Abstract

The wind environment around buildings and the wind pressures on building surfaces are studied by scale-model experiments in wind tunnels and confirmed by full scale measurements. Such experiments are often expensive and time consuming. Computational wind engineering as a new branch of computational fluid dynamics has been developed recently to evaluate the interaction between wind and buildings numerically. In the current study, a systematic examination of wind effects on buildings and wind flow conditions around buildings has been carried out numerically. Contrary to the usual numerical evaluations which were only performed on rectangular buildings, the current study evaluates the wind effects on buildings of different shapes such as L-shapes and Z-shapes. The steady state Reynolds averaged Navier-Stokes equations and the k-$\varepsilon$ turbulence model have been adopted for the numerical studies. These equations have been solved with the SIMPLE method. Some researchers declared that the discrepancies between the computed results with the k-$\varepsilon$ model method and the experimental data on the flat roof of a rectangular building are caused by the coarse grid used. To check this, a systematic evaluation of the k-$\varepsilon$ model method in predicting the wind pressure on flat roofs has been attempted by using grids of various densities. Computations were made for a low building and a taller building. Both the advantages and limitations of this most widely used method in computing the wind pressure on flat roofs under normal and oblique wind conditions have been discussed. This study revealed that these discrepancies can be attributed not only to the coarse grid arrangement but also to the k-$\varepsilon$ model itself. To keep the advantages of k-$\varepsilon$ model in representing the fully turbulent flow in the external region far from solid walls and to avoid its shortcomings for the near wall flows, a two-layer methodology combining the k-$\varepsilon$ model in external flow region with either a one-equation model or a modified k-$\varepsilon$ model in the near wall area has been adopted in this study to predict the wind conditions around a cubic building. The two-layer method based on the modified k-$\varepsilon$ model has not been found effective. The two-layer method based on a one-equation model, however, has been proved very effective in predicting the separation above the roof surface and near the side walls of a cubic building which was not possible with the usual k-$\varepsilon$ model method; as a result, the prediction of the wind pressure on the roof and side walls has also been apparently improved