用时空方法和各向异性网格自适应模拟粘性指进
Simulating Viscous Fingering with a Timespace Method and Anisotropic Mesh Adaptation
AI总结:
本文提出一种时空方法和各向异性网格自适应技术,用于解决二维粘性指进问题,通过避免启发式和插值误差,提高时间步长精度,从而提升计算效率。
摘要:
我们报告了关于使用时空方法和各向异性元素解决二维粘性指进问题的发现。时空方法因时间维度的并行性而受到关注,但在各向异性网格自适应方面也有吸引力;不仅避免了启发式和插值误差,而且斜向的时空元素对应于长且准确的时间步长,即时空中的各向异性可以被利用。我们表明,我们的时空方法受限于最小时间步大小,这由于数值扰动的增长。然而,这个下界相当高,这表明对于实际应用,时间步数可以减少几个数量级。
英文摘要:
We report findings related to a two dimensional viscous fingering problem solved with a timespace method and anisotropic elements. Timespace methods have attracted interest for solution of time dependent partial differential equations due to the implications of parallelism in the temporal dimension, but there are also attractive features in the context of anisotropic mesh adaptation; not only are heuristics and interpolation errors avoided, but slanted elements in timespace also correspond to long and accurate timesteps, i.e. the anisotropy in timespace can be exploited. We show that our timespace method is restricted by a minimum timestep size, which is due to the growth of numerical perturbations. The lower bound on the timestep is, however, quite high, which is indicative that the number of timesteps can be reduced with several orders of magnitude for practical applications.
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。