Learning Forum

[KoShine]

Hello,

 

Has anyone even done a DNS simulation in CFX? I'm carrying out DNS simulation of low pressure turbine cascade (T106A) in CFX. I use Laminar for turbulence option. Would that be considered as a DNS simulation in CFX? I'm also wondering if it is possible to use Central scheme for discretization? 

 

Thank you in advance.

 

Best regards,

Shine

 

[DrAmine]

We do not claim DNS capabilities for our CFD codes, and these codes have not been tested for DNS. Actually, we can try to approach DNS with 2 different methods but the DNS itself is not possible, at least those 2 approaches would be questionable for DNS:

Before I mention the approaches: We are surely missing some of the specialized high order numeric that an actual DNS code would possess. Moreover DNS is very expensive (Re^3). Apporaches:

1. Using the Laminar approach: 

2. You can do that by applying the LES model and set the Smagorinsky constant to 0. 

 

Central Differencing Scheme is not enough for DNS!

 

[KoShine]

Hi Amine,

Thanks for your prompt reply.

I've tried the Laminar approach. I've also done a LES simulation using the LES Smagorinsky model with the default values. Please see the screen shots from the two simulations. Please note that this is just for comparison as the work is still in progress and they might necessarily not be correct. I used High Resolution for the Laminar approach and Central Differencing for the LES simulation. I'm going to try your approach 2 using the LES model and setting the Smagorinsky constant to zero. Would you recommend High Resolution scheme for both approaches?

Thanks,

Shine

[DrAmine]

I recommend not using our solver to make a DNS. Better to user a bounded central scheme to avoid any issues

[KoShine]

Hello Amine,

Thanks for the suggestions. For now, I'll just go ahead with those two approaches and see how it goes. Bounded central scheme is not available in CFX for the Laminar approach, so the approach 2 might be a better option to use a bounded central scheme. 

Many thanks,

Shine

[DrAmine]

It is available: Central Difference and switch on CDS bound.

[DrAmine]

And bear ib mind DNS= RE^3 you need to resolve everything in 3D.

[DrAmine]

Better to rely on SBES (Hybrid LES).

[KoShine]

Hi Amine,

Thank you very much once again for the suggestions. Will add SBES to the options. Well noted on DNS=RE^3 as well. With the laminar approach, only three schemes are available in CFX-Pre solver control - Upwind, High Resolution or Specified Blend Factor. Please see the image below. But I can use Central Difference with the LES model. 

Thanks,

Shine

[DrAmine]

Yes that was what I was talking about.

[KoShine]

Thanks Amine! Appreciated your help.

Many thanks,

Shine

[DrAmine]

Welcome!

Please mark this as "is Solved".