CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   CFX (http://www.cfd-online.com/Forums/cfx/)
-   -   Vortex shedding, FSI-analysis, turbulence numerics (http://www.cfd-online.com/Forums/cfx/75635-vortex-shedding-fsi-analysis-turbulence-numerics.html)

tallknuseren May 1, 2010 07:39

Vortex shedding, FSI-analysis, turbulence numerics
 
Hello,

I'm trying to simulate 2d-vortex shedding behind an asymmetric blade profile, where a part of the blade from the trailing edge and upstream is flexible and free to oscillate in the transversal direction. (It's actually two bodies, one approx. rigid and one flexible, linked together.) Inlet specified by normal velocity and medium turbulence intensity, pressure outlet.

My mesh counts approx. 100k elements, using mostly tri.elements and hex.elements in the boundary layer. Pretty small cells in the near wake downstrem the trailing edge.

My numerics settings: SST turbulence model, high resolution adv. scheme, second order euler. Tried both first order turbulence numerics and high resolution. Timestep is set to approx. 70 steps per shedding cycle. 1-3 coeff. loops and 3-5 stagger iterations per timestep, as recommended in the CFX-manual. RMS residual target = 1e-4 by default.

My residual plots confuse me, especially the Turb. frequency. It's pretty spiky, ranging between 2 orders of magnitude within a fem timesteps. Other things, like momentum and mass converge nicely to approx 1e-5. I've concidered my max. residuals, and they are typically located to the sharp corners at my geometry, which I can understand. About my monitor points - I monitor pressure and transversal velocity at approx. 10 spots in the wake. These plots seems pretty random to me..

When wathcing my results in CFD-post, I'm not able to see the vortex street in the wake. In an early model, where the fluid domain ranged from the start of the flexible blade, a little coarser mesh and no FSI were included, the vortex street came easily. Though, I'm not quite sure about the numerics settings used in this early attempt...:confused:

The images in this folder describes the most about the case:
http://folk.ntnu.no/andrebe/
(Mesh has been further refined after these shots were taken.)

And yes - I know that vortex shedding is widely discussed in this forum, because I've red most of it already. Same about the CFD-manual, regarding the numerics settings etc.. I'm really struggeling with this case, and the long time spent running simulations with (wrong) settings is not very effective.

Best regards,
Andreas Berg
Norway
--
Master graduate student,
this summer entering a job at a Marine thruster manufacturing company.

ghorrocks May 2, 2010 07:15

Nothing leaps out as wrong, but you are trying to do a fair difficult simulation. I would recommend checking the timestep size and mesh size sensitivities.

stumpy May 5, 2010 10:19

It sounds like adding FSI into the mix caused the problems. There are many things that could be wrong in the FSI setup/convergence. It would really need some debugging from an FSI expert, so I would recommend asking your Ansys support contact to take a look at the case.

tallknuseren May 10, 2010 04:31

Quote:

Originally Posted by stumpy (Post 257670)
It sounds like adding FSI into the mix caused the problems. There are many things that could be wrong in the FSI setup/convergence. It would really need some debugging from an FSI expert, so I would recommend asking your Ansys support contact to take a look at the case.

I'm afraid you are pretty right about that. In this moment, I am now running the same analysis, but I have disabled mesh motion and external solver coupling, running a completely "stiff" analysis. Yet the simulation has just began, my convergence plots and monitor points are behaving quite nicely.

Best regards,
Andreas Berg
Norway
--
Master graduate student,
this summer entering a job at a Marine thruster manufacturing company.


All times are GMT -4. The time now is 10:14.