arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2511.20187cs.CE
收藏

双保真稀疏网格插值驱动局部误差估计器

A bifidelity sparse-grid surrogate model driven by a discrepancy indicator

University of Pisa(比萨大学);Consiglio Nazionale delle Ricerche - Istituto di Matematica Applicata e Tecnologie Informatiche “E. Magenes” (CNR-IMATI)(意大利国家研究委员会应用数学与信息技术研究所)

Matteo Rosellini, Filippo Fruzza, Alessandro Mariotti, Maria Vittoria Salvetti, Lorenzo Tamellini

AI总结:

本文提出一种双保真稀疏网格插值方法,通过局部误差估计器优化拉格朗日稀疏网格的效率,提升高维函数近似精度的同时降低成本。

摘要:

基于拉格朗日多项式的稀疏网格已成为近似高维且计算昂贵的函数的一种主要方法,例如在基于偏微分方程的参数化设计探索中。然而,对于需要局部细化的问题,如目标函数表现出局部特征或陡峭梯度时,它们已知效率低下。虽然基于分段线性多项式的局部细化稀疏网格是解决此问题的已知替代方法,但本文提出了一种策略来提高拉格朗日稀疏网格的局部效率。我们通过逐步构建稀疏网格近似,并仅在建议加入函数评估会显著改变近似景观的插值点上评估函数,从而提高效率。其余插值点则由已有的稀疏网格预测值赋值,即采用双保真方法,在降低成本的同时保持精度。该方法的有效性在多个基准分析函数和一个涉及氢燃料穿孔燃烧器中回火现象的工程应用中得到验证。

英文摘要:

Sparse-grid approximations based on Lagrange polynomials are one of the staple methods for approximating functions that are high-dimensional and expensive to evaluate, e.g. in the context of PDE-based parametric design exploration. They are, however, known to have limited granularity: refining these sparse-grid approximations may require multiple additional function evaluations, potentially leading to computational inefficiency. To address this issue, we propose a selection criterion for the candidate collocation points introduced by the refinement procedure: the function is evaluated only at those points for which a suitable (and crucially zero-cost) discrepancy indicator suggests that the additional evaluation would significantly change the resulting approximation. The remaining collocation points are instead assigned values predicted by the already available sparse-grid approximation, i.e., following a bifidelity approach that reduces costs while preserving accuracy. The effectiveness of this methodology is demonstrated on several benchmark functions and on an engineering application concerning flashback phenomena in hydrogen-fueled perforated burners.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑