Hi

When I run LES problem I get variables called reynolds stresses in post. Physically what are reynolds stresses? Thank you.

Dear UndergroundMan,

How did you do LES? To reach the stage where you do LES, one will surely come across Reynolds Stress. TO summarize, it is the cross-correlation coefficient of the fluctuating components of Velocity Field. As an analogy, integral of sin(theta) over 0~360 is zero. However, integral of sin(theta)*sin(theta) over same range in not zero.

You will get some terms as u'u' or u'v' when you put u = u_m + u' and v = v_m + v' in the Navier Stoke's equation (which is valid for Laminar Flow only).

Hope this helps you!

Amod

Dear Amod

Thank you for your answer. One does not have to be einstein to run LES problem. All you have to do is choose a LES model under turbulence models list. I am sure you will agree anyone can do that.

I understand non-linear term in Navier-Stokes equations give rise to reynolds stresses. I wanted to know what are reynolds stresses physically? I understand how we obtain them mathematically.

Reynolds stresses are...stresses! In fact they are the fluctuating part of the stress tensor, related to the presence of turbulence in your fluid. As you know stress TAU = Visc * d(vel)/dt but vel has a turbulent fluctuating part, so the same part has the tensor stress...it's ok?
Try to look at http://en.wikipedia.org/wiki/Reynolds_stresses</p> Bye
Andrea

"One does not have to be einstein to run LES problem. All you have to do is choose a LES model under turbulence models list. I am sure you will agree anyone can do that."

This is the attitude that gives CFD such grief (and is perpetuated by the commercial CFD code vendors who say anyone can to CFD).

Sure, any monkey can click a button, turn on an LES model and get a solution. But if that monkey doesn't understand the physics, what value can you give to the solution? That same monkey might run an Euler solution over a flat plate and conclude there is no drag.

While you are doing the right thing by trying to understand what is going on, I think your comment is weak.

Well with a basic understanding of CFD one can manage. In terms of running the code there is not a huge difference between URANS and LES. All you have to understand is that in case of LES you have to generate turbulence in your domain and when solution is statistically stable you start collecting your statistics. And also resolution of your grid determines what scales you are going to resolve and rest of them are modelled.

Anyone can do CFD even monkeys. You give numbers to computer and it gives you results. You then validate your results with experimental data. Thats it! I learnt CFD in ten days.

So then I assume you know all about the limitations of the LES sgs model and that you need to resolve the turbulence down into the intertial range scales if you want to use it correctly. That being said, it is surprising that you understand the turbulence spectrum yet you don't understand something as basic as Reynolds stresses.

If all you want to do is match experimental data then you are ok. Even a monkey can keep tweaking their model until they get the number they want. What about a case that is outside the box for which you don't have data?

Well its says in text book that smagorinsky's model is very dissipative! Thats excatly what I am saying, all you have to do is tweak buttons like monkey and eventually you will get results. What is the big deal. How would you validate your results for which you dont have a data jean? You can never rely on CFD alone.

What does it mean to say that Smagorinsky is very dissipative? Does it dissipate all of the scales of turbulence? It should not. In LES you have turbulence scales that are fully resolved by your grid. You also have scales that are not resolved. If you didn't then you would be doing a DNS and you wouldn't need an SGS. But, in LES you assume some scales are resolved while others are not. The SGS model is designed to account for the impact of the unresolved scales on the resolved ones. To simplify the model and to make it more universal, an assumption is made that the grid resolves turbulence scales into the intertial range. If you don't resolve into this range then strictly speaking, the Smagorinsky model is not valid. That being said, many people get reasonable results even though they cut corners here and there. It depends on the specifics of the test case and what exactly you are interested in predicting.

If you are a competent practitioner, then you can use LES to do more than reproduce existing results to aid in the interpretation of an experiment. You can actually use it as a predictive tool. Ie, say you have a geometry operating at a certain condition for which you have experimental data. You can validate the model at that condition and then run the model at other conditions to predict the behavior. You can only have confidence in the results at the other conditions if you understand LES.

To provide some info on your original question, there are two components of the Reynolds stresses in an LES. One is due to the resolved scales (that you can calculate by sampling your solution and performing the appropriate correlations) and the other that is due to the unresolved scales modeled by the SGS model. I'm not sure which is given in your output, but I suspect it is that due to the resolved scales.

I hope this helps.

I understand, those scales which are too small to be captured by our grid are dissipated and this happens near the wall boundaries. In DNS our grid is fine enough to resolve all the scales. But these things are so easy, anyone can learn it.

You say:

"You can only have confidence in the results at the other conditions if you understand LES".

Well you look at the flow and see if it makes sense. If it does, then you assume your results are correct. Now what is difficult about that? You really dont have to be einstein to run LES problem. Any monkey can that do. Look at yourself even you have been managing it. Thank you for answer anyway.

You are taking a lot of risk if you think that all aspects of a solution are correct just because you spot checked a few pieces. You can generate some very nice visualizations and show something that looks very reasonable but have things like drag and turbulence levels completely wrong.

I don't know if you are serious or just pulling my leg because your comments are dangerously ill informed (and they are outdated by about 30 years).

In any event, best of luck to you. I just hope our paths never cross and I have to rely on your solutions (but then again, our paths shouldn't cross as I live in the jungle with Tarzan).

- Cheetah

Thank you for your answers Jean. I hope I would have learnt a bit more about LES by the time my solutions start to affect other people's lives.

Let's hope so.