Scholarlink

Accurate microstructural characterizationCharacterization is essential for understanding and optimizing the performance of metallic components produced by powder-based additive manufacturingAdditive manufacturing (PBAM). While two-dimensional (2D) cross-sectional imaging is widely used due to its accessibility, conventional image processing methods are labor-intensive and insufficient for reliably capturing three-dimensional (3D) particle contact features. This study introduces a machine learningMachine learning (ML) framework for predicting 3D particle contact areas (PCAs) directly from 2D microstructural images, eliminating the need for complete 3D X-ray computed tomography (CT) datasets. The framework leverages contact-line statistics extracted from 2D images to predict 3D contact areas. Validation using 316L stainless steelStainless steel PBAM samples demonstrates that the model achieves an average accuracy of 98.5%. By providing a scalable, cost-effective, and accurate method for PCA characterizationCharacterization, this work advances the quantitative analysis of processProcess-induced PCA anisotropyAnisotropy in PBAM and supports improved processProcess optimization and material design.

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

Machine Learning-Enabled Characterization of 3D Particle Contacts from 2D Microstructural Images in Powder-Based Metal Additive Manufacturing

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Nicholas Satterlee,
  • Elisa Torresani,
  • Runjian Jiang,
  • Marc P. F. H. L. van Maris,
  • Diletta Giuntini,
  • Eugene Olevsky,
  • John S. Kang

摘要

Accurate microstructural characterizationCharacterization is essential for understanding and optimizing the performance of metallic components produced by powder-based additive manufacturingAdditive manufacturing (PBAM). While two-dimensional (2D) cross-sectional imaging is widely used due to its accessibility, conventional image processing methods are labor-intensive and insufficient for reliably capturing three-dimensional (3D) particle contact features. This study introduces a machine learningMachine learning (ML) framework for predicting 3D particle contact areas (PCAs) directly from 2D microstructural images, eliminating the need for complete 3D X-ray computed tomography (CT) datasets. The framework leverages contact-line statistics extracted from 2D images to predict 3D contact areas. Validation using 316L stainless steelStainless steel PBAM samples demonstrates that the model achieves an average accuracy of 98.5%. By providing a scalable, cost-effective, and accurate method for PCA characterizationCharacterization, this work advances the quantitative analysis of processProcess-induced PCA anisotropyAnisotropy in PBAM and supports improved processProcess optimization and material design.