Prediction and optimization of mechanical properties of Ni based and Fe–Ni based super alloys via neural network approach with alloying composition parameter

The Optimization of super alloy through alloying element composition control is very challenging since it contains more than ten type of elements. Generally, the optimization process of the mechanical properties of super alloy by composition alteration was performed via trial-and-error which is time consuming and expensive. In this study, the artificial neural network was successfully employed to reveal the correlation between alloying element and the hardness, tensile strength, and melting point of Ni based and Fe–Ni based super alloy. The model showed a good accuracy between predicted and actual values, especially for hardness and melting temperature, with R2 and RMSE values were found to be above 95% and below 3%, respectively. The Pearson Correlation Coefficient and Feature Importance revealed the linear and non-linear correlation of elements in the matrix. The validation of mathematical expression derived from symbolic regression displayed a high reliability associated with RMSE values of approaching 0%. The feature importance derived from atomic and thermal features characterization on Fe–Ni based super alloy as selected sample revealed the dominant effect of Ni matrix on the mechanical properties of the alloy. These results show the potential of our model to assist in the designing of super alloy for industry.