Machine learning assisted design of FeCoNiCrMn high-entropy alloys with ultra-low hydrogen diffusion coefficients

The broad compositional space of high entropy alloys (HEA) is conducive to the design of HEAs with targeted performance. Herein, a data-driven and machine learning (ML) assisted prediction and optimization strategy is proposed to explore the prototype FeCoNiCrMn HEAs with low hydrogen diffusion coefficients. The model for predicting hydrogen solution energies from local HEA chemical environments was constructed via ML algorithms. Based on the inferred correlation between atomic structures and diffusion coefficients of HEAs built using ML models and kinetic Monte Carlo simulations, we employed the whale optimization algorithm to explore HEA atomic structures with low hydrogen diffusion coefficients. HEAs with low H diffusion coefficients were found to have high Co and Mn content. Finally, a quantitative relationship between the diffusion coefficient and chemical composition is proposed to guide the design of HEAs with low H diffusion coefficients and thus strong resistance to hydrogen embrittlement.(c) 2021 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

Automated summary

A data-driven and machine learning assisted strategy was proposed to explore the design of high entropy alloys (HEA) with low hydrogen diffusion coefficients. By constructing a model to predict hydrogen solution energies using machine learning algorithms, the correlation between atomic structures and diffusion coefficients of HEAs was inferred. The whale optimization algorithm was then used to discover HEA atomic structures with low hydrogen diffusion coefficients. Furthermore, a quantitative relationship between diffusion coefficient and chemical composition was proposed to guide the design of HEAs with strong resistance to hydrogen embrittlement.