[RosaRichárd]

Hello everyone!

I have an atmospheric boundary layer (ABL) related question to the community. I am analyzing the flow over a doubly curved surface for wind engineering. I have analyzed multiple wind directions, and the results are compared to wind tunnel test (WTT) results. The surface consists of multiple bays or sections, and each section is connected to the other through a sharp edge. This is an arch-supported tensile membrane surface, and the arches represent sharp edges in the CFD analysis. (I also attached a figure)

The CFD results are based on RANS simulation using k-omega SST turbulence models.

The results are generally in good accordance with the WTT results in every wind direction (0-90°, in 10° increments), but I have a problematic area at a certain part of the surface. The Cp values are calculated using CFD and WTT, and the biggest errors occur behind the last arch/edge at the "leeward side", as you can see on the figure.

The boundary layer on the body's surface is resolved, the y+ is generally less than 5, and I have carried out grid independence study and went down to y+ below 1 on the whole surface (altough the element number was pretty high). I also applied inflation on the bottom wall of the domain because of the ABL.

Some observations: The grid independence study did not show significant improvements on the results. The application of inflation layer and low Re formulation on the bottom ground also did not show any improvement. Only the resolution around the surface mattered.

I do not have any idea what the problem could be, since other parts of the surface have a really good agreement between CFD-WTT, only the last section on the leeward side behind a particular part of the edge.

Can this be solved by tuning the parameters of the turbulence models or applying models other than k-omega SST? Also, the inlet profiles are based on the WTT tests, the velocity, turbulence intensity (the turbulence kinetic energy) and the omega profile was calculated based on the applied wind profile. Or maybe this is the limitation of RANS and turbulence modelling? The results of the WTT can be seen on the figure as well (right), and you can see that the fluctuation components (RMS) are the largest at the problematic zone, meaning a high turbulence at that locations.

Best reqards, RR