K-Epsilon VS Shear Stress Transport (SST) on the same model (Comparison)

Today I will write a few words and I will present you an example I've made regarding two methods for turbulence modeling, K-Epsilon and Shear Stress Transport.

For this experiment, I've used a intentionally badly designed diffuser (so I could create as much eddies as possible) with an inlet inner diameter of 30 [mm] and a sharp step driving the inner diameter to 45 [mm].
  

Ok, so let's present the 1st simulation.

The first mesh attempt will be with a generated "fine" discretization,  max element size = 4 [mm]. As for the inflation I've used the following settings: first layer thickness = 0.003 [mm], inflation algorithm = pre (this is making the inflation layers first) and maximum layers = 10.

For the setup of the analysis, I've used as fluid model: "Air Ideal Gas", as I defined my domain (instead of the default domain). In this domain, we must have the Inlet and the Outlet, that are defined as boundaries. Also, the turbulence model was selected as K-Epsilon and after the first simulation it was changed to Shear Stress Transport (It will be seen in the results).

For the inlet, at "Mass and Momentum", Mass Flow Rate was used as boundary condition for the air flow. (0.166 kg/s = 600 kg/h)

 For the outlet, at "Mass and Momentum", we have to pick Avrage Static Pressuve and Relative Pressure = 1 [bar].

The results:

Here are the results (CFD Post) represented with a streamline (starting Inlet) for the analysis with the K-Epsilon method. This simulation ran for 750 iterations, and although the residuals didn't really got above the threshold, it looked very steady.

Here are the results (CFD Post) represented with a streamline (starting Inlet) for the analysis with the Shear Stress Transport method. This simulation ran for 750 iterations, and although the residuals didn't really got above the threshold, it looked very steady.

CONCLUSION:

It is pretty obvious that the flow is very different from a method to another. Keep in mind that the mesh was exactly the same (100%) in the both analysis. 

By comparing the two flows we can see that the K-Epsilon method is showing a more steady flowing line and almost no eddies. 

If we look at the simulation that ran with Shear Stress Transport method, we can observe a lot more eddies and quite larger. 

This means, either that K-Epsilon method is not the best to use when large eddies are expected, or that SST method is giving exaggerated results. We will prove which method is the best by refining the mesh of the model and running the simulations with a lot more iterations.

............................................................................................................

After refining the mesh, we got this. We ran the simulations again and this is what we got as results.

 CONCLUSION:

As I've said, I'm not a professional, I am learning, but from the results, I think the K-Epsilon method should not be used when it is obvious that high turbulence is going to be created. 

Even with the refined mesh, the Shear Stress Transport method showed no big difference, but the K-Epsilon method was totally different.

This is all for now, I posted the videos of the simulation on youtube, you can check them out. Enjoy. Feel free to ask me, tell me I'm wrong or anything. Have a nice day!!