Anisotropic Shear Behavior of AA7075-T6: Machine Learning Modeling and Failure Mechanism

In this paper, we develop an artificial neural network model optimized by a genetic algorithm to predict the anisotropic shear responses of AA7075-T6. The orientation-dependent failure mechanisms are investigated and analyzed. Before training, classical plastic theory, i.e., the anisotropic quadratic yield function is used as the yield function. The findings show that the quadratic function is unsuitable for modeling the shear anisotropy of material. The machine learning model provides a flexible and accurate prediction of orientation-dependent shear strength. In addition, the qualitative effects of orientation on fracture failure mechanisms are discussed. The influences of orientation are mainly shown as the changes in local ductile characteristics, i.e., the size and growth direction of dimples.