On August 23, 2026, three papers from the NARRATE project were presented at the workshop “Transdisciplinary Digital Manufacturing: Integrating Industrial Practice with AI, IoT and Data-Driven Research”, together with a paper from the sister project MAASive. The workshop took place as part of the annual conference of the Society for Process and Design Sciences (SDPS), which focused on the theme “Transdisciplinary Science in Action”.
Bernd Krämer served as workshop chair and briefly introduced the NARRATE project. He described NARRATE as an approach to transforming conventional, fragmented supply chains into a Smart Manufacturing Network through federated data integration, Digital Twin-based end-to-end visibility, and AI-driven predictive and prescriptive analytics, coordinated by the Intelligent Manufacturing Custodian.
The paper “Towards Semantic Digital Twin Ecosystems for Regenerative, Resilient and AI-Enabled Manufacturing Networks”, presented by Amal Elgammal, outlined the overall architectural vision of NARRATE: a “Blueprint-Based Semantic Digital Twin Ecosystem” that provides the shared semantic foundation for the various components developed in the project.
The architecture organizes manufacturing knowledge into interconnected “Blueprint Views”, including a Core Semantic Backbone, Operational Views, and Governance and Resilience Views. These views are built on a Shared Semantic Foundation and feed into an AI-Driven Analytics and Decision Support Layer. The paper relates the architecture to the NARRATE pilots and illustrates its application through a concrete scenario.
Maurizio Griva presented the paper “Integrating Sustainability KPIs, Resilience Stress Testing and LCA-Based Assessment for Sustainable Manufacturing Decision Support”. He introduced the Resilience, Sustainability and Circularity Stress Testing Tool (RSCST) as the NARRATE component for running “what-if” scenarios against manufacturing network data represented in the Blueprint and Digital Twin. Where required, the tool delegates life-cycle assessment (LCA) calculations to an OpenLCA-based service. This allows the impact of alternative decisions on resilience, sustainability, and circularity to be assessed before they are implemented.
The paper “From Global News to Resilience: An Explainable Early Warning Architecture for Smart Manufacturing Networks” could not be presented because Tarmo Calvet was unable to attend due to illness. The paper proposes a pipeline that transforms GDELT (Global Database of Events, Language, and Tone) news data into manufacturing-relevant disruption alerts.
Raw GDELT records, ranging from approximately 530,000 to 1.95 million per observation window, are reduced through thematic filtering, optional keyword filtering, non-AI pre-scoring, and tone-based prioritization to roughly 1,000 economically relevant articles per month. These articles are then enriched, summarized, and contextualized for specific firms using an LLM, mapped to Harmonized System product codes, scored, and published as structured JSON alerts via MQTT to the IMC, dashboards, and blueprint management components.
The MAASive contribution, entitled “Toward Resilience-Oriented Decision Support for Manufacturing-as-a-Service: A Conceptual Scoreboard”, is a conceptual paper developed using Design Science Research Methodology. It proposes a “Resilience Scoreboard” for Manufacturing-as-a-Service (MaaS) networks. The scoreboard aggregates 37 resilience-related KPIs into six capability scores: Flexibility, Redundancy, Velocity, Visibility, Awareness & Alertness, and Collaboration. It also links these capabilities to 18 risk sources, indicating which capabilities are relevant to a given disruption and which KPIs can be used to assess them.
The workshop concluded with a panel discussion involving the presenters and experts working on related topics. The workshop chair raised several questions for discussion:
- Where do resilience measures such as buffer stocks, multi-sourcing, and redundancy directly conflict with lean and efficiency targets, and how are these trade-offs decided in practice?
- Are current Digital Twin implementations mainly descriptive and monitoring-oriented, or do they already provide genuinely predictive and prescriptive support for disruption response?
- Which collaboration formats, such as co-located teams, shared KPIs, or joint governance, have worked best in NARRATE?
- How do the presented approaches deal with incomplete, siloed, or heterogeneous data across supply chain partners?
- What is the single most transferable lesson from NARRATE and MAASive for manufacturers that are not yet involved in these projects?
The panel’s answer to the last question was a clear call for manufacturers to seriously consider how AI can be applied within their own organizations. The discussion emphasized that AI should not be viewed as an abstract future technology, but as a practical tool that can support resilience, decision-making, and the management of complex manufacturing networks.
The papers will be published in the Springer Lecture Notes series. In addition, the editors of the Journal of Integrated Design & Process Science plan to publish extended versions of selected papers in a special issue.
Authors
