Most accurate turbulence model
 

Hi everyone,

I've got one question to ask you for. I know what are the general guidelines to choose the most accurate turbulence for some cases and meshes. But, according to that, I'm not getting a reasonable result.

I mean, I have an exterior flow over a wall. Re number is over 1e6, and the fluid is air. The problem is that the boundary layer is much larger than I expect (more less 1 m, in a Lx= 21 m, Ly = 6m elliptical shaped wall). I chose standard k-e model with standard wall functions. It is expected that in one point the flow will separate. I know there is a relation between the model and the mesh used related to the value of y+ expected. How can I calculate initially the expected value of y+?

and what is the model that, by experience, you think is the best for such exterior hi Re flows? I'd like your personal opinion, not a link to a guide document. I have enough guide docs.

Thanks in advance and have a great day!

Re: Most accurate turbulence model
 

Hi, I can´t answer the y+ stuff but for the model, I used the Spalart Allmaras model for external flow and gave me goods results. This model was created for aerospace applications, so it should handle very well that Reynolds.

Hope this helps!

Take care.

Re: Most accurate turbulence model
 

Hi,

Fluent manual has a lot of good stuff on turbulence model choice. Turbulence model choice should be based on your computational power, the nature of the flow, and what features you want to resolve. Realizable K-E is generally a very good option. With 30<y+<300, you should use the non-equilibrium wall functions for best results. If you have a simple enough geometry to get a y+ or around 1, SST KW will handle transitional flows as well. Read the documentations for more in depth information about the options available, as there's a hell of a lot there. Hope this helps,

Andy

Re: Most accurate turbulence model
 

Hi, I usually use a excel sheet to have an idea of the mesh refinement in order to get y+ of one. This is based on the flat plate flow without pressure gradient, but afterwards I multiply the results by a secure factor to take into account the curvature and gradients effects. I just input the Re, the desired y+, the reference Length and I get the distance from the wall of the first point and the total length. On your post you wrote 1 m off boundary layer, this is corrected or you meant the wake width? If it is the BL this is really thick for L = 21m ! In relation to the turbulence models I would use SST, Cubic Ke, SA, and EARSM in this order.