Scholarlink

<p>10Cr ferritic steels are widely employed in power generation and petrochemical applications, yet their performance requirements in ultra-supercritical power plants continue to escalate. This study establishes a computational framework for 10Cr ferritic steel design by integrating single-objective surrogate optimization with multivariate regression modeling, aiming to facilitate the development of high-performance alloys. A dataset comprising 274 experimental records was compiled from the MLMD (Materials Learning and Materials Design) platform. A predictive model was constructed through the synergistic combination of K-means clustering and extreme gradient boosting (XGBoost) algorithms, with silhouette coefficient analysis identifying two optimal clusters. Comparative evaluation revealed Cluster 2 as the superior subgroup for model training. Further investigation demonstrated peak predictive performance when employing a 7:3 training-to-testing data partitioning ratio. Among AdaBoost (Adab), random forest (RF), support vector regression (SVR), and XGBoost models, the XGBoost algorithm exhibited exceptional efficacy in predicting ultimate tensile strength (UTS), achieving a coefficient of determination (R<sup>2</sup>) of 0.9355. This framework facilitated the design of a novel alloy maintaining a UTS of 444.46 MPa at 625 °C.</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

High-Performance Alloy Design of 10Cr Ferrite Steel: An Integrated Approach of Single-Objective Optimization and Machine Learning

  • 期刊论文

  • 发表日期：

    2025年9月24日

查看全文
  • Gaolei Xu,
  • Longlong Shen,
  • Chengfeng Wang,
  • Yanfu Chai,
  • Zubin Chen,
  • Qing Guo,
  • Xukai Ren,
  • Zhiyu Chang,
  • Gang Zhu,
  • Qinghang Wang

摘要

10Cr ferritic steels are widely employed in power generation and petrochemical applications, yet their performance requirements in ultra-supercritical power plants continue to escalate. This study establishes a computational framework for 10Cr ferritic steel design by integrating single-objective surrogate optimization with multivariate regression modeling, aiming to facilitate the development of high-performance alloys. A dataset comprising 274 experimental records was compiled from the MLMD (Materials Learning and Materials Design) platform. A predictive model was constructed through the synergistic combination of K-means clustering and extreme gradient boosting (XGBoost) algorithms, with silhouette coefficient analysis identifying two optimal clusters. Comparative evaluation revealed Cluster 2 as the superior subgroup for model training. Further investigation demonstrated peak predictive performance when employing a 7:3 training-to-testing data partitioning ratio. Among AdaBoost (Adab), random forest (RF), support vector regression (SVR), and XGBoost models, the XGBoost algorithm exhibited exceptional efficacy in predicting ultimate tensile strength (UTS), achieving a coefficient of determination (R2) of 0.9355. This framework facilitated the design of a novel alloy maintaining a UTS of 444.46 MPa at 625 °C.