Evaluation of various non-linear k–ɛ models for predicting wind flow around an isolated high-rise building within the surface boundary layer

This paper evaluates the performance of various non-linear k–ɛ models for predicting wind flow around an isolated high-rise building at a 1:1:2 (length: width: height) scale and within the turbulent boundary layer. Two cubic models, proposed by Ehrhard et al. and Craft et al., and one quadratic model proposed by Shih et al., are examined by comparing their simulation results with the experimental data. In this study, the three non-linear models can reproduce the vortex shedding behind building during unsteady calculations. However, these non-linear models cannot achieve a convergent solution in steady calculations. For the Shih model and the Ehrhard model, the magnitudes of the lateral velocity fluctuations normalized by the inlet wind speed at the height of building are only approximately 0.022, and for the Craft model, this dimensionless magnitude increases to approximately 0.11. These three non-linear models can reproduce the reverse flow on building roof. The Craft model slightly overpredicts the reattachment length on building roof. The other two non-linear models produce large roof re-circulation vortexes, which extend over the entire roof without attachment. Among the RANS models studied here, the Craft model shows the best agreement with the experimental data. The Craft model predicts the shortest reattachment length behind building among the revised k–ɛ models, while the other two non-linear models greatly overestimate the reattachment length behind building. The good performance of the Craft non-linear model in the wake region can be interpreted from the magnitude of the total kinetic energy after considering the solved wind fluctuations.