[smog]

I am modeling a cylindrical jet breakup in a cylindrical enclosure with Diameter = 20*diameter of the jet. The jet inlet runs (8*d) before opening up into the larger cylinder. I wish to measure the fanning of the jet at a distance ~100*d downstream. My questions are as follows:

1) I am using LES for modeling the breakup. I succeeded with LES after stupidly using RANS unsuccessfully for a long time, completely forgetting the ensemble averaged nature of the RANS equations (thereby eliminating any eddy formations in the flow leading to the break-up). I am currently using Smagorinsky with a spatial resolution of 0.2mm (a coarser resolution would not get the jet to break-up). I also recently found out that OneEqEddy model gets the jet to break-up at a coarser resolution (0.3 mm). How do I make sure which LES approach is best suited for my case?

2) Since this is a transient simulation, I am not sure how I determine if the solution has converged. What is the convergence criterion for transient simulation? Any links to any literature would be highly appreciated.

3) My mesh runs into ~20 million cells (60 mm diameter , 300mm length). I find this very prohibitive when it comes to run times. Is there a way I can make the mesh more "economical"? I have read that LES needs a consistent mesh throughout, although I have made the mesh to be 0.2 mm at the center of the cylinder and coarser towards the outside. Can I still trust my results?

4) Finally, how can I conduct a mesh dependence study for LES, where essentially the solution is highly mesh dependent (and would stop changing only when we approach DNS)?

Thank you in advance.

[Ardali]

1. When you don't have solid boundary, LES fits well. Try one of dynamic approaches in LES, dynamicOneEq etc.
2. take a probe point somewhere in the domain and plot the pressure, see the pridicity. If you have complete periodicity, the solution converged.
3 and 4. In jests like this case should not need very fine resolutions, fan jets are large vortical dominated. If you have experimental results, plot averaged velocity fluctuations and energy spectrum. If those are fine you have good mesh. If you gonna try mesh study, try a 5 million cells and see.
Good luck
Ardalan

[smog]

Originally Posted by Ardali

1. When you don't have solid boundary, LES fits well. Try one of dynamic approaches in LES, dynamicOneEq etc.

-- What do you mean by a solid boundary? What is the essential difference between the dynamicOneEqEddy and the OneEqEddy, the filter size? How does that affect the break-up?

2. take a probe point somewhere in the domain and plot the pressure, see the pridicity. If you have complete periodicity, the solution converged.

-- Will plot that.

3 and 4. In jests like this case should not need very fine resolutions, fan jets are large vortical dominated. If you have experimental results, plot averaged velocity fluctuations and energy spectrum. If those are fine you have good mesh. If you gonna try mesh study, try a 5 million cells and see.
Good luck
Ardalan

-- Well, that is the main problem, the jet does not break up on coarses meshes!

Thank you so much for responding, your help is highly appreciated!

[Ardali]

What is your Reynolds number?
Solid boundary means 'wall'.
You do not need to have fine resolution because you do not have wall.
The different between those two models is filter function. The dynamic one reduces the need for the filter size.
See the paper by Menon, Yeung and Kim in computers and fluids 1996

[smog]

Well then I do have solid boundaries "walls" in my case. As I said, the fluid runs through the inlet (cylindrical, dia1) into another enclosure (cylindrical, dia2). dia2 = 20 dia1. The fluid is unbounded by walls when it enters the larger cylinder (dia2). The Reynold's number (based on dia1, inlet diameter) is 6500.

[Ardali]

in practice, you don't have wall where breakup happens.
You should be able to capture breakup easily with 20 million cells.

[smog]

What do you mean "you dont walls where breakup happens"? You have to have an enclosure, right? Thats my enclosure, the larger cylinder. And I am able to capture break-up, it is <20 million (18.5 million or so), I just wanted to know if I can lower the cell count and still capture the physics appropriately.

[Ardali]

The drawback of LES is the need for fine resolution close to the wall. your physics (breakup) is far from the wall. So, you don't need fine resolution with y+~1 and delta y+~1. In your interest region you should be able to capture breakup. Make your resolution slightly coarser in streamwise direction, about 10-12 million. It should give good results if all other things, Boundary condition, numerical scheme, time step etc are correct!
good luck
Ardalan

[smog]

I will let you know what I see when I coarsen the mesh in the streamwise direction slightly. I read online in literature that a maxCo <= 0.5 will only work. So I have been following that guideline.

I was also told that the mesh resolution has a strong dependency on the Re Number for LES. If you happen to know any literature that provides direction in this regard, I would be even more grateful for your help! Also, if I was to quote the Re number for my test case, am I correct in using the diameter of inlet (dia1) for calculation?

[smog]

Well, with a coarsened mesh, my mesh is well < 5 million but there is a problem. My jet breaks up properly in 2D but with the same cell size in 3D, it does not. I have kept everything the same- fvSchemes, cell sizes, CO/interface Co. I don't seem to understand the reason why.

P.S: I am using a cell size of 0.5 mm(edge length).