User Project Details

ANNdirect

A new machine learning based method to determine directly validated material model parameters for sheet metal forming simulations

Structural Mechanics

Universität Stuttgart

Institut für Umformtechnik

The cutting-edge approach in identifying material parameters to be used in FE simulations of sheet metal forming processes is based on the full-field optical measurement of test specimen's deformation and on the simulation based inverse approach. It requires expertise in FE analysis, optimization and programming. The validated values of material parameters can then be found only after several time consuming optimization runs to be sure about the convergence. All these drawbacks lead to use of not validated material data for process design, further leading to losses of resources during the process try-out and production. To remove these drawbacks in the determination of validated material parameters we propose a machine-learning based method eliminating the necessity of an inverse approach. A ML regression model trained on simulation results of huge amount of parameter combinations can yield material parameters directly and accurately from a measured deformation field.