Abstract Digital Twin technologies are emerging as powerful enablers for supporting sustainability-oriented decision-making in manufacturing systems. This presentation introduces a conceptual framework that exploits Digital Twin functionalities to monitor, assess, and improve sustainability performance throughout manufacturing processes. The framework combines data collection, virtual representation of physical assets, performance monitoring, and sustainability-related indicators to support informed decision-making and continuous improvement. Particular attention is devoted to the integration of sustainability metrics within the Digital Twin environment, enabling the analysis of resource use, process efficiency, and operational performance. The proposed architecture aims to provide a structured approach for linking digital capabilities with sustainability objectives and facilitating the adoption of data-driven strategies in industrial contexts. The presentation discusses the framework's key components, expected benefits, and potential application scenarios. As an ongoing research effort, the framework is intended to stimulate discussion and future developments on the role of Digital Twins in advancing sustainable manufacturing practices.

Abstract The present work develops process mining system discovery techniques to automatically create digital twins in manufacturing systems characterized by conveyors.

Abstract This research studies Digital Twin validation for production systems through an organised literature review across production systems and related domains. It identifies three main gaps: comparison metrics, update mechanisms, and multi-level validation. As a basis for approaches to the problem, the research proposes a graph-based production system model and an MES-driven Discrete Event Simulation architecture, which support policy-consistent behaviour and more realistic event-sequence prediction, contributing toward a validation-oriented framework for production system Digital Twins.

Abstract Industry 5.0 emphasizes human-centric, resilient, and sustainable manufacturing systems, placing human workers back at the center of production decision-making. This shift requires scheduling models to represent not only machine availability and job constraints, but also the dynamic condition of human operators. However, most existing scheduling methods still treat operators as static resources with deterministic processing times. In real shop-floor environments, operators become fatigued during work, which can affect both their processing speed and the stability of their performance. Ignoring this fatigue-driven variability may lead to hidden bottlenecks, unreliable schedule evaluation, and a planning-execution gap. In this talk, we model human processing time as a fatigue-conditioned stochastic variable, where the processing-time distribution changes according to the operator’s fatigue state. The problem is studied in a dual-resource-constrained flexible job shop scheduling problem with multitasking workers, where operators may process multiple tasks simultaneously. Building on a benchmark setting from previous research, we develop a digital-twin-based evaluation framework that incorporates fatigue accumulation, recovery, and stochastic processing-time variation. Preliminary experiments show that deterministic schedules are fragile under fatigue-driven stochastic execution, highlighting the need to evaluate human dynamic performance in more realistic settings. Further experiments indicate that simply embedding this realistic evaluation into a genetic algorithm does not necessarily improve schedule performance, suggesting that static optimization alone is insufficient. Therefore, we propose a local worker reassignment strategy for dynamic scheduling. Experimental results show that this dynamic scheduling framework improves schedule robustness and achieves better performance than the fixed benchmark schedules under fatigue-driven stochastic execution.

Abstract Injection-molding changeovers trigger production startups where parameters are re-tuned to recover “no visible defects”, generating scrap when qualification settings are reused without adaptation. This work develops a qualification-informed digital twin for startup decision support by capturing defect–parameter sensitivities and updating recommendations with vision-based feedback. Two twin instantiations trained on the same dataset are compared: an interpretable predictive model with delta-anchored tuning and a retrieval-grounded assistant reusing qualification trajectories for constrained updates. On an ABS socket-cover across hydraulic and electric machines and three viscosities, the digital-twin guidance cuts median runs-to-quality by 86% and mean startup scrap by 92% versus the industrial baseline.

Abstract The present work presents a new digital twin architecture specifically designed for wafer lot dispatching in semiconductor manufacturing systems.Acknowledgment: The work presented has received funding from the Chips Joint Undertaking (JU) under grant agreement No. 101097296. The JU receives support from the European Union’s Horizon EU research and innovation programs and Austria, Belgium, Denmark, Finland, France, Germany, Greece, Hungary, Israel, Italy, Netherlands, Romania, Sweden, Switzerland and Turkey.

Abstract Computational Fluid Dynamics (CFD) has become an indispensable tool in engineering, especially for the design of freeform surfaces. Still, the exploitation of results remains limited to two-dimensional visualization, which struggles to convey the inherently three-dimensional structure of flow phenomena. This paper presents an integrated environment for the immersive visualization of CFD data, implemented using the Unreal Engine 5 engine. The proposed framework imports static pressure and flow-velocity fields from an external solver and transfers them into the immersive Digital Shadow environment. Leveraging the Niagara particle system, the user observes flow streamlines around arbitrary geometries in real time while simultaneously retrieving accurate numerical values at selected points in the three-dimensional space, a capability absent from comparable systems. To evaluate the proposed framework, a hybrid decision support methodology was developed, combining multi-criteria decision analysis techniques with the 5 Whys method to ensure that each qualitative criterion is decomposed down to a homogeneous quantitative metric. The system was benchmarked against four representative models from the literature across binary capability, weighted scoring, and objective performance metrics. The results demonstrate the overall superiority of the proposed system, and exact quantitative data retrieval, establishing a cost-effective and accessible solution for immersive CFD analysis in mechanical design.

Abstract The transition to a circular aluminum economy is hindered by inefficient sorting of mixed alloys, which degrades material quality and drives downcycling. While Laser-Induced Breakdown Spectroscopy (LIBS) offers precise, composition-based sorting, its industrial deployment is severely limited by scrap flow instability and measurement sensitivity. This presentation introduces a high-fidelity digital twin framework designed to optimize scrap flow and enable robust LIBS sorting. The proposed architecture integrates the complex, elastoplastic geometries of real-world scrap particles alongside real-time virtual camera sensing and adaptive process control. By achieving stable particle singulation virtually, this "design via digital twin" approach can mitigate physical testing bottlenecks, paving the way for sorting high-specification recycled alloys and advancing industrial circularity.