Shear cutting: Model-based prediction of material parameters based on synthetic process force signals

Data-driven process monitoring is an approach in the field of forming technology for increasing process efficiency. In shear cutting processes surrogate models based on process force signals can be used for process monitoring. Currently, the data basis for developing such models has to be generated within experiments. The generation of synthetic training data using numerical methods seems to be a more efficient alternative approach. In this work, it is investigated whether virtual training data for the prediction of material properties can be generated by numerical methods. An FE model of the investigated shear cutting process has been designed and validated based on experiments. It is shown that especially the consideration of the tool stiffness has a significant influence on the simulated process force signal. The validated FE model is used to generate synthetic training data. Based on this data, different prediction models are trained to predict the material model parameters based on the force signals. Different model types are compared and the hyperparameters are optimized for the preferred model.