interesting topic can provide more information

Tariq Ullah if you can possibly do...

Thanks

Yes for 3-D Simulation SA model is the best option (although you will need grid Independence study). However, in 2D case, K-omega SST model can capture flow around the airfoil surface as well as in the far field region. Amjid Khan

Thanks Tariq Ullah

I agree with Aditya Vedanth ,

It is better to make a literature study to know which turbulence model gives the best results for your case....

The values of y+ are dependent on the resolution of the grid and the Reynolds

number of the flow, and are defi ned only in wall-adjacent cells. The value of y+

in the wall-adjacent cells dictates how wall shear stress is calculated. When you

use the Spalart-Allmaras model, you should check that y+ of the wall-adjacent

cells is either very small (on the order of y+ = 1), or approximately 30 or

greater.

What is your tool/software?

you need a wall model instead of wall function if the Y is low.

Besides, S.A model is more and industrial model with one transport equation to be solved and it is not recommended for 3D and twisted domains with expected secondary flows, swirl flows that possibly exist in the modelling of a wind turbine. This is another point to be considered.

Tariq Ullah Nazaruddin Sinaga , according to the literature, k-omega SST gives good results even for 3D simulation. But, I used S-A because I wanted to test its validity for my case.

Ks Adamsab the problem is that when I wanted to achieve this value of y+ by reducing the first layer thickness, the solution diverge.

Ramin Moradi I use Ansys Fluent ... Yes I have a 3D twisted blade and a rotating domain. So, I think that the source of the divergence is not due to a mesh problem but it is rather due to the turbulence model properties and limits.

Thank you all for your answers and interaction

Somaya Younoussi I doubt that the model is the cause of divergence (however it can be). It would be the wall function/model that you are using when the mesh is refined near walls. If Fluent does not provide wall model and only wall function is available, higher y+ (above 30 for eg) should be considered.

Ramin Moradi , Fluent doesn't provide a wall function for S-A. That is why the mesh must be refined near the wall.

Prueba cambiando los indices de convergencia de malla, o definitivo de condiciones

I agrre with Khadersab answer

Hi Somaye

I guess you are validating your work by another paper that it included experimental and numerical works(Numerical : based on CFD and S-A turbulence model). But , there is no insisting to use of A-S model for your validation. You can use K-W sst model, it is widely used to simulate turbo-machine and you will have good results. Finally you can compare your CFD work by mentioned paper(you are validating your work) and the experimental data. Also the K-w-sst has more accuracy respect to A-s model and more acceptable for your work too.

The divergence of the numerical scheme may come from various sources. Primarly, when you refine the grid near the walls (in the normal direction) the aspect ratio of the cells gets prohibitivly large. It causes instability of the simulation in majority of the CFD codes... Try to compute the aspect ratio of the cells, and later reduce this aspect ratio to a level that was present prior to grid refinement. It would require to alter the number of cells and spasings in one of the other directions than normal to the wall...

Good luck,

Oskar

Oskar Szulc thanks for your answer, I will try this solution and I will tell you if it works with my case or no..

Ramin Alipour , yes I understand what you mean, K-omega SST is used also in the paper that I based on to make my own simulation. My object was to compare between these models that is why I used S-A in my simulation.

Check the mesh and the boundary conditions, your problem may be in one of them.

Good luck@

Somaya Younoussi but to get y+~1 you will need K-omega SST model.

Dear @ Somaya Younoussi

I am not sure if using this model for your purpose could be an appropriate choice. At first, you have to determine the flow field Reynolds number as the most important factor for addressing turbulence in the domain using your model length. Then you could decide which model suits your simulation. K-epsilon is a high Reynolds number model, while K-omega is a low Reynolds number model. A combination of those SST K-omega is enabled to implement k-omega for the surface such as the boundary layer where the local Reynolds number is low and employs to use K-epsilon for farther regions with high Reynolds number. In fact, this capability added to this model by means of switching based on the Reynolds number. As all friends perfectly mentioned this model could be useful for your purpose since it could capture separation of the flow on your body and you could either employ and also not employ Y+ or wall function. If you do not want or cannot employ wall function, you need to start grids right over the surface but before that, you should obtain the boundary layer thickness and viscous sublayer thickness using this formula delta=11.6 *viscosity kinematic/U*, U*=(near-wall shear force/density)^1/2. Then consider 2 layers of mesh for this layer and by 1.2 or 1.3 expansion factor continue your grid toward the boundary layer and onset of turbulent layer, otherwise, you might get the wrong results.

Fluent does have wall function for SA model also. However, please follow guidelines mentioned in its theory manual.

https://www.afs.enea.it/project/neptunius/docs/fluent/html/ug/node410.htm

SA as turbulence model is not good if first cell next to the wall lies in the buffer layer. Either have y+30.

Increase number of cells, decrease CFL number and keep trying. That's all. If it converged earlier, it'll converge now.

If you want to make sure that your mesh is fine enough near the walls, then mesh the geometry manually. This means that start by meshing the elements placing the desired number of point near walls, then surfaces then volumes. SST model is fine but I suggest to start your run as laminar flow, then switch to k-epsilon then SST.

Spalart-Allmaras turbulence model is limited to situations and has limitations that are listed in FLUENT documentations.