[pedroxramos]

Hello there!

What you think about this residuals for the flow simulation around a cylinder?

Q=0.02 m3/s h=0.10 m diameter of cylinder=48mm

transient and LES turbulence model

http://i45.tinypic.com/352je61.jpg

[RodriguezFatz]

Can you show us the grid?

[pedroxramos]

Of course, thanks.

[RodriguezFatz]

Hey,
i) Is this 2d?
ii) Did you try any other turbulence model / stationary case and got better convergence?
iii) Did you try to improve the grid? It looks like you have several "jumps" in your mesh size... although I wouldn't expect the residuals to behave like this due to these jumps...

[pedroxramos]

Hello, thanks for your reply!

Do you think that the residuals are too bad?

This is 3d, i show you the plant of the grid.

I'm obligued to use the LES-model...

Only the continuity value is bad don t you think?

[RodriguezFatz]

I wouldn't really trust a result if "just one" residual looks unusual. Can you localize the point of bad contiunity?
Could you post a picture of what happens during just a few timesteps. I can not really see something in the residual picture.
I guess you did not change fluent's way of normalizing the residuals. You just use the default way, right?

[pedroxramos]

Ya... i'm new in CFD...

There is a problem wich concerns me... the velocities are too high in the initial section next to channel's bed. Like this contour of shear stress in the bed shows



I defined the mass flow rate in the inlet with the flow (20 l/s).

Thanks for your help!

[RodriguezFatz]

I meant the residuals. Can you post a picture of the residuals where you can see just a few time steps? I want to see how each time step converges...

[pedroxramos]

I'm sorry Like this: ?

[RodriguezFatz]

Great... Is each of the peaks the beginning of a new time step?

[pedroxramos]

No, sir... I think that each of the peaks represents one iteration. I defined 20 iterations for time-step...

Do you recommend any example or tutorial for the simulation of the turbulent flow around a body (cylinder)?

Thank you, sir.

[RodriguezFatz]

I don't think you are right. It looks like you have about 10 peaks each 100 iterations, thus 10 iterations per timestep. I guess you did not switch off the convergence critertion. Go to monitors -> residual monitor and set the convergence criterion to "none". You have the default case of "1.0e-3", which means that fluent will stop iterations each time-step, when all residuals are below 1.0e-3. If you switch that off and maybe increase the maxminum number of iterations per time step -> tadaaa.

[pedroxramos]

Thank you for your great advise!

The results improved and, in the other hand, the simulation is too slow now... but i'm patient


I think it is possible improve the residuals. Do you know the meaning of "extrapolate variables?"

[RodriguezFatz]

Fluent Help says, that "extrapolate variables" uses a Taylor series expansion to estimate an initial guess for the next time step. If it is faster, I don't see any drawback...