Thermal management is becoming increasingly important for modern production technology, as climatic conditions cause significant temperature fluctuations that affect production accuracy. Established cooling systems are highly energy-intensive, making them unsustainable. This paper presents the concept of an autonomously switching thermal bridge for passive thermal regulation, based on a paraffin expansion actuator. The system exploits the reversible phase transition to dynamically regulate thermal conductivity without requiring external auxiliary power or active control. Numerical simulations were used to develop the basic principles for utilising the membrane actuators as a thermal bridge. The subsequently realised thermal bridge was tested experimentally in various temperature ranges. It demonstrated good thermal conductivity and autonomous switching capability. The concept is also designed for robustness and production scalability. It therefore holds promise for industrial applications in the field of passive heat management in production environments.

The transition towards sustainable packaging requires reliable forming processes for paperboard, but its anisotropic and hygroscopic nature strongly limits dimensional accuracy in processes such as deep drawing. This study addresses the aforementioned challenge by systematically investigating two complementary strategies: optimizing blank geometry and introducing pretension. A combination of numerical simulations with anisotropic, moisture-dependent plasticity, and experimental validation using a pneumatic press with additively manufactured tools was applied. The base-point method for blank optimization allowed for efficient reduction of flange length deviations and geometric errors by more than 55% in a first iteration and stable convergence within three optimization steps. Pretension strategies, applied either by mechanical pre-stretching or by exploiting hygroexpansion, also reduced anisotropic springback. Hygroexpansion-based pretension proved particularly effective by achieving more homogeneous stress distributions without additional equipment. The results demonstrated that these strategies can reduce springback and increase drawing depth while providing a reproducible approach. Optimized blank geometry ensures a more uniform distribution of blank-holder force, while pretension counteracts anisotropy-induced recovery. Together, these findings provide a pathway toward more accurate and scalable paperboard deep drawing, with relevance for industrial implementation of sustainable three-dimensional packaging.

Abstract Das Rundkneten ist ein energie- und materialeffizientes Verfahren der Kaltmassivumformung, dessen Prozessauslegung durch schwer zugängliche Einflussgrößen wie den real wirksamen Drehvorschubwinkel erschwert wird. Drehschlupf verursacht Abweichungen zwischen Soll- und Ist-Parametern und beeinflusst damit die Bauteilqualität. Dabei entstehen charakteristische Fehler wie die Spiralbildung. In dieser Arbeit werden FEM-Simulationen mit experimentellen Untersuchungen gekoppelt, um diese versteckten Prozessgrößen zu identifizieren und das simulative Modell zu validieren. Die Ergebnisse liefern Schwellenwerte und Toleranzbereiche und bilden die Grundlage für adaptive Strategien zur Prozessführung und sensorintegrierte Messsysteme zur Verbesserung von Prozessverständnis und Bauteilqualität.

The predictive power of machine learning models for process monitoring in sheet metal forming depends strongly on the information content of the sensor signals. This study investigates how force signal characteristics represent process conditions in a multi-stage forming process consisting of deep drawing and ironing, in which surface roughness evolves with a downstream tendency. Indirect and direct force measurement concepts are compared: While indirect sensors are prone to noise, direct sensors often show more clarity. Neural networks are trained on datasets covering multiple roughness and process configurations. Model performance is analysed using classification metrics and explainable AI methods. The results reveal a counter-intuitive finding: Visually smooth force signals with high signal-to-noise ratio can provide limited or misleading information for convolutional neural networks due to temporal misalignment, whereas noisier signals with distributed dynamics show more robust predictions. The study shows the influence of signal clarity for data-driven process monitoring in Industry 4.0-enabled forming.

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.

Progressive die forming is a highly productive method in sheet metal forming, enabling the manufacturing of complex components through multiple sequential forming stages. However, forming processes are typically influenced by a multitude of sources of uncertainty, including variations in the properties of semi-finished products and tool wear. This often results in fluctuating product properties and unstable forming processes. Currently, compensating for these variations requires extensive knowledge from experienced operators. Automated adaptation to varying process conditions presents multiple challenges: In the process of selecting and integrating sensors, it is essential to ensure that relevant process variables are being measured, which is typically associated with proximity to the forming zone. Conversely, the process should be unaffected by sensor integration. Additionally, in progressive dies, signals from multiple stages can interfere with each other. Therefore, the objective of this work is to investigate which sensor technology, in combination with data-driven models, is best suited for closed-loop control of forming processes in progressive dies. The present study proposes a methodology for the control of part angles in an adaptive bending stage. The actuator for the adaptive bending stage is a combination of motor and wedge gear, allowing for continuous adjustment of the bending angle. To model the influence of uncertainty in metal forming, the process is deliberately disturbed by varying the properties of the semi-finished product. Therefore, different types of sensors are integrated into the tool system and used in combination with machine learning models to control the part properties, with the aim of reducing the dependence on manual expertise.

This study investigates an adaptive die concept for cold extrusion that actively modulates radial preload during the main forming and ejection phases. A Gaussian process regression (GPR) surrogate, trained on fewer than 400 finite-element simulations, provides a highly data-efficient model capable of accurately predicting geometric tolerances, residual stresses, and process forces. Experimental spot measurements validate the physical trends captured by the surrogate, demonstrating reliable reproduction of the underlying mechanical interactions. The results show that increased preload during forming enables micrometer-level calibration of final diameters, while higher preload during ejection promotes beneficial compressive residual stresses at the cost of elevated ejector forces. A part-to-part control strategy effectively improves accuracy by independently steering two target properties through separate preload adjustments. Furthermore, a reinforcement learning-based controller, enhanced by flow stress estimates derived from hardness measurements, reduces variance and compensates for stochastic fluctuations in material and friction conditions. Overall, the adaptive die system, combined with surrogate-and RL-based control provides a robust foundation for achieving high dimensional precision and stable product properties under future variability scenarios, such as green steel and sustainable lubrication systems.

Roll forming is a sheet metal forming process characterised by high production rates and high material utilisation. Process-inherent inhomogeneous longitudinal strain distribution across the profiles cross-section requires the use of straighteners to maintain product quality within specifications during manufacturing. The straightening process is adjusted iteratively based on the expertise of line operators. A shortage of skilled workers leads to a loss of experience-based know-how in roll forming operations. To mitigate the effects on process productivity, machine learning (ML) can be applied to provide assistance to less experienced operators. In this context, force and position signals are recorded on a sensorically equipped straightener in order to predict corrective adjustments for line operators using convolutional neural networks. Furthermore, the integration of numerically generated data is investigated to reduce the required amount of labelled experimental data. Applying a Transfer Learning (TL) approach, the incorporation of numerical data reduces the mean error by 20.81% and the mean standard deviation by 36.62% for small experimental datasets.

The integration of sensor technology represents a promising approach for the reliable and cost-effective real-time monitoring of production processes in sheet metal forming. The combination of time series data with Machine Learning algorithms enables the approximation of complex process states. In particular, deep neural networks have demonstrated remarkable performance under laboratory conditions for the monitoring of blanking processes. However, in real-world production environments, the model performance deteriorates due to pervasive environmental factors that induce uncertainty to the underlying data distributions. Therefore, this work presents investigations that show the impact of different sensor modalities (force, acoustic emissions, acceleration) on model accuracy under uncertain manufacturing conditions. Subsequently, the Robust Bayesian Hyperparameter-Optimization, a robustness-focused, novel adaptation of the Bayesian Hyperparameter Optimization is proposed. The evaluation of the signal-specific model structures highlights acceleration signals as the most robust input. Furthermore, the proposed Robust Bayesian Hyperparameter-Optimization approach can increase robustness across all datasets, including a maximum improvement of over 20

In response to the growing demands for both sustainability and precision in metal forming, this study investigates the potential of an adaptive die system for cold forging processes. The system allows of control of die preload during the main forming and ejection phases, thus offering two degrees of freedom to influence product properties. Through a combination of experimental and numerical investigations, the interdependence between the final part diameter, axial residual stresses, and ejection forces is systematically analyzed. It is shown that increasing the preload during forming reduces the final diameter. Conversely, preload applied during ejection has a direct influence on the resulting ejection force and surface stresses. This decoupling capability enables targeted tuning of individual product properties. To experimentally represent the variability of material batches, three different steel grades were selected, spanning a broad range of flow stresses. The resulting process maps reveal how fluctuations in material properties affect forming outcomes, and how the adaptive die system can be used to compensate these effects. The experimental trends were confirmed by finite element simulations, which support the physical interpretation of preload-related elastic and plastic interactions within the tooling system. The study shows that adjusting the preload intelligently enables dimensional corrections and residual stress or ejection force optimization. The primary focus is on understanding and modeling the process-property relationships. The results lay the foundation for potential control strategies, such as inline or part-to-part adaptation. These strategies can be integrated into future forming lines for increased robustness and flexibility.

Friction modeling in finite element simulations of cold forging is commonly based on analytical formulations ranging from the Coulomb and shear friction laws to advanced models incorporating multiple tribological state variables. However, fundamentally different friction formulations are difficult to compare directly because their parameters refer to different physical quantities. This work presents a systematic benchmark of conventional, extended analytical and machine-learning-based friction models under identical tribological boundary conditions. Sliding compression tests covering a wide range of contact conditions are combined with finite element simulations to establish a time-resolved database of local tribological state variables. To enable a physically consistent and model-independent comparison, all models are evaluated on the common level of frictional shear stress. Conventional friction models exhibit pronounced load-dependent residual structures, while extended analytical formulations reduce but do not eliminate systematic deviations. The neural-network-based formulations introduced in this study, predicting either the coefficient of friction, the friction factor or the frictional shear stress directly, achieve the highest predictive accuracy, with nMAE values of 0.85–0.96% relative to the mean flow stress. Despite their different internal parameterizations, these models converge towards highly similar frictional shear stress predictions. Explainability analyses further show that similar predictive behavior on the stress level is achieved through formulation-dependent relationships on the parameter level. The results therefore reveal a distinction between stress-level convergence and parameter-level divergence in data-driven friction modeling.

Many promising incremental forming processes have been proposed in the scientific literature, but only few have found their path to industrial use. The main reason for the disappointing success rate is given by the failure to comply with geometrical requirements. The paper at hand proposes a stepwise approach to enhance the geometrical properties and thereby potential usability of an incremental sheet forming process. The mode of deformation and the effectiveness of the compensation are identified as key aspects. The path from feasibility to usability is exemplified by the novel incremental forming process hole-rolling. During this process, a roller forms a one- or double-sided collar around a hole in a spiral movement, where the contour of the formed part can be circular or an individual geometry, like a polygon shape. The process characteristics are analysed using two different forming tools, which demonstrate different possible process routes. A model-based compensation method, which is based on forming force and an analytical model of the forming tool, is developed and verified using two different geometries, thereby demonstrating its usability.

This review examines how self-adapting principles can be transferred to metal forming technology. It defines self-adapting forming systems as systems that improve process performance through intrinsic physical adaptation mechanisms rather than externally imposed control. Based on concepts from self-engineering systems and soft robotics, a taxonomy is derived that distinguishes adaptation according to system level, energy supply, actuation principle and location of the adaptive effect. Representative examples from bulk and sheet metal forming are analysed, including spring-attached dies in forging, floating dies in bending, adaptive flash gaps in closed-die forging, and self-adjusting blank holder concepts in deep drawing. The reviewed examples indicate that physically embedded adaptation mechanisms can contribute to improved robustness, reduced forming loads and lower control complexity in selected forming processes. Finally, prospective concepts such as smart materials, programmable lubricants and mechanically self-adjusting tool systems are discussed as potential routes for future self-adapting forming systems.

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.

A new process called flexible T-profile rolling has been developed to produce T-profiles with different loadadapted thicknesses along their length. This process provides an environmentally friendly and resourceefficient production method for e.g. aluminium aircraft stringers. This paper outlines the fundamentals of designing the roll stand, the rolls themselves, and the roll movement curves, which are based on preliminary numerical investigations. The new process is then validated experimentally for the first time. It demonstrates that target thicknesses and curves can be effectively achieved through flexible T-profile rolling with multiple passes and thickness transitions. The thickness of the T-profile was reduced by 50%, from 3 mm to 1.5 mm, across the entire cross-section in three rolling passes. Experiments and simulations of the new rolling process show good agreement. Furthermore, the hardness properties achieved in the material are homogeneous across the cross-section of the rolled profiles. However, despite the use of a straightening roll after the rolling gap, the profile still exhibited a small vertical bow due to the inhomogeneous distribution of longitudinal strain.

The uniaxial tensile test is a common and fundamental test in materials science and engineering, in which a specimen is subjected to controlled tension until failure. From this, the stress-strain curve and many property parameters of the material can be calculated, such as tensile strength, ultimate strength, maximum elongation, Young's modulus, Poisson's ratio, and yield strength. As fibrous materials, such as paper and paperboard, become more popular, accurately measuring their mechanical properties becomes essential for developing and applying these materials, especially in packaging. However, since they are anisotropic and inherently inhomogeneous due to the arrangement of the fibers, accurately determining their mechanical properties is not straightforward. This study investigated how several key factors influence the results of tensile tests on fiber-based materials: sample size and deformation measurement techniques using three fiber materials. This study also compared three different strain recording methods: digital image correlation (DIC), video extensometer, and conventional extensometer (Traverse). The DIC technique emphasized the effect of the inherent inhomogeneity of the paperboard on the overall mechanical properties obtained from tensile tests. The results indicated that sample size has a negligible effect on the stress-strain curve, and any apparent influence likely stems from slip at the grips during tensile testing. However, sample size does affect paperboard fracture to some extent. The study also provided recommendations for optimal specimen geometry and deformation recording methods to improve the accuracy and repeatability of tensile testing of fiber-based materials.

In progressive die deep drawing, small deviations in strip positioning and transport can accumulate across stages, leading to significant geometric variations and quality issues. Reliable datasets are therefore essential for data-driven modeling, yet process labels such as the nominal feed length are often affected by machine tolerances and thus introduce uncertainty. This work presents a camera-based methodology for automated feature extraction at the end of the deep drawing process. Using an inline profile projector, relevant geometric features such as circle distances, centroid positions, and local material widths are extracted from high-speed image data. The implemented algorithm combines edge detection, circle fitting, centroid analysis, and directional width measurements to generate consistent feature sets under industrial conditions. Challenges such as material reflections, asymmetric deformation, and non-ideal boundary conditions are addressed, and geometry-specific obstacles for feature detection are discussed. The extracted features are further analyzed with respect to their statistical variability across systematically varied feed length values. Results demonstrate that the variance of image-based features increases with higher feed length, indicating that nominal process labels do not always reflect the actual material position. The study highlights the importance of uncertainty quantification in dataset generation for machine learning in forming technology. By linking process variation to directly observable geometry features, the proposed approach provides both methodological guidance for feature extraction and conceptual insights into label quality in industrial datasets.

The roll forming process generally requires limited adjustments by the operator if the process conditions remain stable. However, in industrial environments, material parameters fluctuate more often; as a result, the product shape quality can vary, and this requires roll position adjustment to bring the product quality back to specification. In industrial production, the incoming material properties and geometric dimensions are usually not monitored continuously, and the influence of real material variation on the product’s quality is yet to be investigated. In this work, a test routine is implemented that continuously monitors the material properties and the sheet thickness of the incoming strip in an industrial roll forming line. This is combined with an inline measurement of the roll forming load and part shape quality. The experimental analysis is complemented with finite element analysis of the process using the commercial software package Copra RF/FEA to generate a dataset with a high variability of equally distributed information. The relationship between the real fluctuating material yield strength and sheet thickness is correlated with roll load and final product shape quality. The acquired dataset is then used to develop a data-driven model for inline shape quality prediction.

Deep drawing of paperboard enables highly productive forming processes for packaging products made of a recyclable material. However, the inherent anisotropy and low elasticity of paperboard pose challenges for deep drawing processes, especially resulting in direction-dependent springback and wrinkling of the formed parts. This paper presents an approach that addresses these challenges by using a segmented blank holder. The goal is to improve component quality both locally and globally by applying different blank holder forces to each segment. To this end, a concept is presented for a pneumatically driven segmented blank holder, along with two different forming geometries. Deep-drawing tests with segment-specific blank holder force distributions were performed and compared with those using a conventional blank holder. Although segmentation of the blank holder did not improve performance when the force per segment was identical, an uneven force distribution was able to improve forming in terms of shape accuracy and wrinkle pattern. Depending on the selected force distribution, significant compensation for springback anisotropy and the creation of wrinkle-free straight areas was shown to be possible.

Although cold forging enables the high-volume production of precision components, extrusion dies are exposed to severe cyclic loading. This promotes fatigue damage and dimensional scatter. Conventional shrink-fitted reinforcements increase tool life by prestressing the die in compression. However, they offer limited flexibility to compensate for variations in material and tribological conditions. Fully adaptive dies can tune external pressure during operation, yet they often require high actuation forces and highly loaded sliding interfaces. To leverage the key advantages of both reinforcement strategies, this paper introduces a hybrid passive–active die reinforcement concept. A passive shrink fit provides a robust baseline prestress, while an active stage superimposes an adjustable external preload. The actuation is realized by a wedge mechanism, so the actuator delivers only an incremental load relative to the passive baseline. A purely analytical dimensioning method is derived for this concept. It couples the wedge model with a Lamé-based compound-cylinder model of a double-ring reinforcement including interference-fit compatibility. The design is optimized across multiple load scenarios. The objectives are compressive die hoop stresses, a minimum characteristic safety factor, and maximum controllable bore diameter change. A deterministic grid search over the interference fits generates contour maps that highlight feasible regions and the trade-off between structural margin and adjustability. Representative design points are then evaluated by spot finite element simulations of a forward extrusion sequence. Die hoop stresses, ejector force, and final part diameter are used as evaluation metrics. The results provide a transparent and computationally efficient framework for selecting geometry, interference fits, and actuation settings. The procedure is transferable to industrial applications that require both structural safety and adjustable compliance.