Machine learning – informed development of high entropy alloys with enhanced corrosion resistance

Abstract

This study demonstrates the use of machine learning as a potential tool to efficiently develop new biomedical alloys with improved corrosion resistance by exploring the whole compositional space in the HfNbTaTiZr system. Owing to the small volume and inherited uncertainty of available corrosion data in the literature, k-fold cross-validation and bootstrapping were used to quantify the uncertainty of models and select a robust one. Potentiodynamic polarization experiments were performed on the predicted composition in simulated body fluid at 37 ± 1 °C for validation, demonstrating the new alloy's superior corrosion properties with a homogeneous microstructure as opposed to the dendritic structure.