CFD Online Discussion Forums

CFD Online Discussion Forums (
-   CFX (
-   -   LES and SST together (

Roland R December 17, 2009 14:26

LES and SST together

I have the next problem:

I would like to study high-frequency pressure fluctuation to an aeroacoustic investigation. I have simulated a simple 3D model which has shown that the LES method is really the most suitable to this study.

But my original flow field is too large, and complicated, I'm sure that it is impossible to calculate the total domain with LES. I have tried the DES, and in the result based on the blending function the solver has applied the LES in environment of the separation and the small vortices, but the frequency of press.fluctuation was more lower than in case of the full LES.

Finally I tried to divide the flow field into two domains with an interface. I have defined the LES in the inside domain , and for the outside domain the SST, but the solver didn't not start...It has stopped after first iteration of first timestep.

Can anybady help me?



Tristan December 17, 2009 21:54


It sounds like you're trying to do your own DES without actually using the DES capability in the code. Do you believe this will give better results than the actual DES did? If you set up your run as you described, CFX would need to know how to transfer/make-up the appropriate variables at the interface since the LES and the SST don't solve for exactly the same variables. This is probably not trivial and is the reason they provide the DES option to you.


ghorrocks December 17, 2009 22:06

As Tristan said what you are planning is a major undertaking. Have you checked the DES or SAS options in CFX are suitable for you?

Roland R December 18, 2009 12:25

Hello Glenn and Tristan,

Thank your answers.

I try to detail my problem better.

I know that the application of "own DES code" is not suitable. By the way I have seen this approach in a study where an airfoil was investigated in terms of aeroacoustic. The flow field was divided into two domains for LES and SST, like in case of my model, therefore I tried to calculate with this solution.

I think that the using of DES method is more suitable. You can see the result to DES on the next pictures:

The invastigated model:
(The pressure was evaluated in the yellow point)
A képet a Képfeltölté tárolja.

Pressure in case of LES model:
A képet a Képfeltölté tárolja.

Pressure in case of DES model:
A képet a Képfeltölté tárolja.

I don't understand because in case of DES based on the bland factor the LES was used by solver in environment of the monitor point and the separated region.:
A képet a Képfeltölté tárolja.

In short I don't understand why does the DES damp the high-frequency fluctuations?


Tristan December 18, 2009 15:26


I would not expect a DES solution and an LES solution to match just because the point where you are making the comparison is in the LES region of the DES. The DES is cheaper, which is why you want to use it, but you're not going to resolve all the same flow features that you get in the LES because you weren't willing to pay for that level of resolution. Think about DNS and LES, you can't just compare LES results to DNS results and expect them to be the same, you have to filter the DNS results. This filtering causes you to lose smaller scale information. A DES is like a filtered LES so you're going to have the same thing happen i.e. some of the flow structures, which in this case provide pressure fluctuations, are not going to be captured as well as they would be in an LES. You have to decide if this loss of resolution is worth the gains you achieve in reduced simulation time.


Roland R December 19, 2009 08:25

Hello Tristan,

I understand what you has said, but it is not clear for me how does DES operate. Based on my last pictures (bland function contour) the LES was applied in blue range. As you has said, the DES method is a filtering. I have thought that this filtering is valid only in the red DES range where RANS was used and by this means the simulation time will be shorter, and the method will be more economical.

However the press.fluctuation shows that the filtering comes into operation in the blue LES range too...

To sum it up, If I understand well, the LES calculates in the blue range, but this region is filtered partially too because of the effect of filtering and damping of the red DES range. In short antecedents of the flow (which develops in the red DES range) has an effect on the LES range to a great extent, therefore the result will be different from the full LES solution.

Am I correct?


Tristan December 21, 2009 13:26


You got it. Basically, there's no free lunch. When you do a more detailed simulation; DES instead of RANS, LES instead of DES, DNS instead of LES; you get more detailed information. The question is whether you need that level of detail in the solution or can you go with a simulation that costs you less in terms of run time.


All times are GMT -4. The time now is 19:07.