Daria Gelbich
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Daria Gelbich has a total of 33 co-authors. Here are the top co-authors by number of shared publications:
- Philipp Niemietz
Manufacturing Technology Institute (MTI), RWTH Aachen University, Aachen, Germany
5 shared publications - Thomas Bergs
Manufacturing Technology Institute—MTI of RWTH Aachen University, Campus-Boulevard 30, 52074 Aachen, Germany
5 shared publications - Jonas Moske
Institute for Production Engineering and Forming Machines (PtU), Technical University of Darmstadt, Darmstadt, Germany
3 shared publications - Markus Schumann
Institute for Production Engineering and Forming Machines (PtU), Technical University of Darmstadt, Darmstadt, Germany
3 shared publications - Antonia Wüst
AIML Lab, Technical University Darmstadt, Darmstadt, Germany
2 shared publications - Felix Divo
AIML Lab, Technical University Darmstadt, Darmstadt, Germany
2 shared publications
Latest Publications
- Open Access
- 13 May 2026
- Published by Springer Nature in Production Engineering
- Vol. 20 (2) , 68
Concept-based force-signature analysis of tool-parameter effects in fine blanking
Fine blanking part quality is highly sensitive to tool parameterization, yet cause–effect relationships are obscured by substantial process noise. This limits the utility of both simulation and scalar process metrics for tool optimization and mechanistic understanding. At the same time, high-resolution force-signature data is information-rich and can be used as evidence for learning physically plausible relationships, provided that interpretability is retained. This paper presents a data-to-knowledge pipeline combining finite element method simulation, neural network prediction, and concept extraction to derive interpretable patterns from force time-series and formulate mechanistic hypotheses linking tool parameters to die-roll formation. Controlled tool-parameter variation generates approximately 27,000 strokes with die-roll metrology. Neural networks predict die-roll height and tool parameters with mean absolute percentage errors of 0.6% and 2.6%, respectively. Concept extraction (ECLAD-ts) and segment-aware attribution localize predictive information to peak-load and post-peak regimes. For die-clearance variation, piecewise-linear analysis within concept windows identifies two mechanism-sensitive slope descriptors (Cliff’s $$|\delta | \ge 0.83$$ ) with ordered progressions mirroring the graded die-roll response. The rising-flank slope captures clearance-dependent shearing resistance, while the post-peak decay slope reflects thermo-mechanical softening, reducing flow stress and accelerating the post-peak force decay. Smaller clearance yields higher resistance, faster decay, and reduced die-roll height, consistent with forming theory. The pipeline enables phase-local assessment of tool-parameter effects under process noise and supports testable mechanistic hypotheses for simulation refinement.- 13 May 2026
- Published by Springer Nature in Production Engineering
- Vol. 20 (2) , 68
Open AccessConcept-based force-signature analysis of tool-parameter effects in fine blanking
Fine blanking part quality is highly sensitive to tool parameterization, yet cause–effect relationships are obscured by substantial process noise. This limits the utility of both simulation and scalar process metrics for tool optimization and mechanistic understanding. At the same time, high-resolution force-signature data is information-rich and can be used as evidence for learning physically plausible relationships, provided that interpretability is retained. This paper presents a data-to-knowledge pipeline combining finite element method simulation, neural network prediction, and concept extraction to derive interpretable patterns from force time-series and formulate mechanistic hypotheses linking tool parameters to die-roll formation. Controlled tool-parameter variation generates approximately 27,000 strokes with die-roll metrology. Neural networks predict die-roll height and tool parameters with mean absolute percentage errors of 0.6% and 2.6%, respectively. Concept extraction (ECLAD-ts) and segment-aware attribution localize predictive information to peak-load and post-peak regimes. For die-clearance variation, piecewise-linear analysis within concept windows identifies two mechanism-sensitive slope descriptors (Cliff’s $$|\delta | \ge 0.83$$ ) with ordered progressions mirroring the graded die-roll response. The rising-flank slope captures clearance-dependent shearing resistance, while the post-peak decay slope reflects thermo-mechanical softening, reducing flow stress and accelerating the post-peak force decay. Smaller clearance yields higher resistance, faster decay, and reduced die-roll height, consistent with forming theory. The pipeline enables phase-local assessment of tool-parameter effects under process noise and supports testable mechanistic hypotheses for simulation refinement. - Open Access
- 7 May 2026
- Published by Springer Nature in Production Engineering
- Vol. 20 (2) , 66
Structured representation of simulation and annotation data for machine learning in forming technologies
The use of machine learning (ML) in manufacturing requires structured, especially standardized, access to both simulation data and domain knowledge. This paper introduces a JSON-based data format for representing synthetic force-time series alongside expert annotations. The schema captures simulation metadata, tool and material parameters, and allows explicit expert knowledge, such as failure indicators, to be linked to signal segments. The proposed structure enables process-aware ML methods that leverage both domain knowledge and raw data for improved learning and generalization. A deep drawing use case illustrates how the format facilitates knowledge-guided learning. The approach aims to bridge the gap between real and simulated production data, supporting scalable integration in modern manufacturing systems.
- 7 May 2026
- Published by Springer Nature in Production Engineering
- Vol. 20 (2) , 66
Open AccessStructured representation of simulation and annotation data for machine learning in forming technologies
The use of machine learning (ML) in manufacturing requires structured, especially standardized, access to both simulation data and domain knowledge. This paper introduces a JSON-based data format for representing synthetic force-time series alongside expert annotations. The schema captures simulation metadata, tool and material parameters, and allows explicit expert knowledge, such as failure indicators, to be linked to signal segments. The proposed structure enables process-aware ML methods that leverage both domain knowledge and raw data for improved learning and generalization. A deep drawing use case illustrates how the format facilitates knowledge-guided learning. The approach aims to bridge the gap between real and simulated production data, supporting scalable integration in modern manufacturing systems.
- Open Access
- 16 April 2026
- Published by Trans Tech Publications, Ltd. in Solid State Phenomena
- Vol. 389, 305-318
Hybrid Analysis of Tool Geometry and Process Noise in Fine Blanking through Simulation and Experiment
In fine blanking, a mass production process for safety-critical components, a discrepancy exists between deterministic modeling approaches and the stochastic nature of real-world measurements, often termed process noise. This work combines Finite Element Method simulations with data from industrial-scale fine blanking experiments, featuring long stroke series across multiple coils with systematically varied die clearances. The analysis shows a strong correlation between force curve characteristics and the formation of tears, a relationship that holds across all tested geometries. In contrast, only weak and inconsistent correlations were found between the force signal and the resulting die roll. This weakness is explained by the finding that multiple physical effects, such as material strength and friction, have competing influences on die roll that are not separable in the single force signal. These results demonstrate that the utility of the force signal for quality prediction is highly dependent on the tool geometry, providing a basis for more reliable tool design strategies in fine blanking.
- 16 April 2026
- Published by Trans Tech Publications, Ltd. in Solid State Phenomena
- Vol. 389, 305-318
Open AccessHybrid Analysis of Tool Geometry and Process Noise in Fine Blanking through Simulation and Experiment
In fine blanking, a mass production process for safety-critical components, a discrepancy exists between deterministic modeling approaches and the stochastic nature of real-world measurements, often termed process noise. This work combines Finite Element Method simulations with data from industrial-scale fine blanking experiments, featuring long stroke series across multiple coils with systematically varied die clearances. The analysis shows a strong correlation between force curve characteristics and the formation of tears, a relationship that holds across all tested geometries. In contrast, only weak and inconsistent correlations were found between the force signal and the resulting die roll. This weakness is explained by the finding that multiple physical effects, such as material strength and friction, have competing influences on die roll that are not separable in the single force signal. These results demonstrate that the utility of the force signal for quality prediction is highly dependent on the tool geometry, providing a basis for more reliable tool design strategies in fine blanking.
- Open Access
- 14 January 2026
- Published by Springer Nature in Research Square
Towards hybrid and explainable data-driven models for forming processes: addressing the gap between simulation, process data, and model validation
Finite element method simulations are foundational to forming process design, but their utility for machine learning is often limited by idealizations that create a gap with stochastic, real-world manufacturing conditions. This paper investigates the systematic integration of simulation-based knowledge into machine learning pipelines to bridge this gap. A structured review of 56 publications reveals systemic limitations in current research, which we classify as the reality, validation, and trust gaps. In response, this work formulates four research questions to address these gaps, examining them through use cases in forging, blanking, and deep drawing. The analysis demonstrates that the impact of simulation simplifications is highly task-dependent, requiring a tailored approach to model selection and data integration. This paper contributes to the methodological discourse by proposing a framework for simulation-informed machine learning. It argues for a shift in focus from maximizing physical realism to a “fitfor-purpose” approach, where simulation complexity is strategically aligned with the requirements of the downstream machine learning task. The paper outlines a path toward more robust, interpretable, and industrially transferable modeling strategies in forming technology.
- 14 January 2026
- Published by Springer Nature in Research Square
Open AccessTowards hybrid and explainable data-driven models for forming processes: addressing the gap between simulation, process data, and model validation
Finite element method simulations are foundational to forming process design, but their utility for machine learning is often limited by idealizations that create a gap with stochastic, real-world manufacturing conditions. This paper investigates the systematic integration of simulation-based knowledge into machine learning pipelines to bridge this gap. A structured review of 56 publications reveals systemic limitations in current research, which we classify as the reality, validation, and trust gaps. In response, this work formulates four research questions to address these gaps, examining them through use cases in forging, blanking, and deep drawing. The analysis demonstrates that the impact of simulation simplifications is highly task-dependent, requiring a tailored approach to model selection and data integration. This paper contributes to the methodological discourse by proposing a framework for simulation-informed machine learning. It argues for a shift in focus from maximizing physical realism to a “fitfor-purpose” approach, where simulation complexity is strategically aligned with the requirements of the downstream machine learning task. The paper outlines a path toward more robust, interpretable, and industrially transferable modeling strategies in forming technology.
- Open Access
- 13 January 2026
- Published by Springer Nature in Research Square
Structured Representation of Simulation and Annotation Data for Machine Learning in Forming Technologies
The use of machine learning (ML) in manufacturing requires structured, especially standardized, access to both simulation data and domain knowledge. This paper introduces a JSON-based data format for representing synthetic force-time series alongside expert annotations. The schema captures simulation metadata, tool and material parameters, and allows explicit expert knowledge, such as failure indicators, to be linked to signal segments.The proposed structure enables process-aware ML methods that leverage both domain knowledge and raw data for improved learning and generalization. A deep drawing use case illustrates how the format facilitates knowledge-guided learning. The approach aims to bridge the gap between real and simulated production data, supporting scalable integration in modern manufacturing systems.
- 13 January 2026
- Published by Springer Nature in Research Square
Open AccessStructured Representation of Simulation and Annotation Data for Machine Learning in Forming Technologies
The use of machine learning (ML) in manufacturing requires structured, especially standardized, access to both simulation data and domain knowledge. This paper introduces a JSON-based data format for representing synthetic force-time series alongside expert annotations. The schema captures simulation metadata, tool and material parameters, and allows explicit expert knowledge, such as failure indicators, to be linked to signal segments.The proposed structure enables process-aware ML methods that leverage both domain knowledge and raw data for improved learning and generalization. A deep drawing use case illustrates how the format facilitates knowledge-guided learning. The approach aims to bridge the gap between real and simulated production data, supporting scalable integration in modern manufacturing systems.