Scholarlink

<p>This study establishes a hybrid experimental-AI framework integrating deep-learning-based XRD reconstruction with residual-stress and microstructural evaluation of LPBF-fabricated IN718 and A286 steels, benchmarked against conventionally rolled armor steels. The approach bridges AI inference and physical validation, enabling simultaneous phase identification and stress estimation through the Williamson-Hall method applied to both experimental and AI-predicted diffraction profiles. Among all alloys, IN718 exhibited the finest sub-grain morphology (4–6&#xa0;µm), lowest surface roughness (Ra&#xa0;≈&#xa0;3.00&#xa0;µm), and highest compressive residual stress (− 654&#xa0;MPa), followed by A286 (Ra&#xa0;≈&#xa0;7.75&#xa0;µm,– 582&#xa0;MPa). A multi-material LSTM validation across FCC (IN718, A286) and BCC (HNS) systems achieved R<sup>2</sup>&#xa0;≈&#xa0;0.98 with &lt; 7% deviation in residual-stress prediction, confirming robust transferability. The framework’s data preprocessing (noise filtering, background subtraction, and augmentation) ensures reproducibility and scalability for unseen alloys via transfer learning. Integrating experimental fidelity with AI inference establishes a unified path for predictive materials characterization and print-to-performance assessment in defence-grade alloys.</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

Hybrid Experimental and AI-Driven Microstructural and Residual Stress Evaluation of Laser Powder Bed Fusion-Fabricated IN718 and A286 Steel Versus Rolled Armor Steels

  • 期刊论文

  • 发表日期：

    2025年12月1日

查看全文
  • D. Akhil Kumar,
  • B. Veera Siva Reddy,
  • C. Chandrasekhara Sastry,
  • S. Suryakumar,
  • A. Hafeezur Rahman

摘要

This study establishes a hybrid experimental-AI framework integrating deep-learning-based XRD reconstruction with residual-stress and microstructural evaluation of LPBF-fabricated IN718 and A286 steels, benchmarked against conventionally rolled armor steels. The approach bridges AI inference and physical validation, enabling simultaneous phase identification and stress estimation through the Williamson-Hall method applied to both experimental and AI-predicted diffraction profiles. Among all alloys, IN718 exhibited the finest sub-grain morphology (4–6 µm), lowest surface roughness (Ra ≈ 3.00 µm), and highest compressive residual stress (− 654 MPa), followed by A286 (Ra ≈ 7.75 µm,– 582 MPa). A multi-material LSTM validation across FCC (IN718, A286) and BCC (HNS) systems achieved R2 ≈ 0.98 with < 7% deviation in residual-stress prediction, confirming robust transferability. The framework’s data preprocessing (noise filtering, background subtraction, and augmentation) ensures reproducibility and scalability for unseen alloys via transfer learning. Integrating experimental fidelity with AI inference establishes a unified path for predictive materials characterization and print-to-performance assessment in defence-grade alloys.