A Method for Crystal Plasticity Model Parameter Calibration Based on Bayesian Optimization

The crystal plasticity model is an efficient method to bridge the mechanical characteristics of a material at the crystallographic scale to the macroscopic mechanical responses. However, the relatively large number of model parameters makes the calibration cumbersome, especially in systems with hexagonal close-packed (HCP) structures like magnesium alloys. This work presents a Bayesian Optimization based approach, which is applied to the calibration of the viscoplastic self-consistent polycrystal plasticity model with twinning and de-twinning scheme (VPSC-TDT) to describe the mechanical behavior of the rare-earth magnesium alloy ZEK100. The result shows that Bayesian Optimization can perform well in such a physical principle-based black-box optimization problem. Combined with a practical tactic, the total trial number can be reduced to around 100, efficiently reducing the time cost. The obtained optimized set of parameters can successfully reproduce the loading path-dependent mechanical behavior of the Mg alloy ZEK100.