Application and analysis of RANS based turbulence models for bluff body aerodynamics

Computational Wind Engineering (CWE) is becoming more popular in the wind engineering community, for prediction of the wind loads on buildings and structures. The most important and basic requirement of CWE is successful modeling of turbulent flow in the simulated atmospheric boundary layer. Many turbulence models have been proposed and tested for a variety of flows, but they are not accurate enough to simulate the bluff body aerodynamics accurately.

In the similar context, five two-equation turbulence models based on the isotropic eddy viscosity concept and Reynolds Averaged Navier Stokes (RANS) modeling approach were tested for the flow around sharp edged building models using Fluent as the solver. The numerical results were compared with the experimental findings in wind tunnel simulations and full scale measurements available in the literature. The results illustrate inaccuracy of the turbulence models to predict the pressure distribution along the building surfaces. The results were analyzed and the effects of different turbulence parameters used in the modeling and analysis were discussed. Each turbulence model was individually reviewed for the correctness of its predictions and the best model in this set was chosen. This model was further applied to different obstacle geometry in different experimental conditions to minimize the effects of unique experimental condition on analysis.

One of the major problems encountered in the modeling was the type of near wall treatment used for simulation. Development of equilibrium boundary layer in the wind tunnel was used as a test condition for studying the deficiencies in different types of near wall modeling approaches. It was concluded that the standard approach of wall functions was insufficient to estimating the effects of roughness on the wall boundary on the mean flow. A new rough wall model proposed by Durbin et al [58] was implemented using the user defined functions capability of Fluent. This model was tested on a variety of flows; like flow through pipe, flow over a backward facing step, etc. This flow was then applied to the models tested in this project to analyze the effects on the rough wall model used for modeling near wall effects. Overall agreement between the computational predictions and experimental findings was acceptable.