Machine learning guided prediction of the yield strength and hardness of multi-principal element alloys

Materials Open Research Pub Date : 2023-10-09 DOI:10.12688/materialsopenres.17476.1

Mohammad Fuad Nur Taufique, Osman Mamun, Ankit Roy, Hrishabh Khakurel, Ganesh Balasubramanian, Gaoyuan Ouyang, Jun Cui, Duane D. Johnson, Ram Devanathan

{"title":"Machine learning guided prediction of the yield strength and hardness of multi-principal element alloys","authors":"Mohammad Fuad Nur Taufique, Osman Mamun, Ankit Roy, Hrishabh Khakurel, Ganesh Balasubramanian, Gaoyuan Ouyang, Jun Cui, Duane D. Johnson, Ram Devanathan","doi":"10.12688/materialsopenres.17476.1","DOIUrl":null,"url":null,"abstract":"<ns4:p><ns4:bold>Background: </ns4:bold>Multi-Principal Element Alloys (MPEAs) have better properties, such as yield strength, hardness, and corrosion resistance compared to conventional alloys. Compositional optimization is a challenging task to obtain desired properties of MPEAs and machine learning is a potential tool to rapidly accelerate the search and design of new materials.</ns4:p><ns4:p> <ns4:bold>Methods:</ns4:bold> We have implemented different machine learning models to predict the yield strength and Vickers hardness of MPEAs at room temperature and quantify the uncertainty of the predictions.</ns4:p><ns4:p> <ns4:bold>Results:</ns4:bold> Our results suggest that valence electron concentration (VEC) is the key feature dominating the yield strength and hardness of MPEAs. Our predicted yield strength and hardness values for the experimental validation set show < 15 % error for most cases with respect to the experimental values.</ns4:p><ns4:p> <ns4:bold>Conclusions:</ns4:bold> Our machine learning model can serve as a useful tool to screen half a trillion MPEAs and down select promising compositions for useful applications.</ns4:p>","PeriodicalId":29806,"journal":{"name":"Materials Open Research","volume":"18 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2023-10-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Materials Open Research","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.12688/materialsopenres.17476.1","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}

引用次数: 0

Abstract

Background: Multi-Principal Element Alloys (MPEAs) have better properties, such as yield strength, hardness, and corrosion resistance compared to conventional alloys. Compositional optimization is a challenging task to obtain desired properties of MPEAs and machine learning is a potential tool to rapidly accelerate the search and design of new materials. Methods: We have implemented different machine learning models to predict the yield strength and Vickers hardness of MPEAs at room temperature and quantify the uncertainty of the predictions. Results: Our results suggest that valence electron concentration (VEC) is the key feature dominating the yield strength and hardness of MPEAs. Our predicted yield strength and hardness values for the experimental validation set show < 15 % error for most cases with respect to the experimental values. Conclusions: Our machine learning model can serve as a useful tool to screen half a trillion MPEAs and down select promising compositions for useful applications.

查看原文本刊更多论文

背景:与传统合金相比，多主元素合金具有更好的性能，如屈服强度、硬度和耐腐蚀性。为了获得理想的mpea性能，组分优化是一项具有挑战性的任务，而机器学习是快速加速新材料搜索和设计的潜在工具。方法:我们实现了不同的机器学习模型来预测mpea在室温下的屈服强度和维氏硬度，并量化预测的不确定性。结果:价电子浓度(VEC)是决定mpea屈服强度和硬度的关键因素。我们对实验验证集的屈服强度和硬度预测值显示<大多数情况下与实验值有15%的误差。结论:我们的机器学习模型可以作为一个有用的工具来筛选5000亿个mpea，并为有用的应用选择有前途的组合物。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

Materials Open Research materials science-

自引率

0.00%

发文量

期刊介绍： Materials Open Research is a rapid open access publishing platform for a broad range of materials science research. The platform welcomes theoretical, experimental, and modelling approaches on the properties, characterization, design, structure, classification, processing, and performance of materials, and their applications. The platform is open to submissions from researchers, practitioners and experts, and all articles will benefit from open peer review.  Materials research underpins many significant and novel technologies which are set to revolutionize our society, and Materials Open Research is well-suited to ensure fast and full access to this research for the benefit of the academic community, industry, and beyond. The platform aims to create a forum for discussion and for the dissemination of research in all areas of materials science and engineering. This includes, but is not limited to, research on the following material classes: ● Biomaterials and biomedical materials ● Composites ● Economic minerals ● Electronic materials ● Glasses & ceramics ● Magnetic materials ● Metals & alloys ● Nanomaterials and nanostructures ● Polymers ● Porous materials ● Quantum materials ● Smart materials ● Soft matter ● Structural materials ● Superconducting materials ● Thin films Materials Open Research also focuses on a range of applications and approaches within materials science, including but not limited to: ● Additive manufacturing ● Computational materials & modelling ● Materials in energy & the environment ● Materials informatics ● Materials synthesis and processing In addition to original Research Articles, Materials Open Research will feature a variety of article types including Method Articles, Study Protocols, Software Tool Articles, Systematic Reviews, Data Notes, Brief Reports, and Opinion Articles. All research is welcome and will be published irrespective of the perceived level of interest or novelty; we accept confirmatory and replication studies, as well as negative and null results.  Materials Open Research is an Open Research Platform. All articles are published open access under a CC-BY license and authors benefit from fully transparent publishing and peer review processes. Where applicable, authors are asked to include detailed descriptions of methods and will receive editorial guidance on making all underlying data openly available in order to improve reproducibility. The platform will also provide the option to publish non-peer reviewed materials including technical reports, training materials, posters, slides, and other documents.