Scholarlink

<p>This investigation presents a framework based on surrogate machine learning models and high- throughput alloy screening to develop new&#xa0;face-centered-cubic structured high-entropy alloys with high thermal stability,&#xa0;high strength, low density, and low cost.&#xa0;The study demonstrates the feasibility of computational alloy design using a combination of thermodynamic and empirical model-based calculations applied to a library of 1451 quinary alloys in a Co–Cr–Fe–Ni–Mn system. The developed machine learning models based on noted calculations were employed for phase prediction of 72,501 quaternary and quinary alloys. High-throughput alloy screening was performed based on the predictions of the XGBoost model, found to be the best performing model, to identify alloys with a single&#xa0;face-centered-cubic&#xa0;phase.&#xa0;The results indicated remarkable phase prediction accuracy of the XGBoost model, which identified 28,948 single FCC-structured alloys. A physics-based strength model was subsequently used to&#xa0;estimate the strength of the selected single-phase&#xa0;high-entropy alloys. High-throughput screening was performed using criteria including strength &gt; 250 MPa, density ≤ 8&#xa0;g/cm<sup>3</sup>,&#xa0;and cost &lt; 6 USD per kg. The maximum strength of 403 MPa was estimated for an Fe34Mn26Co24Cr8Ni8 (at.%) HEA with a density of 7.69&#xa0;g/cm<sup>3</sup>. This investigation demonstrates&#xa0;the potential of surrogate machine learning models for phase prediction of large alloy spaces followed by high-throughput screening based on application specific property requirements, to enable the development of high-performance alloys.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Machine learning-assisted high-throughput exploration of single-phase high-entropy alloys

  • 期刊论文

  • 发表日期：

    2026年6月15日

查看全文
  • Hamed Shahmir,
  • Farsad Forghani

摘要

This investigation presents a framework based on surrogate machine learning models and high- throughput alloy screening to develop new face-centered-cubic structured high-entropy alloys with high thermal stability, high strength, low density, and low cost. The study demonstrates the feasibility of computational alloy design using a combination of thermodynamic and empirical model-based calculations applied to a library of 1451 quinary alloys in a Co–Cr–Fe–Ni–Mn system. The developed machine learning models based on noted calculations were employed for phase prediction of 72,501 quaternary and quinary alloys. High-throughput alloy screening was performed based on the predictions of the XGBoost model, found to be the best performing model, to identify alloys with a single face-centered-cubic phase. The results indicated remarkable phase prediction accuracy of the XGBoost model, which identified 28,948 single FCC-structured alloys. A physics-based strength model was subsequently used to estimate the strength of the selected single-phase high-entropy alloys. High-throughput screening was performed using criteria including strength > 250 MPa, density ≤ 8 g/cm3, and cost < 6 USD per kg. The maximum strength of 403 MPa was estimated for an Fe34Mn26Co24Cr8Ni8 (at.%) HEA with a density of 7.69 g/cm3. This investigation demonstrates the potential of surrogate machine learning models for phase prediction of large alloy spaces followed by high-throughput screening based on application specific property requirements, to enable the development of high-performance alloys.