DMaaST: Cognitive Digital Twins for Resilient and Sustainable Manufacturing
Karlsruhe Institute of Technology (KIT) is part of the DMaaST European consortium and is leading WP3 of the DMaaST project, which focuses on the development of novel methodologies and the delivery of cognitive Digital Twins in the domain of smart manufacturing.
The manufacturing industry is experiencing increasing system complexity due to growing product customizations, tighter integration of technologies, dynamic demand patterns and strong interdependencies across manufacturing systems and value networks. Within the DMaaST project, the objective is to support European manufacturing industries in achieving resilient manufacturing value networks by leveraging Digital Twins, enabling digitalization, and fostering sustainability within Manufacturing as a Service (MaaS) paradigm.
Digital Twins provide continuously updated digital representations of physical systems, integrating IoT data in (near) real-time and enabling bidirectional interaction between physical and digital environments. This supports monitoring, simulation, analysis, and optimization, leading to better-informed decision making, reduced costs and lead times, and enhanced operational efficiency, reliability and resilience.
A key objective of WP3 is the development of a manufacturing ecosystem Digital Twin together with our partner IND. This ecosystem integrates two perspectives: (1) value-chain-oriented Digital Twins, and (2) manufacturing-service-oriented Digital Twins. Combining these perspectives enables more comprehensive and context-aware models, improving decision-making across interconnected manufacturing environments.
WP3 is tightly integrated with other work packages through the Smart Manufacturing Assessment Platform (SMAP). Within SMAP, Digital Twin models, e.g., receive data from Decentralized Knowledge Graphs (WP2) and provide simulation output to multi-objective decision support (WP4), enabling end-to-end data-driven decision process.
A central contribution of WP3 is the development of methodologies for extracting Digital Twin models from both IoT data and expert knowledge. This approach extends conventional data-driven techniques by systematically integrating domain expertise into the model extraction process. Purely data-driven approaches often face limitations due to data quality issues, incomplete IoT coverage and challenges in interpretation. Integrating expert knowledge provides context-understanding and domain-specific reasoning, resulting in more accurate, robust, and explainable Digital Twin models.
This approach is referred to as data-knowledge-fused Digital Twin model extraction. To achieve the goal of the DMaaST project, WP3 develops ground-truth models, a framework for extracting cognitive Digital Twin models, a structured model extraction methodology, and algorithms for combining IoT data and expert knowledge, including automated processing of natural-language-based expert input using language models.
Overall, this work delivers a systematic and scalable methodology for the automated extraction of cognitive Digital Twin, supporting their deployment in complex manufacturing ecosystems.
Future work will focus on large-scale validation and integration of the proposed methodologies across industrial use cases.
