Scholarlink

<p>To address the limitations in accuracy and reliability of the optimization model for the heat treatment process parameters in 2618 aluminum alloy, a data-driven model system (DDMS) that integrates data-driven design, big data analysis, and advanced machine learning techniques is proposed in this paper. Initially, a dataset encompassing technological parameters and performance indicators was constructed. Subsequently, a series of advanced machine learning algorithms were employed for model training and evaluation, including Xtreme gradient boosting, random forest, long short-term memory, and artificial protozoa optimizer-backpropagation (APO-BP). Comprehensive evaluation revealed that the APO-BP model exhibited superior accuracy in predicting UTS, YS, and elongation (<i>δ</i>), with <i>R</i><sup>2</sup> values of 0.99, 0.98, and 0.97, respectively. Therefore, the APO-BP algorithm was employed in the DDMS for technological parameter design and performance prediction. The DDMS achieved significant reductions in standard deviations across all process parameters compared to the baseline module. Specifically, for solution temperature, the value decreased from 3.16 to 0.28; for solution time, from 0.09 to 0.01; for aging temperature, from 1.25 to 0.22; and for aging time, from 0.50 to 0.10. Consequently, the DDMS served as an effective tool for optimizing process parameters in the heat treatment of aluminum alloys and laid a solid theoretical and practical foundation for future research and industrial applications in the field of metallurgical materials.</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 Modeling and Optimization of Heat Treatment Parameters for Aluminum Alloys Using Advanced Machine Learning

  • 期刊论文

  • 发表日期：

    2025年10月27日

查看全文
  • Kai Yan,
  • Shun Hu Zhang,
  • Wei Jian Chen,
  • Ming Kun Ge,
  • Zi Huang

摘要

To address the limitations in accuracy and reliability of the optimization model for the heat treatment process parameters in 2618 aluminum alloy, a data-driven model system (DDMS) that integrates data-driven design, big data analysis, and advanced machine learning techniques is proposed in this paper. Initially, a dataset encompassing technological parameters and performance indicators was constructed. Subsequently, a series of advanced machine learning algorithms were employed for model training and evaluation, including Xtreme gradient boosting, random forest, long short-term memory, and artificial protozoa optimizer-backpropagation (APO-BP). Comprehensive evaluation revealed that the APO-BP model exhibited superior accuracy in predicting UTS, YS, and elongation (δ), with R2 values of 0.99, 0.98, and 0.97, respectively. Therefore, the APO-BP algorithm was employed in the DDMS for technological parameter design and performance prediction. The DDMS achieved significant reductions in standard deviations across all process parameters compared to the baseline module. Specifically, for solution temperature, the value decreased from 3.16 to 0.28; for solution time, from 0.09 to 0.01; for aging temperature, from 1.25 to 0.22; and for aging time, from 0.50 to 0.10. Consequently, the DDMS served as an effective tool for optimizing process parameters in the heat treatment of aluminum alloys and laid a solid theoretical and practical foundation for future research and industrial applications in the field of metallurgical materials.