Wall-modeled large-eddy simulation of channel flow via recursive regularized lattice Boltzmann method

This study presents a wall-modeled large-eddy simulation (WMLES) framework that integrates a recursive regularized lattice Boltzmann method (RR-LBM) with a quasi-analytical wall function, mitigating numerical instability and alleviating the stringent near-wall grid-resolution demands inherent in high-Reynolds-number, wall-resolved simulations for wall-bounded turbulence. The WMLES approach employs either the wall-adapting local eddy-viscosity (WALE) or Smagorinsky models for subgrid-scale closure. Direct numerical simulations of turbulent channel flow were conducted at friction Reynolds numbers of Reτ = 180 and 395 using the RR-LBM and validated against benchmarks, confirming its accuracy. Based on the validation, WMLES simulations were performed to assess the feasibility of the method and to clarify the effects of the WALE coefficients and subgrid-scale closures on wall-bounded turbulence. The WMLES predictions remain consistent with benchmarks at Reτ = 950–20 000 with reduced number of grid nodes by an order of magnitude, demonstrating its viability at high Reynolds numbers. In the wall-modeled RR-LBM framework, the WALE model coefficient exhibits a Reynolds-number dependence with values of 0.06 (Reτ = 950), 0.20 (Reτ = 2000), 0.30 (Reτ = 10 000), and 0.35 (Reτ = 20 000), indicating that a single universal constant value is inadequate within the examined range 950 ≤ Reτ ≤ 20 000. Furthermore, the WALE model outperforms the Smagorinsky model for 950 ≤ Reτ ≤ 20 000, whereas the Smagorinsky model underestimates mean velocity by 8.1% at Reτ = 20 000, exhibiting excessive dissipation and necessitating additional wall treatment in practice. The proposed wall-modeled RR-LBM enables statistically accurate WMLES of high-Reynolds-number wall-bounded turbulence without the wall-resolved grids.