# Efficient and Robust Crystal Plasticity Parameter Identification via RSM-GA Coupling: Application to AZ31 Magnesium Alloy with Bimodal Non-Basal Texture

## TL;DR

A new method using RSM-GA coupling efficiently identifies crystal plasticity parameters for AZ31 magnesium alloy with bimodal non-basal texture.

## Contribution

A novel RSM-GA framework with weight-amplified multiscale calibration for efficient parameter identification in crystal plasticity models.

## Key findings

- The RSM-GA framework effectively captures plastic deformation and microstructural evolution in AZ31 Mg alloy.
- The optimization procedure converges rapidly with a stable fitness value of ~80 within 200 iterations.
- The method shows good validity in mechanical response and deformation mechanism activities.

## Abstract

RSM-GA framework with a weight-amplified multiscale calibration strategy is developed for efficient crystal plasticity parameter identification.

Multi-orientation calibration is essential for accurate cross-scenario mechanical and microstructural predictions.

Material parameters based on the constructed RSM-GA framework effectively capture the plastic deformation behavior and microstructural evolution of AZ31 Mg alloy with bimodal non-basal texture under a variety of loading paths.

Identifying material parameters in crystal plasticity constitutive models with high precision and high efficiency can be especially complicated due to the ever-increasing complexity of the models. These material parameters are typically calibrated through the fitting of macroscale experimental data, such as true stress–strain curves, while microscale experimental data, including phase evolution, twinning volume fraction, and so on, are rarely considered and used for verification. In the present study, a novel and computationally efficient optimization procedure for material parameters identification in a crystal plasticity constitutive model has been proposed, which couples a response surface model (RSM) and a genetic algorithm (GA). Specifically, 34 macroscopic true stress–strain data (21 for rolling direction, RD, and 13 for transverse direction, TD) and 4 microscale {10-12} extension twin (ET) volume fraction data have been utilized for multi-objective training. Furthermore, the objective function has been optimized in the present study by tailoring the weights of macroscale stress–strain data and microscale volume fractions for {10-12} ET. The proposed optimization methodology has been verified via visco-plastic self-consistent (VPSC) simulation of tensile deformation for AZ31 magnesium (Mg) alloy sheet with bimodal non-basal texture at room temperature. Results show that the fitness value of the optimization procedure would rapidly converge to a stable value of ~80 within 200 iterations. The obtained material parameters for VPSC simulation on the basis of RD-tensile and TD-tensile experimental data show good validity and applicability in aspects of mechanical response, activities of involved deformation mechanisms, evolution of volume fraction for {10-12} ET, and characteristics of texture evolution.