A crystal plasticity-informed data-driven model for magnesium alloys

In the past few years, data-driven models based on artificial neural network (ANN) have been successfully developed and applied to investigate the macro- and micro-mechanical behaviors of various materials. However, these data-driven models are either too complex in structure or lack interpretable physical insights. In the present work, a crystal plasticity-informed data-driven (CPIDD) model is proposed, which updates the microstructural information and parameters associated with the macroscopic constitutive model using a parallel ANN structure, and combines conventional constitutive equations to obtain the stress-strain response, ensuring efficient and stable calculations. In conjunction with the finite element (FE) method, the FE-CPIDD model simulates the micro- and macro-mechanical behaviors of magnesium (Mg) alloys under uniaxial loading, non-proportional loading, four-point bending and unloading. The comparison between the simulations and available experiments (or crystal plasticity simulations) demonstrates the accuracy and effectiveness of the proposed CPIDD model. Using Mg alloys as a representative case, the CPIDD model provides an operational and extensional tool for the design, fabrication, manufacturing, and service of the metallic components.