Scholarlink

<p>The optimization of structural steel compositions for high-temperature applications requires a systematic approach to balance the resulting microstructure with competing mechanical properties such as proof stress, ultimate tensile strength (UTS), and elongation. In this study, a machine learning (ML)-driven framework is developed to predict and optimize mechanical properties of steels across a 22-dimensional composition space. The dataset includes over 3000 compositions sourced from the National Institute for Materials Science (NIMS), Japan database. An Extreme Gradient-Boosted decision tree (XGBoost) model is developed to predict mechanical properties with high accuracy, achieving R<sup>2</sup> values of 0.98, 0.98, and 0.94 and RMSE values of 18.8&#xa0;MPa, 17.5&#xa0;MPa, and 3.7% for 0.2% proof stress, UTS, and elongation, respectively. The ML model is integrated with the Non-dominated Sorting Genetic Algorithm II (NSGA-II) to optimize compositions for single and multi-objective scenarios. Single-objective optimization focuses on maximizing individual properties, while multi-objective optimization explores pareto-optimal solutions to identify the most effective compromises between strength and ductility over a wide temperature range (27&#xa0;°C, 400&#xa0;°C, and 1000&#xa0;°C). The optimized outputs are presented as candidate trade-off solutions suggested by the model, indicating theoretical pathways for enhanced performance. The optimal solutions also demonstrate combined effects of elements to the tensile properties. This work highlights the potential of data-driven alloy design frameworks to accelerate materials discovery by reducing reliance on trial-and-error experiments. The methodology developed here is scalable to other alloy systems, enabling the design of next-generation structural materials for applications in aerospace, nuclear reactors, and renewable energy systems.</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

Data-driven framework for alloy design of steels in high-temperature applications

  • 期刊论文

  • 发表日期：

    2026年7月1日

查看全文
  • Anish Atey,
  • Pikee Priya

摘要

The optimization of structural steel compositions for high-temperature applications requires a systematic approach to balance the resulting microstructure with competing mechanical properties such as proof stress, ultimate tensile strength (UTS), and elongation. In this study, a machine learning (ML)-driven framework is developed to predict and optimize mechanical properties of steels across a 22-dimensional composition space. The dataset includes over 3000 compositions sourced from the National Institute for Materials Science (NIMS), Japan database. An Extreme Gradient-Boosted decision tree (XGBoost) model is developed to predict mechanical properties with high accuracy, achieving R2 values of 0.98, 0.98, and 0.94 and RMSE values of 18.8 MPa, 17.5 MPa, and 3.7% for 0.2% proof stress, UTS, and elongation, respectively. The ML model is integrated with the Non-dominated Sorting Genetic Algorithm II (NSGA-II) to optimize compositions for single and multi-objective scenarios. Single-objective optimization focuses on maximizing individual properties, while multi-objective optimization explores pareto-optimal solutions to identify the most effective compromises between strength and ductility over a wide temperature range (27 °C, 400 °C, and 1000 °C). The optimized outputs are presented as candidate trade-off solutions suggested by the model, indicating theoretical pathways for enhanced performance. The optimal solutions also demonstrate combined effects of elements to the tensile properties. This work highlights the potential of data-driven alloy design frameworks to accelerate materials discovery by reducing reliance on trial-and-error experiments. The methodology developed here is scalable to other alloy systems, enabling the design of next-generation structural materials for applications in aerospace, nuclear reactors, and renewable energy systems.