Figure 1.

Machine learning training strategy diagram. Based on the thermodynamic database and experimental data, ANN I and ANN II artificial neural networks were used to establish the “composition-microstructure” and “microstructure-properties” relationships of the target alloys. Based on the Bayesian optimization process, a Gaussian simulation process was performed, and then the inference and validated results were obtained based on the simulation results. The new trail should be used in the next iteration.