Scholarlink

<p>In this study, machine learning techniques were employed for inverse analysis of the underlying mechanisms of local failure in dual-phase (DP) steel, and the local hotspot regions with different sizes were dynamically extracted from numerical predictions. Two-point correlation calculations and principal component analysis (PCA) were applied to obtain high-value, low-dimensional representations of the microstructural features in local regions. Random forest (RF) and Boruta feature selection algorithm were then employed to rank and select the principal components (PCs) based on their importance, with genetic algorithm (GA) used for RF hyperparameter tuning. Finally, Monte Carlo algorithm and individual conditional expectation (ICE) plots were employed to provide qualitative explanations and quantitative analysis of the physical significance of the important PCs. The effectiveness of the framework was validated using DP590 steel as a case study, demonstrating its general application potential in DP composite material 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

A Systematic Framework for Inverse Analysis of Critical Microstructural Features Governing Ductile Fracture in Dual-Phase Steels: Local-Scale

  • 期刊论文

  • 发表日期：

    2025年11月18日

查看全文
  • Lisha Shi,
  • Shujun Yang,
  • Chen Yuan,
  • Ning Nie,
  • Yu Liu,
  • Hui Wang

摘要

In this study, machine learning techniques were employed for inverse analysis of the underlying mechanisms of local failure in dual-phase (DP) steel, and the local hotspot regions with different sizes were dynamically extracted from numerical predictions. Two-point correlation calculations and principal component analysis (PCA) were applied to obtain high-value, low-dimensional representations of the microstructural features in local regions. Random forest (RF) and Boruta feature selection algorithm were then employed to rank and select the principal components (PCs) based on their importance, with genetic algorithm (GA) used for RF hyperparameter tuning. Finally, Monte Carlo algorithm and individual conditional expectation (ICE) plots were employed to provide qualitative explanations and quantitative analysis of the physical significance of the important PCs. The effectiveness of the framework was validated using DP590 steel as a case study, demonstrating its general application potential in DP composite material systems.