Scholarlink

<p>This paper introduces a new prediction model that utilizes the knowledge of frosting mechanism and proper orthogonal decomposition (POD) to predict heat flux variations due to frosting on a cryogenic vertical cold plate under forced convection conditions. The proposed method divides the heat flux curve into two zones based on the transition point, where the frosting mechanism changes, and utilizes different metamodels for each zone to better capture the nonlinear behavior and the overall trend of heat flux. Performance comparisons using mean absolute error (MAE), interquartile range (IQR), normalized dynamic time warping (nDTW), and relative error of total heat transfer show that the proposed method enhances MAE by 24.4 % and 15.2 %, IQR by 11.4 % and 27.9 %, nDTW by 47 % and 85.5 %, and relative error by 72.6 % and 51.7 % compared to the two previous models. Consequently, the proposed method effectively improves the accuracy and robustness of heat flux predictions.</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

Heat transfer prediction model of cryogenic vertical cold plate under forced convection conditions using the knowledge of frosting mechanism and proper orthogonal decomposition

  • 期刊论文

  • 发表日期：

    2025年4月3日

查看全文
  • Ikhyun Ryu,
  • Hobin Son,
  • Dongheum Ryu,
  • Yongbin Lee

摘要

This paper introduces a new prediction model that utilizes the knowledge of frosting mechanism and proper orthogonal decomposition (POD) to predict heat flux variations due to frosting on a cryogenic vertical cold plate under forced convection conditions. The proposed method divides the heat flux curve into two zones based on the transition point, where the frosting mechanism changes, and utilizes different metamodels for each zone to better capture the nonlinear behavior and the overall trend of heat flux. Performance comparisons using mean absolute error (MAE), interquartile range (IQR), normalized dynamic time warping (nDTW), and relative error of total heat transfer show that the proposed method enhances MAE by 24.4 % and 15.2 %, IQR by 11.4 % and 27.9 %, nDTW by 47 % and 85.5 %, and relative error by 72.6 % and 51.7 % compared to the two previous models. Consequently, the proposed method effectively improves the accuracy and robustness of heat flux predictions.