Machine learning assisted design of novel refractory high entropy alloys with enhanced mechanical properties

authored by: A. A. Catal, E. Bedir, R. Yilmaz, M. A. Swider, C. Lee, O. El-Atwani, H. J. Maier, H. C. Ozdemir, D. Canadinc
Abstract: This paper details an alloy design effort by machine learning (ML) attempting to design novel refractory high entropy alloys (RHEAs) with exceptional mechanical properties at elevated temperatures and good room temperature ductility. For this purpose, four datasets were generated by mining the data available in literature, containing room temperature strength, high temperature strength, room temperature ductility and hardness, which were trained by three different ML models, namely the support vector regression, random forest, and artificial neural network. As a result, three novel RHEA compositions were predicted, and their performances were experimentally validated. Specifically, the Ti₈Nb₂₁Zr₂₇Ta₁₃Mo₁₉V₁₂, Ti₁₀Nb₁₉Zr₁₅Ta₄₃Mo₇V₆, and Ti₁₀Nb₂₀Zr₃₇Mo₂₁V₁₂ RHEAs were produced and subjected to room-temperature and high-temperature compression, and room-temperature hardness tests, which have demonstrated that especially the Ti₈Nb₂₁Zr₂₇Ta₁₃Mo₁₉V₁₂ and the Ti₁₀Nb₂₀Zr₃₇Mo₂₁V₁₂ RHEAs exhibit both high strength at elevated temperatures and good room-temperature ductility. The current study not only contributes to the literature by presenting three novel RHEAs, but also constitutes an encouraging example of efficient alloy design by ML for demanding applications.
Organisation(s): Institute of Materials Science
External Organisation(s): Koc University
Eskişehir Technical University
Los Alamos National Laboratory Materials Science and Technology Division
Auburn University (AU)
Type: Article
Journal: Computational materials science
Volume: 231
ISSN: 0927-0256
Publication date: 05.01.2024
Publication status: Published
Peer reviewed: Yes
ASJC Scopus subject areas: General Computer Science, General Chemistry, General Materials Science, Mechanics of Materials, General Physics and Astronomy, Computational Mathematics
Electronic version(s): https://doi.org/10.1016/j.commatsci.2023.112612 (Access: Closed)

BibTeX

@article{2a86c7a1c6b14a99ba565970b97d76b6,
title = "Machine learning assisted design of novel refractory high entropy alloys with enhanced mechanical properties",
abstract = "This paper details an alloy design effort by machine learning (ML) attempting to design novel refractory high entropy alloys (RHEAs) with exceptional mechanical properties at elevated temperatures and good room temperature ductility. For this purpose, four datasets were generated by mining the data available in literature, containing room temperature strength, high temperature strength, room temperature ductility and hardness, which were trained by three different ML models, namely the support vector regression, random forest, and artificial neural network. As a result, three novel RHEA compositions were predicted, and their performances were experimentally validated. Specifically, the Ti8Nb21Zr27Ta13Mo19V12, Ti10Nb19Zr15Ta43Mo7V6, and Ti10Nb20Zr37Mo21V12 RHEAs were produced and subjected to room-temperature and high-temperature compression, and room-temperature hardness tests, which have demonstrated that especially the Ti8Nb21Zr27Ta13Mo19V12 and the Ti10Nb20Zr37Mo21V12 RHEAs exhibit both high strength at elevated temperatures and good room-temperature ductility. The current study not only contributes to the literature by presenting three novel RHEAs, but also constitutes an encouraging example of efficient alloy design by ML for demanding applications.",
keywords = "Alloy design, Ductility, High-temperature strength, Machine learning, Refractory high entropy alloy",
author = "Catal, {A. A.} and E. Bedir and R. Yilmaz and Swider, {M. A.} and C. Lee and O. El-Atwani and Maier, {H. J.} and Ozdemir, {H. C.} and D. Canadinc",
note = "Funding Information: D. Canadinc acknowledges the support by the Alexander von Humboldt Foundation (Germany) within the scope of Humboldt Research Award. H.J. Maier acknowledges financial support by Deutsche Forschungsgemeinschaft ( project #388671975 ) (Germany). ",
year = "2024",
month = jan,
day = "5",
doi = "10.1016/j.commatsci.2023.112612",
language = "English",
volume = "231",
journal = "Computational materials science",
issn = "0927-0256",
publisher = "Elsevier",
}