[MovinJ]

Hi All,

I am trying to run a multiphase CFX simulation on this geometry and I am having a hard time with various problems. I am implementing just two phases, one continuous fluid phase and one dispersed solid phase. I want to model dispersed solid particles of 1 micron (1e-6m) to represent finely aerosolized droplets. I had a relatively coarse mesh of mesh size 1mm earlier and I was having a lot of trouble getting the model to converge.

After reading the theory guide and understanding that there's a discretization algorithm under the hood that's trying to interpolate the location of a particle to make calculations between the Lagrangian method (used by CFX for dispersed solid) and Euler method (for continuous phase), I came to the hypothesis that maybe the mesh needs to be more fine than 1mm.

I then tried to just change the mesh size, use regular face sizing etc to make it more fine but the meshing was taking way too long, and looked like it would take longer. The size I was aiming for was 1e-4m, 0.1mm. I then took to googling to find out why it was taking so long and how to make meshing faster. It was mentioned somewhere that if you use things like match controls, edge sizing, specifying the exact amount of nodes on an edge, sweep methods, breaking the body up to smaller parts etc. that the meshing would take significantly less longer since the software doesn't have to decide for itself.

However, I have never really had to do this for other geometry before (just cause the automeshing on ANSYS is just so good on its own), so I have no experience with the various tools/methods that ANSYS has built in to make a more specifically-defined mesh.

I have attached two pictures of the geometry and given some info about the various boundaries. I was wondering if I could get some advice on what the best way to do this would be, and maybe if people online could impart their wise wisdom on me as a beginner who's only just starting to get into more complicated meshing.

PS. The tubes that are going into the domain are half inch tubes (12.7mm), and the overall diameter of main body is 100mm, if it helps give the scale of the problem.

Thank you all

M

[Gert-Jan]

- If you go from 1mm element size to 0.1mm, your number of elements will roughly increase by 10x10x10=1000. I can imagine that you have to wait a long time. Possibly your computer might start to complain as well?
- How do you judge convergence? There is more than just residuals. Did you read this: https://www.cfd-online.com/Wiki/Ansys_FAQ
- Refine the mesh mostly does not help in reducing residuals. The finer the mesh, the more details in the flow, the more likely you'll hit a transient phenomena, which will increase the residuals.
- Create a hex mesh. That mostly helps a lot, but is quite a hassle.

[MovinJ]

Hey,
Yes, my computer really wasn't handling it well when I decreased the mesh size by a factor of 10, and it's really obvious now, since it's actually 1000 times more work for a factor of 10 increase in quality. I settled for a decrease to 0.75mm, which hopefully may be enough.
Earlier, what I was trying to do was run a transient simulation of the model. It would do about 8-12 convergence loops before jumping to the next timestep. When I said

Originally Posted by MovinJ

I had a relatively coarse mesh of mesh size 1mm earlier and I was having a lot of trouble getting the model to converge.

What was happening was the run was failing at the end of a time step if the target of residuals weren't met within the convergence loops.
I converted my model to a steady state problem and I had a go at it, and I realized that attempting to converge in 8 to 12 loops was ambitious of me.
When you say "How do you judge convergence", I had a look at the FAQ and I think it was a bit unrelated to the transient problem I was attempting, but was very related to the steady state attempt I have just done; I am seeing that the solver wasn't able to converge the solution, and the residuals reached a constant value much higher than my target. This is definitely to do with the the shedding occurring now due to the fine mesh I believe.
Ultimately for the meshing, I used a tetrahedral method for the one I just ran, and turned off as many auto meshing features as possible, and it went very well, much better than expected. I will try the hex mesh as well if I can.
Thanks for directing me to the FAQ, that was a very helpful link, and thank you for your support!

[pavdarkel]

Hi, Mr.MovinJ
What time step are you using in the transient calculation?