Accurate prediction of unsteady flows past cylindrical structures is critical for marine engineering applications, yet the reliability of large eddy simulation (LES) in such contexts remains insufficiently quantified. This study develops and applies a comprehensive verification and validation (V&V) framework to quantify predictive uncertainties in LES of three-dimensional flow around a stationary cylinder. An in-house solver integrating the boundary data immersion method with Cartesian grids and the Smagorinsky subgrid-scale (SGS) model is employed. Through systematic grid refinement across five configurations at Re numbers from 80 to 500, the total simulation error is decomposed into the numerical discretization error and the SGS modeling error. Results show that numerical error constitutes the dominant uncertainty source, accounting for over 75% of the total error magnitude, while a compensatory trend between the two error components is observed. The predicted Strouhal numbers agree with experimental data within uncertainty intervals of ±0.004 to ±0.017 and velocity profiles in the far wake fall within the estimated error bands. The framework establishes rigorous accuracy benchmarks for bluff body flow simulations and provides a transferable methodology for error quantification in computational fluid dynamics, representing a novel contribution to V&V applications in LES of cylinder flows.
Zhao et al. (Wed,) studied this question.