Scholarlink

<p>The prediction of mechanical strength using morphological scanning electron microscopy (SEM) images would accelerate the development of bone-substitute materials. However, only a few studies have confirmed the relationship between the sintering temperature and surface properties of sintered bodies by analyzing their SEM images. Therefore, in this study, we attempted to classify and predict properties based on SEM images of β-tricalcium phosphate (β-TCP), which was liquid crystal display (LCD) 3D printed and sintered, by deep learning, for the purpose of applying it to the optimization of sintering temperature and estimation of mechanical strength. Synthesized β-TCP powder was mixed with light-curing resin and polyethylene glycol and poured into a vat of the LCD 3D printer. The CAD data of the gyroid structure with interconnecting holes was prepared for 3D printing. The printed green bodies were sintered at various temperatures for 3&#xa0;h to obtain sintered bodies. As the sintering temperature increased, the degree of sintering of β-TCP progressed, the size of the sintered body decreased, and the maximum compressive strength increased. It was found that the clustering of the SEM images of 3D printed β-TCP sintered at different temperatures can be made by transfer learning and hierarchical cluster analysis. Images of the sintered β-TCP surface can be classified by deep learning (transfer learning) and cluster analysis. The constructed model in this study successfully predicted the compressive strength of the 3D printed sintered bodies from their SEM images. The proposed deep learning method can be used for future surface property analyses of ceramic materials and sintering condition optimization.</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

Mechanical property prediction of 3D-printed bioceramics by deep learning for sintering condition optimization

  • 期刊论文

  • 发表日期：

    2026年5月11日

查看全文
  • Sayaka Ito,
  • Yuta Otsuka,
  • Hiroshi Kono,
  • Kazuyuki Noguchi,
  • Masafumi Kikuchi

摘要

The prediction of mechanical strength using morphological scanning electron microscopy (SEM) images would accelerate the development of bone-substitute materials. However, only a few studies have confirmed the relationship between the sintering temperature and surface properties of sintered bodies by analyzing their SEM images. Therefore, in this study, we attempted to classify and predict properties based on SEM images of β-tricalcium phosphate (β-TCP), which was liquid crystal display (LCD) 3D printed and sintered, by deep learning, for the purpose of applying it to the optimization of sintering temperature and estimation of mechanical strength. Synthesized β-TCP powder was mixed with light-curing resin and polyethylene glycol and poured into a vat of the LCD 3D printer. The CAD data of the gyroid structure with interconnecting holes was prepared for 3D printing. The printed green bodies were sintered at various temperatures for 3 h to obtain sintered bodies. As the sintering temperature increased, the degree of sintering of β-TCP progressed, the size of the sintered body decreased, and the maximum compressive strength increased. It was found that the clustering of the SEM images of 3D printed β-TCP sintered at different temperatures can be made by transfer learning and hierarchical cluster analysis. Images of the sintered β-TCP surface can be classified by deep learning (transfer learning) and cluster analysis. The constructed model in this study successfully predicted the compressive strength of the 3D printed sintered bodies from their SEM images. The proposed deep learning method can be used for future surface property analyses of ceramic materials and sintering condition optimization.