An evolutionary algorithm-based numerical material testing system based on a crystal plasticity finite element model

Determining all necessary material parameters by simple material tests through numerical material testing method is effective for efficient forming simulation because it allows using complex material models with minimal effort. The proposed numerical material testing method is based on the concept of multi-scale virtual material testing with a crystal plasticity (CP) model. Since we use a crystal plasticity finite element model (CPFEM) of BCC materials, there are many microscopic parameters, including the non-Schmid parameter, as unknown variables. The CPFEM we use is a finite element polycrystal model (FEPM) based on the successive accumulation method to determine the activity of all slip systems. We have developed a numerical material testing system that uses an evolutionary algorithm to predict microscopic parameters from only a few in-plane tests. By learning work hardening-related parameters and texture-related parameters represented by Euler angles using in-plane test results as training data, the CP-based virtual material model acquires generalization capability and is able to predict unlearned biaxial test results with sufficient accuracy. The effectiveness of the proposed method is confirmed by simple forming simulations using the predicted material parameters.