Quantitative analysis of implicit large-eddy simulation: The role of grid resolution

This paper presents a quantitative comparison between implicit large-eddy simulation (iLES) and explicit eddy-viscosity LES (eLES) across multiple grid resolutions. The weakly compressible Taylor–Green vortex is used as a benchmark, and iLES, together with three representative eLES models, is evaluated within a fourth-order finite-volume gas-kinetic scheme. The analysis focuses on dissipation mechanisms and resolved-scale flow structures. Total subgrid-scale (SGS) dissipation is decomposed into explicitly modeled SGS dissipation and intrinsic numerical SGS dissipation, enabling their resolution-dependent roles to be examined separately. The results show that, with grid coarsening, explicit SGS dissipation rapidly loses relative importance, whereas numerical dissipation becomes the dominant SGS mechanism on under-resolved meshes. To assess the direct influence of SGS modeling on the resolved field, probability density functions (PDFs) of velocity gradients are analyzed using direct numerical simulation (DNS) data as a reference. The PDF-based results show that iLES provides consistently robust agreement with DNS across the tested grid resolutions and better preserves the statistical characteristics of resolved velocity gradients overall. In the present high-order framework, intrinsic numerical dissipation provides an effective SGS contribution, allowing iLES to maintain resolved-scale statistics and local flow structures more robustly than the tested explicit SGS models. In contrast, the additional dissipation introduced by the tested explicit SGS closures provides no systematic improvement and, particularly on the coarse and intermediate grids, is associated with larger discrepancies in the resolved velocity-gradient statistics relative to the DNS. These findings indicate that large-eddy simulation assessment should move beyond integral statistics and emphasize grid-dependent model behavior, SGS effects on resolved-scale physics, and the potential of iLES as a robust scale-resolving strategy.