Fast parameter optimization for high-fidelity crystal plasticity simulation using active learning

Crystal plasticity (CP) simulation is a powerful tool for studying and understanding the mechanical behavior of materials. A critical aspect of this method is the accurate determination of CP parameters, which ensures that the constitutive model accurately represents the real deformation behavior of a material, especially in high-fidelity simulations. However, identifying these parameters poses a significant challenge due to the high computational cost and the difficulty of finding optimal solutions within a vast and complex parameter space. To address these challenges, we propose a fast search strategy that leverages active learning (AL) and experimental data to accelerate the optimization of CP parameters. Using the Al-Cu eutectic materials as a case study, we introduced a quantitative index, C ecp, to measure the consistency between simulated and experimental stress-strain curves. We demonstrated that Gaussian process regression (GPR) serves as the most appropriate surrogate model for relating CP parameters and C ecp based on our dataset. After only six iterations guided by AL, the optimal CP parameters were successfully identified, resulting in a high-fidelity CP model for analyzing the mechanical behavior of Al-Cu eutectic materials. The machine learning-enhanced strategy is far superior to traditional methods in terms of both efficiency and accuracy. It advances our understanding of the macro-micro relationships of materials and accelerates the material design process.