Prediction and compensation of subsequent deformation in robotbased incremental sheet forming by application of machine learning

• Applicant: Professor Dr.-Ing. Bernd Kuhlenkötter
• Subject Area: Primary Shaping and Reshaping Technology, Additive Manufacturing
• Term: since 2021
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 457407945

Incremental sheet forming is a flexible, workpiece-independent process for manufacturing sheet metal parts in small lot sizes. An industrial application has yet rare or not all taken place due to the still low geometric accuracy of the process. The reason for this is mainly the missing possibility for a precise simulation of the forming process, hindering the use of compensation approaches for springback and subsequent deformation. While there are multiple FEM-based simulation approaches, their application is prevented by summing up simulation errors, caused by the incremental nature of the process. During the first funding period of the research project, a data driven approach was pursued by the usage of machine learning, which, in contrast to FEM-simulations, does not need a detailed modelling of the forming process. A multi-layer artificial neural network was build up for predicting the resulting geometric accuracy of a forming experiment based on common process parameters, part geometry and the course of the tool path. The mean absolute forming error was reduced by up to 68 percent thanks to a prediction-based adjustment of the tool path. These results will be built upon in the second funding period in order to further increase the accuracy of the prediction of forming accuracy. To this end, gaps in the process database are first identified using a cluster analysis. Based on these, new components are derived, formed and added to the process database. Several artificial neural networks are then trained based on the geometry representations developed in the first funding period and their predictions are combined into a joint prediction using ensemble learning. The artificial neural networks are also used to predict the optimal combination of the two unique parameters of roboforming, the support angle and the support force, and to adapt these dynamically during the forming process. This requires an extension of the process control to prevent oscillations of the system. Finally, a further artificial neural network is trained, which predicts the increase in forming accuracy through a prediction-based adaptation of the tool path and a dynamic adaptation of the process parameters and thus the achievable forming accuracy. If the achieved improvement of the forming accuracy is high enough, transfer learning shall be used during a project extension to extend the prediction qualities of the artificial neural networks to new combinations of sheet materials and thicknesses.

• DFG Programme: Research Grants