[MikeBravo]

Has anyone developed a way of introducing synthetic turbulence at the inlet of an LES domain? I've seen very little about this for CFX and haven't found a way. Any advice would be appreciated!

[ghorrocks]

I presume you are talking about coherent turbulent eddy structures, not sub grid turbulence. You can add sub grid turbulence by simply specifying it.

CFX has no built in mechanism to do it. There is a lot of literature on this topic so I recommend you do a literature review on it, so hopefully that will assist you in developing your own model.

Alternately - you could use the DES, SAS or related advanced turbulence models and the transition to LES occurs within the domain. This means you can specify simple RANS inlet boundary conditions and the solver will handle the coherent turbulent structures generated when it switches to the LES sub-model.

[MikeBravo]

Yes, I want to have coherent eddies at the inlet. The documentation states that CFX uses the Vortex Method, which would be fine if I could figure out how to do it, but I've found nothing on how that works... Just curious if anyone could lend insight. I'll be comparing predictions with STAR, which can use the Vortex Method or SEM, so was hoping to set up something comparable.

I've looked into using DES with a SAS SST domain at the inlet, but I'm skeptical that the bulk turbulence would get washed out instead of forming coherent structures through the LES domain. I also can't figure how the Zonal LES Model beta feature for DES works or how to set it up.

My model would actually lend itself toward a periodic BC at the outlet matched to the inlet for this, but that feels more like a trick than a modelling the physics. I'd be more open to modelling a length of the inlet in LES and mapping the transient velocities onto the inlet BC if there was a good way to do that.

Appreciate the help.

[ghorrocks]

I would ask ANSYS support on this. They may well have an example of using these features.

[MikeBravo]

Unfortunately this is for academic research, and I no longer have my ANSYS Customer Portal account. I'll try to reach out by phone, but not sure I'll get very far.

If anyone on the forum has experience using the Vortex Method for synthetic turbulence or DES with Zonal LES option, I'd appreciate learning how it was set up. Thanks

[ghorrocks]

Academic research is normally done under a research license. A research license does give you access to support. The academic license normally is just for teaching and that allows only minimal access to support. At least that is how it worked when I was at uni.

So investigate whether your uni has a research license. If you don't have a research license then you better look into whether your uni is infringing its license agreement and needs to get a research license.

How's that for an answer which turns a little question into a big headache?

[MikeBravo]

Actually, we do have a research license and a customer number. I created a Customer Portal account and got my answer!

Answer:
- Turning on Beta options in CFX (in the tree > Case Options > General > Enable Beta Features)
- Setting a transient run
- Choosing the SST model in the Fluid domain
- Below the SST Model, in the Advanced Turbulence Control, activate “Zonal LES model (Beta)”
- In the Option 'Forcing with Harmonic Flow Generator', defining the wanted LES zone. For example with an if expression, like this one: if(x>-0.005[m] && x<-0.004[m] && z>=0[m],1,0) . It has to be 1 for the LES zone and 0 elsewhere
- In the Solver Control, choosing Central Difference Scheme (for Advection Scheme), and selecting Bounded CDS if wanted (
see advice in the CFX Help)

[rellis2012]

If i have followed this correctly, in using Beta features you can specify regions of an SST model to be treated as LES? Is this also the case if the turbulence model of the fluid domain was initially set to the DES option?

My simulation depends heavily on the upstream turbulence as i am trying to look at wake recovery behind turbines and so need to capture the ambient turbulence, however with my current setup it appears that the inlet region is treated as RANS and so any turbulence specified at the inlet seems to dissipate, ultimately causing an under prediction in wake recovery.

If there are any other suggestions as to how to generate this turbulence at the inlet and get it to carry downstream to the turbine then i am all ears.

Any help given is much appreciated!

[ghorrocks]

My first suggestion is to try SAS or DES models as they can generate turbulent structures from simple inlet conditions. But as this thread states there is no built in way of generating coherent, resolved turbulent structures at an inlet for a LES model in CFX. You could develop one (which is cutting-edge research, recommended for experts only). Alternately contact ANSYS support to see if they have any methods here.

[rellis2012]

Unfortunately DES and SAS do not seem to give the desired turbulence, there is not enough instability upstream for the transition to occur. I do not think i have the time (and probably also the skill) to develop one at this stage.

ANSYS stated that a velocity profile at the inlet would do the trick but it appears this has also not worked due to the domain depth being quite small as it is replicating experimental work. I am in the process of trying a model with a free surface as i think this should provide the necessary instability but i will see how it goes.

If i were to take a plane from the outlet of a 'converged' (i use converged loosely here for turbulent structures) model and use it as the input for the new DES model, would that help?

Again, thanks for the help. T

[ghorrocks]

Yes that would help. But the turbulent structures need to be coherent in time and space, so you have to define a boundary which changes in time and space. Is this what you had in mind?

[MikeBravo]

Originally Posted by rellis2012

Unfortunately DES and SAS do not seem to give the desired turbulence, there is not enough instability upstream for the transition to occur. I do not think i have the time (and probably also the skill) to develop one at this stage.

What do you mean that DES and SAS are not giving the desired turbulence? I have had the issue before where DES and even the new SBES do not switch to LES early enough to model the turbulent flows of interest. But RANS transports all of the k and Omega (with the exception of residuals lost) so you're not going to lose anything in RANS, it just won't be resolved. LES on the other hand, will wash out turbulence if your mesh isn't fine enough.

From my understanding of your problem, it sounds like a good application for the zonal LES approach, where an SST inlet domain can be established with Turbulence Intensity and Length Scale specified. At the RANS-LES interface, CFX converts the k-Omega to synthetic coherent turbulence in the LES domain. This way you can specify as high of turbulence level as you want with just 2 input variables, you can even make it a function of time. You can also make a RANS domain at the outlet which will convert the coherent turbulence back into k and Omega, and thus give you a nice value for the downstream turbulence.

Let me know if you have any questions about implementing zonal LES in CFX. Regards.

[rellis2012]

Originally Posted by ghorrocks

Yes that would help. But the turbulent structures need to be coherent in time and space, so you have to define a boundary which changes in time and space. Is this what you had in mind?

That was my initial thought, although given the opportunity I would ideally not have to use this method.

Originally Posted by MikeBravo

I have had the issue before where DES and even the new SBES do not switch to LES early enough to model the turbulent flows of interest.

I think you could be right with this. It seems as though the transition behind the turbine with DES sometime doesn't occur early enough to catch the complex nature of the flow directly behind the turbine as you have found, ultimately leading to slower velocity recovery further downstream.

Originally Posted by MikeBravo

From my understanding of your problem, it sounds like a good application for the zonal LES approach, where an SST inlet domain can be established with Turbulence Intensity and Length Scale specified.

I suppose in using the Zonal LES this would still be similar to the DES model except the LES would be controlled by the zone specified, rather than the grid sizing in the regions of interest.

I played around with the Zonal LES on the weekend for a more 'basic' model and didn't have too many issues in the setup at least so i will see how it goes on my other model. Your description of setting it up earlier in this thread was very useful, would have had a hard time finding that otherwise.

Cheers for the help both, think this should keep me occupied for the time being!

[Jiandong]

Originally Posted by MikeBravo

What do you mean that DES and SAS are not giving the desired turbulence? I have had the issue before where DES and even the new SBES do not switch to LES early enough to model the turbulent flows of interest. But RANS transports all of the k and Omega (with the exception of residuals lost) so you're not going to lose anything in RANS, it just won't be resolved. LES on the other hand, will wash out turbulence if your mesh isn't fine enough.

From my understanding of your problem, it sounds like a good application for the zonal LES approach, where an SST inlet domain can be established with Turbulence Intensity and Length Scale specified. At the RANS-LES interface, CFX converts the k-Omega to synthetic coherent turbulence in the LES domain. This way you can specify as high of turbulence level as you want with just 2 input variables, you can even make it a function of time. You can also make a RANS domain at the outlet which will convert the coherent turbulence back into k and Omega, and thus give you a nice value for the downstream turbulence.

Let me know if you have any questions about implementing zonal LES in CFX. Regards.

Hi, Mark,
I am currently using the zonal LES model. Few Questions about this model:
1' Any mesh requirement for the LES zone? (like Y-plus)
2' Any other setting needed about the model? expect for those settings:3' How can i know i have applied the zonal LES model successfully?

It been long time since you propose this question.
I will be more than thankful if I can receive a reply.