Scholarlink

<p>Reconstructing three-dimensional grain structures from planar imaging data remains a central challenge in materials science, as real microstructures are inherently three-dimensional while most characterization techniques provide only two-dimensional maps. This study presents an artificial intelligence-driven framework for reconstructing three-dimensional polycrystalline microstructures from a single inverse pole figure image. A genetic algorithm, guided by a fitness function based on the distance of statistical descriptors, is employed to evolve grain structures that match the reference microstructure without requiring large training datasets or deep learning architectures. The method is validated on both synthetic data and an experimental inverse pole figure map from an Inconel 718H superalloy. Results show high fidelity, with grain size distribution peaks in the synthetic case deviating by less than 0.2 units from the reference. A parameter sensitivity analysis confirms the robustness of the approach. The proposed method provides a computationally efficient, data-light alternative for microstructure reconstruction, offering direct relevance to materials design, digital twin development, and structure–property modeling workflows.</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

Evolutionary artificial intelligence for 3D microstructure reconstruction based on inverse pole figure images

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年11月25日

查看全文
  • Fatih Uzun,
  • Alexander M. Korsunsky

摘要

Reconstructing three-dimensional grain structures from planar imaging data remains a central challenge in materials science, as real microstructures are inherently three-dimensional while most characterization techniques provide only two-dimensional maps. This study presents an artificial intelligence-driven framework for reconstructing three-dimensional polycrystalline microstructures from a single inverse pole figure image. A genetic algorithm, guided by a fitness function based on the distance of statistical descriptors, is employed to evolve grain structures that match the reference microstructure without requiring large training datasets or deep learning architectures. The method is validated on both synthetic data and an experimental inverse pole figure map from an Inconel 718H superalloy. Results show high fidelity, with grain size distribution peaks in the synthetic case deviating by less than 0.2 units from the reference. A parameter sensitivity analysis confirms the robustness of the approach. The proposed method provides a computationally efficient, data-light alternative for microstructure reconstruction, offering direct relevance to materials design, digital twin development, and structure–property modeling workflows.