[haconk]

Could some of you please let me know what are sensible parameters to begin with, based on your experience?

I'm doing a transient 2D-simulation of a thin plate, 1000 x 20 mm in a 5000 x 3000mm domain, 100 mm deep (1 layer). The plate is located in the first half of the domain.

What sort of mesh size, edge length, and timestep should I choose. The motion amplitude is 25mm in heave, 5° in pitch (separate) and the circular motion frequency is between 1.2 and 6 Hz.

I know I'm supposed to look for grid-independent and grid-convergent solutions by quadrupling the number of mesh elements in each run but what do you think, what numbers should I start with and what number will I end up with for an acceptable solution?

Many thanks

[michael_owen]

I might start with a cell size of 20mm at the plate. This is essentially the same as neglecting the width of the plate, given that it is not resolved. This should at least allow you to set up the run, make sure the dynamics are working correctly, get a feel for the fluid response, test your timestepping, etc.

As far as timestepping, there will be several timescales involved. The fluid residence time, the period of the oscillations of the plate, vortex shedding periods, etc. Choose a fraction of the shortest of these periods to start.

[haconk]

Thanks Michael,

I started with some steady runs but the funny thing is that the coarse mesh yields better results than the fine mesh, or put differently, the coarse mesh converges within few iterations whereas the fine mesh stops at the 100th iteration.

These are the number of elements I used:
1st run: 2125
2nd run: 3094
3rd run: 8258
4th run: 53926

Any views on this much appreciated.

[ghorrocks]

Hi,

With finer meshes you may be resolving finer detailed flow features making convergence more difficult. But you cannot say then that the coarse mesh is more accurate!

Have a look at http://journaltool.asme.org/Template...umAccuracy.pdf , or even better read Roache's book "Computational Fluid Dynamics" for a more complete coverage of accuracy issues in CFD.

Glenn Horrocks

[haconk]

Thanks Glenn,

Just ordered two Roache books:
1. Verification and validation in computational science and engineering
Patrick J. Roache. - Albuquerque, NM : Hermosa, 1998
2. Computational fluid dynamics
Patrick J. Roache. - rev. engl. ed. - Albuquerque, N.M. : Hermosa, 1985

But do you think 2000 elements are reasonable ? Or should it be more 50,000?

[ghorrocks]

Hi,

After you have read the texts you will be able to tell us.

In CFD the finest mesh you can run is never quite enough to resolve some sort of feature you would like to look at. In other words I bet you need 50k elements but I bet it still does not capture everything. But whether it captures enough to demonstrate what you are looking for - you will have to judge that.

And a further point is that CFD is meant to converge to the exact solution as the mesh gets finer. So the finest mesh you can run is your best bet for a mesh independant solution straight off.

Glenn Horrocks

[Spinassien]

I am running a 2D-Simulation of a chinese Windmill and i want to check the torque. I chose a Velocity specified condition at the inlet, static pressure at the outlet and Openings on the left and right with the Entrainment option. How can i be shure that the Boundary conditions do not have influence on the torque? My strategy is to make the Domain larger and check the Influence. If the influence is below 1% it would be ok in my opinion. Are there better ways to check the influence of Boundary conditions? How does a Professional deal with this issues?
Thank you for your advice.

[flotus1]

Your approach sounds quite professional to me.