Information Content Analysis of Direct and Indirect Force Measurements for Machine Learning-Based Process State Classification in Multi-stage Sheet Metal Forming

The predictive power of machine learning mod- els for process monitoring in sheet metal forming depends strongly on the information content of the sensor signals. This study investigates how force signal characteristics rep- resent 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 Metallurgy Materials Engineering Markus Schumann and Jonas Moske have contributed equally to this work. * Markus Schumann markus.schumann@ptu.tu-darmstadt.de Jonas Moske jonas.moske@ptu.tu-darmstadt.de Felix Divo felix.divo@tu-darmstadt.de Antonia Wüst antonia.wuest@tu-darmstadt.de Kristian Kersting kristian.kersting@tu-darmstadt.de Peter Groche peter.groche@ptu.tu-darmstadt.de 1 Institute for Production Engineering and Forming Machines (PtU), Technical University of Darmstadt, Otto-Berndt-Straße 2, 64287 Darmstadt, Hessen, Germany 2 AIML Lab, Technical University Darmstadt, Hochschulstraße 1, 64289 Darmstadt, Germany 3 Hessian Center for AI (Hessian.AI), Technical University Darmstadt, Karolinenplatz 5, 64289 Darmstadt, Germany 4 German Research Center for AI (DFKI), Landwehrstraße 50A, 64293 Darmstadt, Germany Trans Indian Inst Met (2026) 79:140 140 Page 2 of 13 This paper evaluates which force-sensing concepts and process phases provide high discriminative power, detect - ing die roughness states in a multi-stage process via neural networks. By combining a factorial experimental study with machine learning-based signal analysis, the classification performance is quantified and informative signal regions are identified. Contributions: • A systematic investigation of direct and indirect force signals under defined roughness states. • A comparative evaluation of sensor concepts regarding their discriminative capability. • An explainable machine learning analysis identifying informative process phases. • Guidance on sensor selection for robust surface condition monitoring. 2 Experimental Setup 2.1 Process and To