Time to converge with free surface (PICTURE)
 

I had posted a little ways back and got past my initial problems with my free surface flow. I believe it is now being solved appropriately, but it will not converge (as I read in the manual, its very likely free surface solutions will not converge due to little waves).

What I'm wondering is how long I should allow my simulation to solve for accurate results. I have stopped it after 400 iterations with a physical time scale of 0.01 [s] but the solution is not yet close to what I would expect from steady state.

Here is my setup: http://www.pdi4life.com/images/bowl.gif

It is essentially a 25" diameter bowl, spinning at 100 RPM with water being fed in to it through a pipe and that water climbing up the walls due to centrifugal force.

Re: Time to converge with free surface (PICTURE)
 

This is what my Momentum and Mass plot looks like after almost 500 iterations.

http://www.pdi4life.com/images/converge.gif

Re: Time to converge with free surface (PICTURE)
 

Dunno where your surface actually ends up, but it looks to me like you have a good boundary mesh for the wall interaction but the water surface will be somewhere in the middle where the mesh is very coarse?

You propably need a very fine mesh along the water surface too to get convergence.

Re: Time to converge with free surface (PICTURE)
 

Would you suggest I use a structured mesh throughout the domain?

Re: Time to converge with free surface (PICTURE)
 

Whatever suits you in order to get a mesh that has good resolution perpendicular to where you expect to find your free surface. But yes, this is usually easier with a structured mesh.

I would also suggest a run with Interface Compression = 2 to see if that changes anything.

Re: Time to converge with free surface (PICTURE)
 

This is easy to mesh with a structured mesh, so it would be a very good idea to do so. Once you have a good idea of where the free surface is, try to construct a mesh that is close to parallel to the free surface in that region, with very fine vertical spacing.

Re: Time to converge with free surface (PICTURE)
 

Hi Mark, a few additional suggestions that may help:

- It looks like you can reduce your problem size by taking a small slice instead of a 90 degree slice. I would make the mesh 1 element thick in the circumferential direction. You can create a 2D extruded mesh in CFX-Mesh to do this.

- Estimate how long it would take to reach steady-state if you actually had this set up. How much water do you expect to get in the bowl, how much do you start with and how much are you introducing? 400 iterations at 0.01[s] gives you 4[s]; if in reality it would take longer than this to reach a steady-state condition then you'll likely need more iterations, or an initial guess closer to the solution. Hope this helps, Mike

Re: Time to converge with free surface (PICTURE)
 

Mike, Charles, Chebeba,

Thank you all for your help. I was able to get the solution to converge after a great amount of iterations and the reason it was taking so long is because my initial guess was so far off (glad I did the simulation!).

Since I have to do a number of simulations like this, I will take your advice Mike and try it as 1 element thick. I will also make the mesh structured and more fine.

Again, thank you all.