Scholarlink

<p>This study presents a data-driven predictive model capable of using microstructural data to estimate properties such as ultimate tensile strength, yield strength, and fracture elongation of additively manufactured (AM) 316L stainless steel (SS). Artificial Neural Networks (ANN) are trained to establish the correlations of cellular structure, grain size, and grain shape cellular structure, and grain shape correlations to corresponding mechanical property attributes. The training data are sourced from existing literature on additively manufactured SS316L and additional data generated by experiments that involved building SS316L samples by AM and performing microstructural analysis with mechanical testing. The fidelity of the trained ANN is evaluated by predicting mechanical properties for out-of-sample testing, which yielded correlation coefficients of <i>R</i><sup><i>2</i></sup><i> = 0.96</i> for ultimate tensile strength, <i>R</i><sup><i>2</i></sup><i> = 0.93</i> for yield strength, and <i>R</i><sup><i>2</i></sup><i> = 0.90</i> for strain at ultimate tensile strength. These results highlight the effectiveness of the data-driven model in leveraging domain-specific knowledge to establish microstructure–property correlations. This offers a reliable tool for estimating the spatial and temporal variation of mechanical properties in AM-processed components, thus reducing the necessity for extensive testing and long lead time.</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

Prediction of Mechanical Properties of Additively Manufactured 316L Steel Based on Microstructural Features Using Data-Driven Approaches

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年9月3日

查看全文
  • Akanksha Parmar,
  • Rohit Surya Kumar,
  • Mark Atwater,
  • Yung C. Shin

摘要

This study presents a data-driven predictive model capable of using microstructural data to estimate properties such as ultimate tensile strength, yield strength, and fracture elongation of additively manufactured (AM) 316L stainless steel (SS). Artificial Neural Networks (ANN) are trained to establish the correlations of cellular structure, grain size, and grain shape cellular structure, and grain shape correlations to corresponding mechanical property attributes. The training data are sourced from existing literature on additively manufactured SS316L and additional data generated by experiments that involved building SS316L samples by AM and performing microstructural analysis with mechanical testing. The fidelity of the trained ANN is evaluated by predicting mechanical properties for out-of-sample testing, which yielded correlation coefficients of R2 = 0.96 for ultimate tensile strength, R2 = 0.93 for yield strength, and R2 = 0.90 for strain at ultimate tensile strength. These results highlight the effectiveness of the data-driven model in leveraging domain-specific knowledge to establish microstructure–property correlations. This offers a reliable tool for estimating the spatial and temporal variation of mechanical properties in AM-processed components, thus reducing the necessity for extensive testing and long lead time.