Learning Forum

[senthilathiban]

Hi all, 

I have done a simulation in RANS and LES using ANSYS Fluent 18.0 and compared with experimental results. The RANS results seems good, while the LES results are unacceptable. Please see the attached pdf. Any help to rectify the error is appreciated. 

 

Thanks,

senthil. 

[Rob]

As staff we're unable to open attachments, so I've not looked at the results. 

Have you used a suitable mesh and time step in the LES calculation? Did both models converge to a good level? Is the experimental data good (always worth checking)? 

[senthilathiban]

computational domain.

 

LES.. time averaged at 65 sec. 

 

The above image shows a velocity probe near the point (0.6,0.2,0.34), 0.6 m distance from the inlet. As it can be seen that the velocity converges nearly to 0.225 m/s which equals the inlet velocity. There is no reduction in the velocity throughout the flow and as the period of simulation increases, the time averaged velocity also increases. This is the main problem in LES simulation.

 

The mesh is fine. The experimental results have been validated by many researchers in RANS simulations. I am trying to validate it in LES.

[Rob]

Have you got the same plot from the RANS model?  Please can you also post the residual plots. 

[senthilathiban]

Rans results

 

 

RANS residuals

 

 

 

LES residuals.

 

[seeta gunti]

Hello,

Did you use the same mesh for RANS and LES? 

What kind of LES model did you use? I guess it is a confined flow. If it so, you can use either Dynamic LES or WMLES model and mesh needs to be very very fine.

How did you do the sampling?  We have some best practices to run LES model. WE have few tutorials on LES. Please go through the below tutorials.

Regards,

Seeta

[Keyur Kanade]

As mentioned earlier, the mesh and time step are important factors here. 

Can you please insert some images for mesh used. 

[DrAmine]

I want first of all recommend having a look into the Best Practice Document for Scale Resolving Simulations. Parts of it could be find in the CFX Best Practice Documentation.

[senthilathiban]

Mesh details.

 

 

 

 

The mesh was made for LES - Smagorinsky lilly model with time step 0.01 (Courant number stays less than one).  And the same mesh was used for RANS simulation.

[DrAmine]

Again I recommend going through the SRS guidance. Temporal and spatial resolution  for LES are by fore more strict then the ones for RANS.