Scholarlink

<p>Centered on high-entropy alloys (HEAs), this study develops an interpretable machine learning framework for AlFeNi-based high-entropy alloys that simultaneously predicts phase composition, hardness, and corrosion current density. By integrating SHapley Additive exPlanations (SHAP) with physical descriptors, we refined 17 initial features to 7-8 core parameters. The optimized models demonstrated high accuracy: Random Forest (91% for phase classification), Gaussian Process Regression with GMM augmentation (<i>R</i><sup>2</sup> = 0.96 for hardness), and K-Nearest Neighbors (<i>R</i><sup>2</sup> = 0.88 for corrosion current density). SHAP analysis identified <i>T</i><sub>m</sub>/FIE and EA/Δ<i>H</i> as the dominant factors governing hardness and corrosion behavior, respectively. Guided by these insights, we designed and experimentally validated three novel HEAs: Al<sub>0.1</sub>Fe<sub>2</sub>Ni<sub>2</sub>V<sub>0.1</sub>Si<sub>0.05</sub>, AlFe<sub>2</sub>Ni<sub>2</sub>VSi<sub>0.05</sub>, and AlCoFe<sub>2</sub>Ni<sub>2</sub>Si<sub>0.05</sub>. These alloys exhibit tailored hardness and superior corrosion resistance. This work establishes an interpretability-driven paradigm that integrates machine learning with experimental validation to accelerate the design of high-performance HEAs.</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

Effect of V/Si and Co/Si Co-doping on AlFeNi-based HEAs Using Machine Learning with SHAP Explainability

  • 期刊论文

  • 发表日期：

    2026年3月5日

查看全文
  • Zhuoyi Wang,
  • Hanqing Xu,
  • Mengdi Zhang,
  • Tianming Li,
  • Xin Zhao

摘要

Centered on high-entropy alloys (HEAs), this study develops an interpretable machine learning framework for AlFeNi-based high-entropy alloys that simultaneously predicts phase composition, hardness, and corrosion current density. By integrating SHapley Additive exPlanations (SHAP) with physical descriptors, we refined 17 initial features to 7-8 core parameters. The optimized models demonstrated high accuracy: Random Forest (91% for phase classification), Gaussian Process Regression with GMM augmentation (R2 = 0.96 for hardness), and K-Nearest Neighbors (R2 = 0.88 for corrosion current density). SHAP analysis identified Tm/FIE and EA/ΔH as the dominant factors governing hardness and corrosion behavior, respectively. Guided by these insights, we designed and experimentally validated three novel HEAs: Al0.1Fe2Ni2V0.1Si0.05, AlFe2Ni2VSi0.05, and AlCoFe2Ni2Si0.05. These alloys exhibit tailored hardness and superior corrosion resistance. This work establishes an interpretability-driven paradigm that integrates machine learning with experimental validation to accelerate the design of high-performance HEAs.