Spatial network models are used as a simplified discrete representation in a wide range of applications, e.g., flow in blood vessels, elasticity of fiber based materials, and pore network models of porous materials. Nevertheless, the resulting linear systems are typically large and poorly conditioned and their numerical solution is challenging. This paper proposes a numerical homogenization technique for spatial network models which is based on the Super Localized Orthogonal Decomposition (SLOD), recently introduced for elliptic multiscale partial differential equations. It provides accurate coarse solution spaces with approximation properties independent of the smoothness of the material data. A unique selling point of the SLOD is that it constructs an almost local basis of these coarse spaces, requiring less computations on the fine scale and achieving improved sparsity on the coarse scale compared to other state-of-the-art methods. We provide an a-posteriori analysis of the proposed method and numerically confirm the method's unique localization properties. In addition, we show its applicability also for high-contrast channeled material data.
翻译:空间网络模型在广泛的应用领域,例如血管流动、纤维材料弹性和孔径网络多孔材料模型中被用作简化的离散代表,然而,由此产生的线性系统通常规模大,条件差,而且其数字解决方案具有挑战性。本文件建议对空间网络模型采用基于超局部矫形分解(SLOD)的数字同质化技术,该技术最近为椭圆多尺度局部分异方程引入。它提供了精确的粗略溶解空间,其近似特性与材料数据的顺畅无关。SLOD的一个独特的销售点是,它几乎就地建造了这些粗糙空间,比其他最先进的方法要求较少的微量计算,在粗粗度上实现更宽度的宽度。我们对拟议方法进行了表面分析,并以数字方式确认了该方法独特的本地化特性。此外,我们还展示了该方法对于高频通道材料数据的可适用性。