A New Way for the Adaption of Inverse Identified GTN-Parameters to Bending Processes

One major challenge in metal forming exists in sheet metal bending of modern lightweight materials like high‑strength low-alloyed steels (HSLA), since conventional methods of predicting failure in numerical simulation, like the forming limit diagram (FLD), can generally not be applied to bending processes. Moreover, fracture mechanisms are mainly depending on the microstructure, which is very fine-grained in HSLA steels composed with different alloying elements compared to established mild steels. Consequently the damage and failure behaviour of HSLA steels are changing. Especially for small curvature bending processes characterised by high gradients of strain and stress over the sheet thickness other failure criteria than the FLD have to be utilised. Within this paper a numerical study of the micromechanical based damage model Gurson-Tvergaard-Needleman (GTN, *MAT_120) is performed in LS-DYNA®, in order to realise an effective adaptability of the model for bending operations on HSLA steels. The material dependent damage parameters are determined by commonly used methodology of inverse numerical identification re-calculating the uniaxial tensile test. The minimisation of the mean squared error (MSE) of experimental and numerical global load displacement curves is realised by an optimisation algorithm using commercial software LS‑OPT®. For the adaption of the GTN-Model to the bending operation a strain-based calibration method is developed. This method is based on the comparison and adaption of the numerically calculated and the experimentally measured deformation field on the outer surface of the bent specimen. In this context the parameters are systematically varied again in the optimisation software LS-OPT. Their influence on the strain and damage evolution is analysed and discussed. On the one hand it is shown that it is possible to represent the strain evolution by adapting only one parameter instead of all parameters of the model and thus reducing the modelling effort for the user. On the other hand a big effect on the damage evolution and distribution can be identified.