CFD Reliability Issues in the Prediction of Airflows in a Naturally Ventilated Building

The potential for prediction error when using computational fluid dynamics (CFD) for investigating internal building airflows is assessed in the current paper. The ability of a proprietary CFD code, CFX, to simulate buoyant and forced airflow regimes, typical of a naturally ventilated building, are investigated using two experimental case studies from the literature. Comparisons are then made between simulated and measured airflows for a naturally ventilated building. Results from the two case studies indicate that structured meshes are less dependent on mesh density and yield consistent convergence and accuracy when coupled with the k-e or k-? turbulence models. Comparison of CFD predicted airflows with the full-scale building airflows was challenging due to scatter in the measured data. A structured mesh in conjunction with either the k-e or k-? turbulence models, showed reasonable correlation with the measured airflows, with both models performing equally.