Robust active surface design for multi-stage sheet metal forming processes based on data- and computation-driven surrogate modelling of springback behaviour

• Applicants: Professor Dr.-Ing. Mathias Liewald; Professor Dr.-Ing. Michael Weyrich
• Subject Area: Primary Shaping and Reshaping Technology, Additive Manufacturing ; Automation, Mechatronics, Control Systems, Intelligent Technical Systems, Robotics ; Data Management, Data-Intensive Systems, Computer Science Methods in Business Informatics
• Term: since 2023
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 520459230

Springback effects cause geometrical deviations in sheet metal forming, which reduce the dimensional quality of manufactured sheet metal parts and lead to rework or scrap. For this reason, costly and time-consuming adjustments often have to be made to already finished forming tools in order to meet the part accuracy required in industrial practice. FEM simulation methods applied in tool development can significantly shorten the real tryout process and reduce costs through virtual iteration loops. Nevertheless, due to existing deviations between calculated simulation results and real manufacturing results, it is not possible to completely eliminate manual tool tryout due to a multitude of simplifications and assumptions made during modelling. In order to further reduce development cycles and tryout phases in tool making, it is therefore necessary to increase the numerical prediction quality of forming results on the one hand and to reduce the required calculation effort on the other hand. Therefore, the aim of the first funding period of the proposed project is the development of a novel design method for active die surfaces based on a data- and calculation-driven surrogate model. The data basis for this is created by active surface geometries from CAD and FE simulation results from a large number of springback calculations of multi-stage formed sheet metal components. With this data, a surrogate model is to be built using GAN or DDPM meshes, which can learn and map correlations between the active surface geometry, process parameters and the part springback. However, the springback part geometries determined by means of FE simulation always provide only an approximate calculation of reality. Therefore, a delta modelling is additionally built up based on the surrogate modelling, which is to learn the deviations between FE simulation and real forming processes by means of test data from endurance runs. Here, data of real forming processes are obtained during systematic endurance tests by means of sensor measurements and thus consider measurable process parameters and interference effects. The FE simulation thus first learns known and modelled interrelationships by machine, which are subsequently extended by experimental process data. In the second funding period, the surrogate and delta modelling will provide the starting point for a die surface generator that can generate improved modelling of die surfaces based on springback part geometries, which already includes the springback effects to be expected in reality.

• DFG Programme: Priority Programmes
• Subproject of: SPP 2422: Data-driven process modelling in metal forming technology