Scholarlink

<p>Considering the effect of parameter changes in Finite-Element (FE) models requires repeated dynamic FE simulations, which is very common in fields such as optimisation. Reduced-order modelling is an effective method to alleviate computational costs in simulations.&#xa0;Indirect forced-based methods, such as the Implicit Condensation and Expansion (ICE), are widely used for constructing Reduced Order Models&#xa0;(ROMs), which can capture the dynamics of FE models if the accuracy of static fitting is ensured.&#xa0;In this paper, we show that variations in parameters within an FE model can be efficiently considered by ICE-based ROMs.&#xa0;A parameter analysis of a wing-inspired FE model is conducted to demonstrate the technique.&#xa0;The results show that, when varying parameters only related to nonconservative forces, the ROM can accurately capture the dynamics of the FE model within a wide parameter range whilst achieving high computational efficiency of around 2,000 times that of dynamic simulations in FE software.&#xa0;Regarding parameters related to conservative terms, the ROM needs to be reconstructed to adjust for the variations in these parameters, leading to additional costs.&#xa0;However, the ROMs are still around 10 times faster than dynamic FE simulations for the proposed FE model.&#xa0;Considering the results in the parameter analysis, we demonstrate the computational efficiency of ROMs in gradient estimation, and illustrate their potential in optimisation.</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

Comparing the Efficiency of Finite Element Models and Reduced-order Models with Varying Parameters

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年8月1日

查看全文
  • Xiao Xiao,
  • Thomas L. Hill,
  • Simon A. Neild

摘要

Considering the effect of parameter changes in Finite-Element (FE) models requires repeated dynamic FE simulations, which is very common in fields such as optimisation. Reduced-order modelling is an effective method to alleviate computational costs in simulations. Indirect forced-based methods, such as the Implicit Condensation and Expansion (ICE), are widely used for constructing Reduced Order Models (ROMs), which can capture the dynamics of FE models if the accuracy of static fitting is ensured. In this paper, we show that variations in parameters within an FE model can be efficiently considered by ICE-based ROMs. A parameter analysis of a wing-inspired FE model is conducted to demonstrate the technique. The results show that, when varying parameters only related to nonconservative forces, the ROM can accurately capture the dynamics of the FE model within a wide parameter range whilst achieving high computational efficiency of around 2,000 times that of dynamic simulations in FE software. Regarding parameters related to conservative terms, the ROM needs to be reconstructed to adjust for the variations in these parameters, leading to additional costs. However, the ROMs are still around 10 times faster than dynamic FE simulations for the proposed FE model. Considering the results in the parameter analysis, we demonstrate the computational efficiency of ROMs in gradient estimation, and illustrate their potential in optimisation.