Key Engineering Materials Vol. 1049

Abstract: The stretch-reducing mill is a forming process to manufacture tubes by progressive metal deformation. This process is characterized by high complexity and a high number of variables that are strongly interconnected. To overcome the limitations and substantial simplifications of the traditional modelling and offer higher flexibility and suitability for real-time control, an Artificial Neural Networks approach is employed. By defining three parallel networks, they were predicted the milling status, the tube thickness and the angular speeds of the stands composing the process. With the models results, an optimization algorithm is employed to determine the best configuration of angular speeds of the stands to obtain a defined final tube thickness. The Artificial Neural Networks show extremely low RMSE across training, validation, and test sets, confirming their ability to model complex nonlinear dependencies. The optimisation stage reaches the target thickness with only 0.0079% error while preventing unstable operating conditions. The overall methodology provides a tool for implementing the intelligent and data-driven control of the stretch-reducing mill.

Abstract: This contribution presents a proof-of-concept and benchmark analysis between two different approaches for multiple domain modeling, employing numerical and experimental results relevant to the Radial-Axial Ring Rolling (RARR). First, a finite element method (FEM) simulation data-based Deep Neural Network (DNN) was employed as base model for a subsequent transfer learning (TL), carried out by employing 40 experimental (EXP) data belonging to a different RARR domain. The DNN-TL model is benchmarked against the proposed solution, in terms of a Proxy-Physics Informed Neural Network (P-PINN), defined using proxy equations for the outer diameter (OD) expansion, the radial force (RF), and the axial force (AF) and trained on 30 FEM and 30 EXP data. The proxy equations for the P-PINN are based on known knowledge and analytical formulations developed for the RARR. The results show that employing the whole 218 cases strong FEM database for the training of the base DNN model results in high prediction accuracy on the FEM cases and can be adapted well through TL. When employing only 30 FEM cases for the training of the base DNN model results in slight improvements for RF and in a significant drop of performance for AF, after TL. Instead, in a limited data scenario, the P-PINN model, powered equally by data and proxy equations steering the learning process, is capable of modeling well and simultaneously both FEM and EXP cases with reasonable accuracy, averaging at 3.7% for OD and 24.0% for RF/AF.

Abstract: This study proposes a Bayesian data assimilation approach to estimate material model parameters based on deformation fields measured via digital image correlation in a biaxial tensile test using a cruciform specimen. The anisotropy parameters and exponent of the Yld2000-2d yield function for a A5052P-H32 aluminum alloy are identified. The results indicate that the proposed method can estimate parameters with high accuracy—comparable to those identified via conventional multiaxial testing methods—while requiring only a single biaxial test. The proposed method offers an efficient framework for material modeling by minimizing a cost function via Bayesian optimization, enabling parameter identification from a single biaxial tensile test for sheet metal forming applications.

Abstract: Material Testing 2.0 (MT2.0) couples full‑field deformation measurements (Digital Image Correlation, DIC) with inverse identification methods (Virtual Fields Method, VFM) to extract constitutive parameters from a small number of heterogeneous experiments. This paper presents the Cut‑Clamp‑Play concept: an integrated industrial MT2.0 solution that unifies specimen design, automated testing hardware, and a computationally efficient VFM identification chain to deliver fast, user‑friendly sheet‑metal characterization. A perforated cruciform specimen is optimized for parameter identifiability of the Yld2000‑2d anisotropic yield function and used in a single biaxial test. A working prototype has been built at KU Leuven and used to collect representative DIC data; the measured displacement/strain response is double‑symmetric, confirming correct mechanical operation. Projected and early prototype results indicate that the Cut‑Clamp‑Play approach can reduce operator actions by roughly 70% and produce identification results within one hour for typical sheet‑metal cases, while further work is required to make the fully automated “Play” stage robust for industrial deployment.

Abstract: A common challenge of punch-bending is the control and compensation of process deviations which are induced, e.g., by different material batches. The resulting deviations are currently countered iteratively using expert knowledge in the design phase as well as during the production process. However, this is not always successful due to the complex interactions between semi-finished product properties and forming tools. To automate the optimization of tool geometries and process adjustments, an accurate prediction model of correlations between process/tool parameters and product properties is necessary. In that regard, an application programming interface (API) aligned with the Asset Administration Shell (AAS) specification is developed utilizing a hybrid data-driven approach. It enables the transformation of various data formats e.g., from sensor and simulation data into a machine-readable structure. The API is linked to a digital twin and a database, to automatically gather requested information and provide the new structured data to a machine learning (ML) model. To validate the developed method, hybrid punch-bending data is automatically transferred to an ML model which then returns tool geometry suggestions for a defined product geometry.

Abstract: The quality and dimensional accuracy of sheet metal components are strongly influenced by various sources of uncertainty, including variations in material properties, tool geometry and process parameters. Determining the specific source responsible for deviations in bending outcomes is usually costly and time-consuming, especially in industrial settings where numerous factors interact. In this study, a machine learning framework that can detect and quantify the impact of uncertainties in both air and bottom bending processes is presented. A dataset comprising forming results such as bending angles, final thickness and measured deviations, is used to train two neural networks metamodels (one for each process) that link input uncertainties to process outcomes. The predictive performance of these models was evaluated using different metrics achieving high predictive accuracy, with coefficients of determination close to 1 for most uncertainty sources in air bending and values above 0.95 for the majority of parameters in bottom bending. These results demonstrate the capability of the methodology to reliably identify dominant sources of uncertainty and support robust process optimization.

Abstract: Air bending is a critical operation in the metalworking industry, where dimensional accuracy and process efficiency are essential to ensure product quality and economic viability. This work proposes an AI-driven design and optimization strategy which couples artificial intelligence, specifically artificial neural networks, with a quasi-random search algorithm for the metamodeling and optimization of the air bending process. An extensive simulation database was generated by varying geometrical, material, and process parameters, and neural-network-based metamodels were trained to predict the maximum punch force, maximum thickness reduction, and final bending angle, achieving high predictive accuracy with R² values exceeding 0.96. The metamodel was subsequently used to optimize process configurations by simultaneously minimizing the maximum punch force and the maximum thickness reduction while ensuring the target bending angle, leading on average to reductions of 46.7% in maximum force and 31.5% in thickness reduction compared to non-optimized cases. The results demonstrate that artificial intelligence provides an efficient and effective tool for the design and optimization of the bending process, significantly accelerating parameter selection while improving process quality and reducing manufacturing costs.

Abstract: Numerical simulation, particularly through the Finite Element Method (FEM), is widely applied in the design and optimization of metal forming processes. However, certain real-process effects are not fully captured by numerical models alone, creating a need for complementary data-driven approaches. This study presents a hybrid modelling framework that integrates FEM simulations with machine-data-based predictive modelling to improve defect prediction in hot forging. Experimental data were collected from automated forging trials on an SMS SPPE 6.3 screw press equipped with position, force, strain, acceleration, and frame-deflection sensors. A series of trials with varied process parameters enabled the development of a data-driven classification model for detecting the “underfilling” defect. In parallel, the FEM model was refined by incorporating additional real-process phenomena, including ram tilt, press-frame deflection, and air entrapment in die cavities. The combined approach significantly enhanced defect prediction accuracy and provided deeper insight into the mechanisms driving defect formation. These results demonstrate the effectiveness of hybrid numerical–data-driven modelling for improving the robustness of metal forming process design.

Abstract: In this work, a new methodology for the identification of the CPB06 yield criterion parameters is presented. This methodology is based on the application of Machine Learning models and the Levenberg-Marquardt optimization algorithm. The proposed methodology relies on data obtained from a biaxial tensile test with a cruciform specimen and aims to overcome some of the challenges usually faced during material characterization with the CPB06 yield criterion. The predictive performances achieved were positive overall, when comparing the yield surfaces obtained for testing cases, highlighting the potential of the proposed methodology.