Estimation of Texture-dependent Stress－Strain Curve and <i>r</i>-value of Aluminum Alloy Sheet Using Deep Learning

Abstract

Deformation of an aluminum alloy sheet is affected by its underlying crystallographic texture and has been widely studied by the crystal plasticity finite element method (CPFEM). The numerical material test based on the CPFEM allows us to quantitatively estimate the stress－strain curve and the Lankford value (r-value), which depend on the texture of aluminum alloy sheets. However, in the use of the numerical material test as a means of optimizing the texture to design aluminum alloys, the CPFEM is computationally expensive. We propose a methodology for rapidly estimating the stress －strain curve and r-value of aluminum alloy sheets using deep learning with a neural network. We train the neural network with synthetic texture and stress－strain curves calculated by the numerical material test. To capture the features of synthetic texture from a {111} pole figure image, the neural network incorporates a convolution neural network. Using the trained neural network, we can estimate the uniaxial stress－strain curve and the in-plane anisotropy of the r-value for various textures that contain Cube and S components. The results indicate that the neural network trained with the results of the numerical material test is a promising methodology for rapidly estimating the deformation of aluminum alloy sheets.