Phase Transformation Prediction by Machine Learning-Crystal Plasticity Finite Element Model (ML-CPFEM)

The mechanical properties of alloy are affected significantly by its microstructure. Efforts to touch white area in Ashby’s chart for desired material indices to enhance alloy properties have been investigated for the past decades. This target is coming true through improvement in recognizing microstructure of materials and synergy between microstructure and its effect in bulk materials. Plastic deformation originating from defects in materials mingles with microstructural phenomena, including transformation induced plasticity (TRIP), slip, and twining. It is observed that active microstructural phenomena make materials harder and increase their strength. To understand the mechanism, development of crystal plasticity finite element models (CPFEM) enables us to solve deformation field with the given crystallographic orientation and loading conditions. Over the past few years, emerging research in various fields has been growing in using machine learning (ML) to investigate fundamental constitutive relationships and to accelerate classic numerical modeling. In this study, we leverage a ML model on local crystallography and CPFEM for FCC to HCP phase transformation prediction in a high entropy alloy (HEA). The training data of the ML uses electron backscatter diffraction (EBSD) experimental data before and after deformation and CPFEM simulation results of localized stress and texture evolution during deformation. We use an extreme gradient boosting (XGBoost) ML approach, which repeatedly builds new models and combine them into an ensemble model based on a decision tree. This approach effectively enhances performance for finding optimum model and accelerates searching process. We implement this predictive model in multiple experimental measurements to validate our model. To further understand the driving force for phase transformation, we harness the ML model prediction accuracy to evaluate several different forms of proposed driving force.