Scholarlink

<p>UNet-CFD model, incorporating a Logically Extended-Integrated (LEI) dataset and flow feature functions, is proposed to predict tundish flow field. The key idea of LEI dataset comes from the turbulence theory, and the flow feature functions include fluid region function and boundary distance function. UNet model can capture small vortices accurately, achieve MAPEs of 2.08&#xa0;pct (<i>u</i>) and 1.52&#xa0;pct (<i>v</i>). Training and prediction times are only 492 and 2.03 seconds, respectively.</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

Generative Flow Field in the Tundish by UNet-CFD Architecture with LEI Dataset

  • 期刊论文

  • 发表日期：

    2025年7月28日

查看全文
  • Yili Sun,
  • Hong Lei,
  • Changyou Ding,
  • Haoyu You,
  • Qiang Li,
  • Yan Zhao

摘要

UNet-CFD model, incorporating a Logically Extended-Integrated (LEI) dataset and flow feature functions, is proposed to predict tundish flow field. The key idea of LEI dataset comes from the turbulence theory, and the flow feature functions include fluid region function and boundary distance function. UNet model can capture small vortices accurately, achieve MAPEs of 2.08 pct (u) and 1.52 pct (v). Training and prediction times are only 492 and 2.03 seconds, respectively.