On the Variability of Results for Cylinder Flow: A Systematic Analysis of Parameter Sensitivity Within the Lattice Boltzmann Method

ABSTRACT

In computational fluid dynamics (CFD), obstacle flows are a pivotal benchmark, providing essential insights into complex fluid dynamics phenomena. CFD frameworks extensively employ these flows to analyze boundary conditions, offering a comprehensive understanding of fluid‐structure interactions. Throughout the development of CFD over the last decades, the stationary circular cylinder has been one of the most frequently simulated obstacle flows, combining simplicity for methodological work and similarity as well as transferability to practical applications. Despite its prevalence, published results exhibit significant and sometimes unexplained variations. This study addresses this issue by conducting a thorough investigation into how sensitive simulations are to system and model parameters, using the example of the lattice Boltzmann method. We summarize the state of the art and conduct additional parameter studies to examine the sensitivities of essential metrics such as force coefficients, Strouhal number, average velocity, and Reynolds stress profiles across a range of Reynolds numbers encompassing laminar and turbulent flows. Our findings indicate non‐monotonous convergence and fluctuations in response to grid resolution changes, with parameter choices becoming increasingly critical at higher Reynolds numbers. Furthermore, we captured the transition from two‐dimensional periodic to inherently three‐dimensional flow, underlining the necessity of 3D simulations even in a geometrically 2D environment for Reynolds numbers at or above the transition regime.