Automatic Meshing Based on Turbulence

[knewsomeN8P59]

Hey all,

I'll preface the question with some info about my simulation:

1. It is about 19 million elements before adaptation. This is after some manual refinement and turning surface-gap refinement on.
2. There is heavy heat transfer, around 300kW, and ambient cooling as well, about -20kW. For this application, this only results in ~20-30C temp rise, no phase change.
3. The fluid is oil, the only solid is copper (300kW generation), the flow is slow (~4m/s max), and I have simplified the geometry to remove small parts and small gaps to help convergence.
4. The ideal end goal is k-epsilon turbulence, Intelligent Wall Formulation (IWF) turned on, Advection 5. In practice this gives the most realistic results compared to my measurements.
5. I realized recently that the issue with my convergence that I've been struggling on-off with for a while is IWF. Previously I have ramped up my simulations from Laminar Advection 1 to k-e Advection 5, but I'm unsure if this is necessary once I turn IWF off, i.e., I think I can disregard the ramp-up and "cold start" my simulations k-e Advection 5 IWF off, then once that converges, maybe turn IWF on. I have not tested this yet. I believe the issue with IWF turning on is my mesh, although it could be other things (please let me know if you have other ideas)

My question is this: Does automatic meshing depend on the turbulence model selected when meshed? If so, I think my IWF issues could be caused by my mesh being generated on a laminar advection 1 solve, instead of k-e advection 5. I really need IWF turned on for the final run, so if you guys think of any other ways I could achieve that, please let me know. The only other thought I have is running an adaption cycle first so the mesh is more intelligent than the base mesh I'm using.

Thanks in advance!

[knewsomeN8P59]

Note: I looked through my mesh and found a very small gap (not in Inventor, not sure why porting it to CFD introduced it) that skyrocketed the nodal aspect ratios to around 30,000 there. I assume this is where the issue lies, but I'm going to first try running the sim from the converged k-e advection 5 IWF off solution, with IWF on and velocity, pressure, and temperature at half the relaxation values in hopes to sidestep fixing the geometry, although I know that would help a lot.

[chris_smith1]

Howdy there- so if you're talking about the voids in plane view that's more of a visualization thing. The results don't show there are voids in the fluid domain so you should be good to ignore those.

For the wall formation question, i'm a bit confused on why you're looking into this. With such low flows I'm stumped on why you wouldn't be getting convergence. Could you share your conv. plot? If you want to create a support Case- I'd be glad to go over your files in a more secure channel. 

My understanding to your primary quesiton: My question is this: Does automatic meshing depend on the turbulence model selected when meshed?
My gut would tell me that it shouldn't given that you can take a mesh and run it with various models, I'll double check with development, but I'm pretty sure that it should be independent. 

[lauri_barnhart]

Welcome to the community, @knewsomeN8P59.

Hope you are well. Following up to see if the information provided by chris_smith1 helped clarify your question. If it did, please consider clicking on the 'Accept Solution' button.

Thanks,