Revision as of 10:29, 29 October 2025

Abstract

In this paper, we investigate the performance of different boundary condition (BC) schemes for curved walls within the framework of the Lattice Boltzmann Method (LBM). A canonical benchmark problem—the flow past a circular cylinder in a channel—is considered, with Reynolds numbers ranging from 50 to 300. While prior studies have examined similar configurations, this work provides a novel comparative analysis under realistic conditions—using dimensional LBM simulations with actual fluid properties and consumer-grade hardware, rather than idealized lattice units. Additionally, we introduce an in-house GPU-accelerated solver, enabling efficient high-fidelity simulations without reliance on specialized computational resources. Four wall boundary conditions—the standard bounce-back scheme, the non-equilibrium extrapolation scheme, the fictitious equilibrium scheme and a one-point scheme—are implemented and analyzed through their influence on the time-averaged drag coefficient of the cylinder. The results are compared against both experimental and Navier-Stokes-based numerical data to assess accuracy. Additionally, the study evaluates the relative impact of outlet BC selection on simulation fidelity. The findings show that all tested solid wall boundary schemes can produce reasonable predictions under suitable conditions. Furthermore, based on our results, the accuracy of LBM simulations is notably more sensitive to the choice of the outlet boundary condition when compared to the choice of the ones used at the immersed body.OPEN ACCESS Received: 08/05/2025 Accepted: 24/06/2025 Published: 27/10/2025

Document