首页|期刊导航|Journal of Magnesium and Alloys|Deep learning-driven microstructure characterization and Vickers-hardness prediction of Mg-Gd alloys

Deep learning-driven microstructure characterization and Vickers-hardness prediction of Mg-Gd alloysOA

中文摘要

In high-strength Mg-rare earth(RE)alloys,solution treatment and aging processes have pronounced effects on microstructure and properties,yet traditional experimental and physical approaches still struggle to establish a quantitative mapping among processing parameters,microstructural features,and property response.Using high-strength Mg-Gd alloys as a case study,we establish a quantitative“process(solidsolution and aging)-microstructure-property”framework based on image recognition and deep learning.Specifically,the mechanical property of solid-solution Mg-Gd alloys depends on Gd content,grain boundary,and the presence of second-phase,while the mechanical property of aged alloys depends similarly on Gd content,aging parameters and the presence of precipitates.To establish the aforementioned mapping,we propose a two-stage multimodal fusion framework that integrates elemental composition and process parameters with image-derived microstructural features to predict hardness.In particular,deep learning is used to extract features from the alloy images at different states,including grain size,second-phase,and precipitate.These features are combined with composition and processing parameters to construct a solid-solution dataset and an aging dataset.The former is used to predict solid-solution hardness,whereas the latter is used to predict the hardness increment induced by aging.The models achieve R2 values of 0.90 and 0.89,respectively.Validated against manual analysis,the two-stage framework provides an automated approach for predicting the final room-temperature hardness of Mg-Gd alloys,thereby reducing manual effort.The proposed framework also provides a new insight for establishing a quantitative and physically interpretable linkage between microstructure and macroscopic property.

Lu Wang;Hongcan Chen;Bing Wang;Yanlin Guo;Qian Li;Qun Luo;Yuexing Han

School of Computer Engineering and Science,Shanghai University,99 Shangda Road,Shanghai,200444,ChinaState Key Laboratory of Materials for Advanced Nuclear Energy&School of Materials Science and Engineering,Shanghai University,99 Shangda Road,Shanghai 200444,ChinaSchool of Computer Engineering and Science,Shanghai University,99 Shangda Road,Shanghai,200444,ChinaGrinm Metal Composites(Beijing)Co.,Ltd,Beijing,101400,ChinaState Key Laboratory of Materials for Advanced Nuclear Energy&School of Materials Science and Engineering,Shanghai University,99 Shangda Road,Shanghai 200444,China College of Materials Science and Engineering,Chongqing University,Chongqing,400044,China National Engineering Research Center for Magnesium Alloys,Chongqing University,Chongqing,400044,China National Key Laboratory of Advanced Casting Technologies,Chongqing University,Chongqing,400044,ChinaState Key Laboratory of Materials for Advanced Nuclear Energy&School of Materials Science and Engineering,Shanghai University,99 Shangda Road,Shanghai 200444,ChinaSchool of Computer Engineering and Science,Shanghai University,99 Shangda Road,Shanghai,200444,China

矿业与冶金

Mg-Gd alloyPrecipitation hardeningDeep learningMicrostructure analysisMechanical property prediction

《Journal of Magnesium and Alloys》 2026 (7)

P.308-330,23

sponsored by theNational Natural Science Foundation of China(Grant No.52273228,52422407).

10.1016/j.jma.2026.102106

评论