ii) A digital twin of the shop-floor to test the AI models; (iii) A data management framework deployed from the edge to the cloud ensuring data quality, privacy and confidentiality, building a data safe fog continuum; (iv) Human-AI Collaboration and Domain Knowledge Fusion tools for domain experts to inject their experience into the system to trigger an automatic discovery of knowledge that allows the system to adapt automatically to system changes; (v) A set of standardization mechanisms for the exchange of trained AI-models from one context to another; (vi) A knowledge marketplace platform to distribute and interchange AI trained models.
With respect to objective 1: "Develop an AI-centric software architecture to support agile manufacturing scenarios", the work performed covered various tasks, from user need analysis and scenario definition to vision, specification and system architecture.

With respect to objective 2: "Simplify the development, integration and uptake of smarter and more frugal AI", the work involved designing the data collection platform component and its interfaces, including identifying functional and architectural requirements, the design of the component and the design of the unified data model to create a common data format in the platform.

Concerning objective 3: "Develop a deployment model capable of distributing data mining and analytic services across the compute", the task focused on developing a prototype for the knowledge discovery engine to perform unsupervised AI functionality.

In addressing objective 4: "Develop an interactive and multi-sided knowledge", the work was undertaken by making a first operational version of the knowlEdge ,arketplace available, including providing user interfaces for exploring AI models. The front-end design was based on best practices related to graphical user interfaces development, with contributions from various tasks, including the knowlEdge repository of WP5.

Regarding the implementation of objective 5: "Provide efficient and secure communication, data management and governance infrastructure", the objective was addressed with all tasks of WP3, ranging from work on data collection, integration, interoperability and data models to real-time retrieval, access and data governance.

With respect to objective 6: "Develop advanced user-facing applications and services based on real-world use-cases from the digital manufacturing domain", the required infrastructure in the form of the knowlEdge platform was set up. In this reporting period, the first version of digital twin for Parmalat’s scheduling was developed; integration, deployment and commissioning technologies are identified; security requirements are identified and implemented.

To implement objective 7: "To demonstrate and evaluate the knowlEdge framework in proof-of-concept use cases", the work focused on developing an implementation plan and evaluation methodology. The evaluation methodology explores all potential areas of interest to be more efficient and precise, including user satisfaction and KPI’s for AI impacts.

Concerning objective 8: "Communicate project outcomes, make contributions to standardization and wider awareness", tasks included defining the dissemination strategy and documenting corresponding activities. Work also included the initial framework for the business models, defining exploitation strategy and contributing to standardisation activities. Efforts were also undertaken to collaborate with other ICT-38 projects.
Progress beyond the start of the art includes automated knowlEdge discovery. Distributed AI and data analytics techniques will be developed based on deep learning of neural networks representing and interacting with real world industrial operators.

New methods for data placement strategies across the edge and the cloud are developed. As a result, an edge storage platform is developed, targeting low end devices running in the industrial domain, supporting offline and on-premises processing of the data. Privacy and confidentiality are ensured by masking confidential data and by providing access restrictions.

The project also aims to making products and services usable in a wide range of manufacturing scenarios. The variety of pilots and the platform’s open architecture and the possibility to trade knowlEdge solutions through marketplace lead to agile production processes and improved quality of products and processes.

Further, the project addresses the need to make humans work together with AI in optimal complementarity. This includes the platform’s ability to allow knowledge fusion routines and decision support. New business models are developed and educational events held to support adapting organizational structures. This can make humans more creative and productive and make plant operations and value chains more efficient and resilient. The solutions are tested, validated and standardized, are GDPR compliant, transparent and accountable. For example, the knowlEdge platform will help Parmalat employees in scheduling activities regarding production planning and sales forecasting. Other partners replace purely manual defect inspections with AI supported routines.

Partners’ absorptive capacity and exploitation potential are increased by creating new technological knowledge on AI, data mining & management. The project structure allows for cross-disciplinary and mixed teamwork. It encourages openness among partners, securing access to their background and the potential for future exploitation. Focus has, furthermore, been on value roadmaps through identification of value streams. Thus, insights based on multi-disciplinary research are created. Moreover, indirect impacts include helping EU manufacturing companies to boost their competitiveness by introducing reference architecture and AI technologies and create growing employment opportunities.

Environmental impacts include energy efficient data management, computation etc., increased process efficiency, less waste, zero-defect manufacturing. In Parmalat’s case the production scheduling will be improved, reducing waste in both terms of resources, products and time. In the case of Kautex-Textron and Bonfiglioli, the solutions aim at zero-defect manufacturing. Partly, defects will be handled and corrected almost in real-time without the need of re-work on them, reducing unnecessary resources and energy.

Related documents