Scholarlink

<p>Hot compression tests of 304 stainless steel were performed under a range of temperatures (800–1200&#xa0;°C) and strain rates (0.01–10&#xa0;s<sup>−1</sup>). A hybrid physics-informed machine-learning framework was developed, in which a PSO-optimized backpropagation neural network provides a data-driven correction factor to enhance the classical Arrhenius constitutive model. Compared with the Arrhenius model, the improved model exhibits higher predictive accuracy, achieving a correlation coefficient of 0.9989 and an average absolute relative error of only 2.06%. A hot working map for a strain of 0.6 was created using information from the enhanced model. According to this map, the optimal hot deformation parameters for 304 stainless steel were identified within the temperature range of 1100–1200&#xa0;°C and strain rate range of 0.05–0.37&#xa0;s<sup>−1</sup>.</p> Graphical abstract <p></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

Optimization of constitutive model and processing map for 304 stainless steel using backpropagation neural network

  • 期刊论文

  • 发表日期：

    2026年1月8日

查看全文
  • Hongyan Duan,
  • Shufan Li,
  • Yuanji Gao

摘要

Hot compression tests of 304 stainless steel were performed under a range of temperatures (800–1200 °C) and strain rates (0.01–10 s−1). A hybrid physics-informed machine-learning framework was developed, in which a PSO-optimized backpropagation neural network provides a data-driven correction factor to enhance the classical Arrhenius constitutive model. Compared with the Arrhenius model, the improved model exhibits higher predictive accuracy, achieving a correlation coefficient of 0.9989 and an average absolute relative error of only 2.06%. A hot working map for a strain of 0.6 was created using information from the enhanced model. According to this map, the optimal hot deformation parameters for 304 stainless steel were identified within the temperature range of 1100–1200 °C and strain rate range of 0.05–0.37 s−1.

Graphical abstract