can any body explain was is kato-launder modification.

1. It is use for turbomachinery if yes then why.?

2. What we gain in terms of solution accuracy?

3. Will this take more time than standard turbulence models?

4. what is value of y+ for this model if exits any?.

5. Is this low reynolds number model if yes then why ?

Did you check out CFD-Wiki? You can find a basic introduction of the Kato-Launder model here:

http://www.cfd-online.com/Wiki/Kato-...r_modification

ya i have already refered it. It is very good reference . and after reading i have these question. plz could u answer them and tell me which turbulence model is best for turbomachinery

Okay, here comes the answers:

1. The Kato-Launder modification is often used in turbomachinery applications. It was very common about 10-15 years ago. Lately it has been replaced by more modern variants though, I'm thinking of various non-linear models and realizability constraints that inherently avoid the problems with stagnation regions and regions with strong acceleration or decelleration.

2. You avoid the excessive overproduction of turbulent energy that classical k-epsilon models tend to produce in stagnation regions, shock regions and regions with very large normal strain.

3. The Kato-Launder modification usually makes the simulation slightly more stable since it avoids the excessive overproduction of k in problematic regions.

4. Which y+ you need depends on what model that you use as basis before you add the Kato-Launder modification.

5. See question 4. It all depends on which model you use as a basis. The Kato-Launder modification can be used both with a high-Re wall-function simulation and with a low-Re simulation.

which model is best for cfd of turbomachinry. My advisor always prefer the SA model...??????

The SA model has become very popular in the last few years. The main reason is that it is a fairly simple and robust model which rarely leads to completely unhysical results. People are tired of getting completely unphysical results with the k-epsilon models and instead turn to the more simple SA model. However, the SA model is a one-equation model which is not that good at predicting convection and diffusion of turbulent properties. The SA model is often used for cases with fairly low free-stream turbulence models, like fans, wings etc. A validated and well tuned two-equation model should be better than the SA model. However, the SA model is always good to start with since it gives you a fairly quick and fairly reliable result before you decide if you need to also test a more advanced model.

I don't think that you can say which turbulence model is best for turbomachinery applications. There are so many different applications - separation prediction, heat-transfer, performance prediction, secondary flow predictions, ... In the last few years Menter's SST k-w model has become more and more popular, as has the more simple SA model as you suggested. In terms of turbulence modeling for turbomachinery I'd say that one clear trend in the last years has been to start to do transient simulations, perhaps running a DES model, instead. The original DES model is based ona a SA variant, but there are also several other hybrid models that use two-equation modesl like SST k-omega as a basis.

right. But SST model requires y0 = 1 which means atleast 1 milion modes for one components in other words for one stage of compressor we need atleast 2 millions nodes to apply SST.

now question is that if we need to only find out performance not the secondary or seprated flow, then is it advisible to use y + = 1 or sst model??

waiting for ur preciuos reply

regards sam

First, both the SST k-w model and the SA model are low-Re models and as such they both require y+ to be around 1 for the first cell. However, these models can also be used together with a wall-function approach at the walls in order to save grid points. Hence, you can't select between these models based on their low-Re capabilities only.

If you are predicting on-design performance for a fan or compressor using a code and models that you have validated against similar applications you can most likely get fairly good results with a high-Re model using wall functions. With todays computers it is not a problem to run a steady state 2-million cell simulation though. What is really important is that you have a validated code and model and use it approriately, otherwise both a low-Re and a high-Re simulation might produce bad results. My experience is that when you have fairly complex separations and shapes, with wall-curvature etc., you most often get better results with a low-Re simulation, whereas with a simple on-design simulation of a compressor cascade with no complex shock interactions etc. you might also get good performance from a high-Re simulation.

I am using Fluet and I have used K-E RNG and SA. They both produce same results.

Unfortunely i have no test data. so switching to different models is time vast?

thanks jonas. please tell me more abt kato launder

Nope, it is not a vaste of time to switch models. That you get similar results with different models is one indication that your choice of model is not that important. You can of course get good results with an unvalidated code, but then you of course need to have done all the details right (right mesh density, right boundary conditions, right convergence criteria, right discretization schemes, ...). You also need some form of information that the models and your code you is suitable for the cases you predict.

About Fluent models. I'd only use the RNG version for strongly rotating flows (cyclons etc.). For most turbomachinery application I would prefer either the simpler SA model or the more advanced Realizable k-epsilon model or the SST k-omega model (only well implemented if Fluent versions 6.2 and later though)

Kato launder is only used with standard k-epslon? or with SST as well? Where can I find out results of different turbulence models and thier discussion as applied to typical turbomachinery.

One request with jonas: Can I get geometry of nasa rotor 67 and 35 along with their validation data(i.e test data)

Yep, you can use the Kato-Launder modification also with the SST k-omega model. However, this isn't very common since the SST k-omega model doesn't have as big problems as the standard k-epsilon model has in flows with large normal strain. For tricky cases with strong stagnations etc. also the SST k-omega can produce too much turbulent energy though, and then it might be good to use the Kato-Launder modification. I'd always recommend using another type of limiter though (some realizability contraint from Shih's works or perhaps Durbin's limiter). I think that the Kato-Launder modifcation is too crude and basically turns off the turbulence model outside of the boundary layers.

About good references. There aren't that many complete references. A good classical turbulence modeling book is Wilcox's (see http://www.cfd-online.com/Books/show_book.php?book_id=1 ). Otherwise your best bet to stay on top is to read papers in journals like ASME Journal of Turbomachinery, AIAA Journal etc.

About the NASA rotors. There are versions of this data avilable I think. You have to ask the ones who owns the data if you can get a copy. It is a great way to start. Perhaps someone else with more information about this can help you here also.

Jonas,

since I have you on the line, a quick question...

in "two-equation turbulence modles for turbine blade heat transfer simulations" you state

f_mu = 1 - exp ( -3.4/(1 + Rt/50)^2)

this gives f_mu = (approx) 1.0 at Rt = 0.0 and f_mu = 0.0 at Rt = infinity...

in "an experimental and numerical CFD study of turbulence in tundish container" P. Gardin, M. Brunet, J.F. Domgin, K.Pericleous use...

f_mu = exp ( -3.4/(1 + Rt/50)^2) (reversing f_mu values)

amongst CFD papers there is probably a 60%/40% split!

It seems logical to me that f_mu tends to zero as Rt tends to zero (i.e. you are dampening your dissipation term in the dissipation equation as you approach the wall).

So who is right? I am guessing you have repeated a typo probably from the original Launder and Sharma paper?

Iain