Work

Pit-stop manufacturing: framework extension for decision support during production ramp-up

Year: 2026
Type: dissertation
Abstract: Uncertainty, complexity, and variability are increasingly challenging production ramp-up in (re-)manufacturing systems. The shift toward circular manufacturing and the intensifying pressure for resource efficiency have exposed the limitations of traditional ramp-up strategies, especially in environments with high product variance, irregular core availability, and fluctuating operator skills. These factors demand flexible, data-driven approaches to managing ramp-up under uncertainty. This thesis addresses these challenges by proposing a novel Decision Support System (DSS) that combines graph-based decision modelling, stochastic Dynamic Programming (SDP), Digital Twins (DT), and Machine Learning (ML) to enhance decision-making during the ramp-up phase. The developed framework, called the Pit-Stop Manufacturing (PSM) extended framework, enables production planners to dynamically simulate, evaluate, and optimise ramp-up strategies in complex (re-)manufacturing environments. The system's decision-making core is a graph-based model that maps possible ramp-up trajectories across different production states. Using SDP, optimal decision paths are identified based on expected performance under uncertainty. A digital twin defines the current and future system states, integrating layout, process, and operator information in real time. Machine learning models complement the digital twin by approximating state transitions, reducing the need for exhaustive simulations and accelerating strategy testing. The developed system was validated through industrially inspired use cases, reflecting conditions typical of battery (re-)manufacturing. Results demonstrate that the DSS can reduce ramp-up time, maintain target throughput, and improve overall resilience, even in highly variable scenarios. Moreover, the modular architecture allows for easy adaptation to different production environments and data sources, supporting broader applications in flexible and circular manufacturing. By integrating simulation, data-driven modelling, and human-in-the-loop design, this thesis offers a comprehensive framework for supporting effective ramp-up decisions in uncertain (re-)manufacturing systems. The results confirm that combining digital twins, SDP, and machine learning can significantly improve ramp-up effectiveness and efficiency while reducing the dependency on expert knowledge.
Language: English
Open Access status: green
OpenAlex ID: W7226000572
Cites:
Cited by:
Related to: 10