Uncertainty Quantification for RANS Predictions of Wind Loads on Buildings

Computational fluid dynamics simulations to calculate wind pressure loads on buildings can be strongly influenced by uncertainty in the inflow boundary conditions and the turbulence model. In the present work we investigate these uncertainties in Reynolds-averaged Navier-Stokes predictions for wind pressure coefficients of a high-rise building, and compare the results to wind tunnel measurements. The uncertainty in the inflow boundary condition is characterized using three uncertain parameters, the reference velocity, roughness length and model orientation, and propagated to the quantities of interest using a non-intrusive polynomial chaos expansion approach. The results indicate that the uncertainty in the inflow conditions is non negligible, but insufficient to explain the discrepancy with the wind tunnel data, in particular where flow separation occurs. The uncertainty related to the turbulence model is investigated by introducing perturbations in the Reynolds stress tensor. The results confirm that the turbulence model form uncertainty is dominant near the separation region that forms downstream of the windward building edge.