Scholarlink

<p>Honeycomb structures are highly effective in shock and vibration isolation due to their light weight and small material usage. However, they suffer from plastic buckling after deformation, preventing reuse. To address this issue, researchers have focused on negative stiffness honeycomb (NSH), which feature curved-beams that allow for elastic buckling and snapthrough behavior, enabling reuse after deformation. Existing studies primarily employ theoretical approaches and finite element analysis (FEA), but they do not consider performance prediction using both FEA and experimental data under manufacturing and experimental uncertainties. Therefore, this study develops a multi-fidelity modeling approach by combining low-fidelity FEA data and high-fidelity experimental data to reduce experimental costs while addressing manufacturing and experimental uncertainties. This approach is applied to a modified Bayesian classifier to identify negative-stiffness (NS) regions of an NSH unit cell and subsequently to a kriging model to predict its specific energy absorption (SEA). The multi-fidelity model achieved a mean absolute percentage error (MAPE) below 5 % while requiried 11 fewer experimental data points compared to single-fidelity models, demonstrating enhanced prediction accuracy and efficiency.</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

Performance classification and prediction of additively manufactured negative stiffness honeycomb unit cell using a multi-fidelity model

  • 期刊论文

  • 发表日期：

    2026年1月15日

查看全文
  • Taemin Noh,
  • Hyungdo Kim,
  • Young-Jin Kang,
  • Seunghun Baek,
  • Yoojeong Noh

摘要

Honeycomb structures are highly effective in shock and vibration isolation due to their light weight and small material usage. However, they suffer from plastic buckling after deformation, preventing reuse. To address this issue, researchers have focused on negative stiffness honeycomb (NSH), which feature curved-beams that allow for elastic buckling and snapthrough behavior, enabling reuse after deformation. Existing studies primarily employ theoretical approaches and finite element analysis (FEA), but they do not consider performance prediction using both FEA and experimental data under manufacturing and experimental uncertainties. Therefore, this study develops a multi-fidelity modeling approach by combining low-fidelity FEA data and high-fidelity experimental data to reduce experimental costs while addressing manufacturing and experimental uncertainties. This approach is applied to a modified Bayesian classifier to identify negative-stiffness (NS) regions of an NSH unit cell and subsequently to a kriging model to predict its specific energy absorption (SEA). The multi-fidelity model achieved a mean absolute percentage error (MAPE) below 5 % while requiried 11 fewer experimental data points compared to single-fidelity models, demonstrating enhanced prediction accuracy and efficiency.