[pandora]

Hi there

Here goes a question that has been puzzling me for some time:

The best way to put my question is thinking on a turbulent flow around a circular cylinder placed vertically in a flume. For some Reynold numbers the flow field is unsteady, so a steady-state numerical model will not simulate the transient eddies behind the cylinder. Your solver would give you a nice (wrong) solution though. I assume that the only way to get around this is by running a time dependent calculation. Now for this particular problem the the reyonlds number for which we get this transient vortex sheding are well known and it doesn't pose any problem.

The question is, is there a way to predict this behavior for a more complex problem? You will always be able to get a steady-state solution but it would be unsatisfactory if it's masking these kind of effects. I guess this may be some sort of well known problem but as a newbie I would appreciate any advice you guys could give me.

Thanks in advance

[ghorrocks]

There are no general rules that I am aware of. The Re number regimes key changes take place for a cylinder are as good a guess as any for other shape.

There has also been a lot of work on bluff bodies, such as rectangles, squares and the ahmed body. You might be able to get equivalent transitions from these shapes.

[pandora]

Hi ghorrocks

Thank you so much. I understand that there's a lot of work on different bodies and that for all of them the Re numbers can help. My concern is that the geometry and problem type I'm dealing with are extremely different to that of the flow around an object placed in a flume.

To start with is a natural convection problem with buoyant flow, where unstable regions are created by the presence of a lighter fluid below a heavy one. For some conditions this can give rise to unsteady flow with coherent structures. Something like a lava lamp. The question is how to predict these kind of behaviors on beforehand when their existence is not so straightforward. Is this possible? How do people deal with these problems?

Already thank you!

[joey2007]

Roughly spoken: engineers believe every flow is fully turbulent. Subsequently they do not deal with that problem.

As far as I remember some papers report that you get good results on such flows with laminar approach. You have to resolve time and length scales of the instability. If you miss to resolve, you will get nicely converged results which are wrong. If you then refine, you may get the instability, however it will change your results completely. You can never be sure about such results. I ask myself if you choose your name because of that? ( http://en.wikipedia.org/wiki/Pandora%27s_box )

[pandora]

;-) Not quite, but now that you say it my avatar matches my problem!

Thanks for the reply! I would appreciate if you could give those references you're talking about. Anyway, I'm not sure to understand the answer and how assuming a fully turbulent flow would make the problem easier. Actually, when I think on the shedding of coherent vortex from the typical alternating Karman vortex street, it seems to me like the problem itself has a turbulent nature.

I would appreciate if you could develop a bit more your answer.

Thanks
Antonio

[ghorrocks]

In my experience natural convection flows with gross flow motion are hard to model with turbulence models. You may well have to directly model the flow, such as LES. But give it a go, you don't want to do LES unless you really have to.

[pandora]

Ghorrocks, what do you exactly mean by hard. Is it that is difficult to reach convergence or that your converged solution may be wrong? Anyhow, which is the physical explanation for this?

Btw, is there some kind of sign that should lead me to opt for a LES or is it more of an art?

[ghorrocks]

By hard I mean difficult to get accurate results. You can simply turn on a turbulence model and you will get a result, but it will probably be totally wrong. Validation and verification of a turbulence model to give accurate natural convection results is hard.


For a natural convection flow, if a transient run was showing flow features of a significant size (they often like to flap wildly about) then I would starting thinking about either a transient laminar (if low Re number) or LES simulation (if high Re number).

[pandora]

I may be misunderstanding something here but wouldn't these large flow features flappig wildly (that in fact I have) be a sign of turbulent regime? In this case my only option would be using DES or LES right?

[ghorrocks]

The flapping is not necessarily turbulent. It could be laminar flow structures. If they are laminar structures then a turbulence model is inappropriate. If they are turbulent then it is still unlikely that a traditional turbulence model will capture the physics properly.

So if you have a flow which is flapping wildly then you should do a transient laminar simulation (if laminar) or a LES/DES simulation (if turbulent).

[pandora]

Guys, let me ask one more thing. Do you know some refeences books/papers.. where to gain some insight on this topic? I would really appreciate it!

[ghorrocks]

Look for stuff under "Bluff body flow", examples being the Ahmed body, and flow over various cylinders (square, round, rectangle etc). Should be lots of studies of this stuff in the literature.