[alquimista]

Hello everybody.

As referenced in the announcement of the new release version of OpenFOAM a new turbulence model has been implemented:

Version 2.1.0 includes the k-kl-omega (low-Re) model for Reynolds-average simulation of incompressible flows with a boundary layer that undergoes a transition from laminar to turbulent flow. The model is described in D. K. Walters and D. Cokljat, J. Fluids Eng. 130:121401 (2008) and is based on the model k-w with an additional transport equation included to predict low-frequency velocity fluctuations that trigger transition in the boundary layer.

I have been looking for some additional information about it in the user guide but it doesn't appear. I just want to highlight that issue to discuss it and ask if someone has test it. As soon as I check the model I'll report it.

Since I'm not an expert.... Is it an "automatic near wall teatment" (like the near wall treatment implemented in ANSYS CFX)?...then It resolves the viscous layer near the wall (mesh refinement is needed)?

Thanks

[RuiVO]

Boas Alquimista (Or hi Alquimist )

I'm also going to begin using the transition model, I was also expecting to find some clarification about this in a tutorial or in any reference besides the one given in the openFoam site which is referencing an article of ASME.

The thing is, I suspect that the boundary conditions for "kt" and "w" are similar to a regular k-w model, though the boundary conditions for kl, since it represents the laminar fluctuations kinematic energy should be set to zeroGradient at the walls , uniform zero in the domain (for initial conditions), fixed value uniform zero at the inlet and zeroGradient at outlet.

My thoughts of course ..

Best regards

RuiVO

[klausb]

Hello Gentlemen,

let's join forces and setup a reference case using the kklomega model.

How about a 2D case for a wing?

I'd prefer a case based on a stl file so it can be adapted for different geometries.

Interested?

Klaus

[alquimista]

Boas RuiVo (or Hola jejeje)

It's nice to see someone else interested in it. By now I needn't (or I can't) use this model since I realised that my case is fully turbulent but I'm interested on the results of the model.

In my first topic I talked about the near wall treatment but I was wrong mixing the turbulence model itself with the boundary conditions specified for the wall treatment.

I think it's a good idea suggested by klausb. I propose to use the object shown in the anouncement



It would be nice to use a geometry present in some tutorial. I'm trying to find something appropiate.

[lakeat]

Anyone knows how to set the boudndary condition for kl and kt?

For now I just copy k file and then rename to kl and kt.

I am running a square case, and so far , the strouhal number, the mean and RMS value of force coeff look quite good!

[klausb]

Hello,

would you mind sharing your case (post your case directory).

I'd like to play with it over the weekend.

Klaus

[Simon Lapointe]

Originally Posted by lakeat

Anyone knows how to set the boudndary condition for kl and kt?

According to the article presenting the model, you should use kl=kt=0 and omega =zeroGradient at walls. At the inlet, k=0 and kt and omega are similar to a k-w model so based on desired inlet turbulence intensity and viscosity ratio.

[lakeat]

Originally Posted by Simon Lapointe

According to the article presenting the model, you should use kl=kt=0 and omega =zeroGradient at walls. At the inlet, k=0 and kt and omega are similar to a k-w model so based on desired inlet turbulence intensity and viscosity ratio.

Thanks, that's exactly what I have set right now.

But just for a future discussion.

I have done a lot of tests, and generally
I found kklOmega works well when it can work. But I found it do have convergence problems, especiall kl will grow very large within vortices.

Relatively, I found other LRN models are more stable.

[egp]

I would vote for two test cases: 1) a classic zero-pressure gradient flat-plate boundary layer (Figs. 2-6 of Walters and Cokljat) and 2) S809 wind-turbine airfoil (Fig. 15-16 of Walters and Cokljat).

I have meshes for both, and will try to reproduce the results in this paper. If I am successful, I post case files. If not, I'll post questions and case file and we can work on this as a group.

[lakeat]

And also a reminder, that I found the following utility is needed for using these kind of LRN turb model.

[sandy]

Yes, after my code skipped the time =0, it became convergent very well.

In addition, lakeat, you mean all LRN models can also solve transition flows (from laminar to turbulence), right?

[lakeat]

1. Turbulence;
2. (Re-)Attached or Separated Turbulence;

And IMO, I think the three-equation model still needs extensive tests, and it needs to be bounded.

(Chinese) 新年快乐！

[sandy]

Hi 老魏, 春节快乐！

I want to find the model to simulation the transition flow, I guess kklomega model is good for it. However, you think, which scheme should be chose to bound the computation values? I find, in OF, almost all schemes are bounded schemes. What about your thought?

Sandy

[lakeat]

Originally Posted by lakeat

And IMO, I think the three-equation model still needs extensive tests, and it needs to be bounded.

Correction: I dont know if it really needs to be limited. I am struggling testing with some other convection schemes.

[sandy]

Yes, lakeat, except Guass upwind, I will change my code schemes into bounded schemes ...

[egp]

I've set up the Schubauer and Klebanoff case for natural transition (as opposed to the ERCOFTAC bypass transition cases). I'm getting closer on getting kkLOmega to work, but the results still do not look good. Here is the Cf vs. Rex plot:



You can download my case from here:

http://dl.dropbox.com/u/2182201/kklOmega.tgz

Note, freestream parameters are set as follows:
U = 50.1 m/s
Tu = 0.3
kt = 0.03388 m2/s2
kl = 0 m2/s2
nut/nu = 1
nu = 1.5e-5 m2/s
omega = 2259 1/s

wall boundary conditions are set to:
kt = 0 m2/s2
kl = 0 m2/s2
omega = zeroGradient

I'd be interested to see how others would try to set this case up.

Originally Posted by egp

I would vote for two test cases: 1) a classic zero-pressure gradient flat-plate boundary layer (Figs. 2-6 of Walters and Cokljat) and 2) S809 wind-turbine airfoil (Fig. 15-16 of Walters and Cokljat).

I have meshes for both, and will try to reproduce the results in this paper. If I am successful, I post case files. If not, I'll post questions and case file and we can work on this as a group.

[egp]

OK, found a mistake in the 0/nut. wall b.c. for nut was incorrectly set from a fully turbulent case to nutWallFunction, and is now set to zeroGradient. This improves the solution, but laminar Cf and transition location are still off quite a bit.

Originally Posted by egp

I've set up the Schubauer and Klebanoff case for natural transition (as opposed to the ERCOFTAC bypass transition cases). I'm getting closer on getting kkLOmega to work, but the results still do not look good. Here is the Cf vs. Rex plot:

You can download my case from here:

http://dl.dropbox.com/u/2182201/kklOmega.tgz

Note, freestream parameters are set as follows:
U = 50.1 m/s
Tu = 0.3
kt = 0.03388 m2/s2
kl = 0 m2/s2
nut/nu = 1
nu = 1.5e-5 m2/s
omega = 2259 1/s

wall boundary conditions are set to:
kt = 0 m2/s2
kl = 0 m2/s2
omega = zeroGradient

I'd be interested to see how others would try to set this case up.

[sandy]

Hi egp, I guess, you gave wrong the kt value. How did you calculate the kt and omega values?

[egp]

Originally Posted by sandy

Hi egp, I guess, you gave wrong the kt value. How did you calculate the kt and omega values?

Why would you guess this? The Schubauer and Klebanoff (S&K) case is a well-known experiment with very low free-stream turbulence, i.e.,

U=50.1 m/s
Tu = 0.3%
nut/nu = 1

Given these parameters, I compute kt and omega to be:

kt = 1.5*(0.003*50.1)^2 = 0.0339
omega = k/nut = 0.0339/1.5e-5 = 2259

It's interesting to note that the paper by Walters and Cokljat (2008) did not use the S&K dataset for validation and only used the ERCOFTAC T3A-, T3A, T3B benchmarks. These latter cases have a varying degree of free-stream turbulence and are typically used to demonstrate bypass transition. In contrast, the S&K with low FST is more a test of natural transition.

[vinz]

Dear Eric,

Very interesting test case. did you try different mesh refinement to see if it has a large impact on the solution? spanwise and wall refinement?