Convergence problem.
Hi
Iīm using the SimpleFoam and buoyantBoussinesqSimpleFoam.at the moment, I donīt consider the turbulence: RASModel laminar; turbulence off; I think that the BC are well defined, so I donīt really understand why i have such a problem in steady case. It could be nice if someone can take a look, maybe you notice where is my mistake(s). fvSolution Code:
solvers fvSchemes Code:
ddtSchemes Code:
Selecting incompressible transport model Newtonian Iīve been reading in the forum, and a lot of people recommend to use the GAMG preconditioner in order to calculate p, any suggestion? increase the number of iterations and the relaxation factors? . If you think that is necessary i could also show my BC files, I donīt do it know to avoid an excess of information. Thanks a lot, Carles |
Dear CRT,
I would change your relaxation factors to the settings below: The settings you have seem a bit too hard for the solver. If it does not run then, maybe there is some other problem. Cheers, Tom relaxationFactors { fields { p 0.3; } equations { U 0.7; T 0.5; "(k|epsilon|R)" 0.8; } } |
Dear Tompf,
Thanks for your replay. I've changed the relaxation factors and it doesn't seem to change a lot. What it seems to work is to use the GAMG solver for the pressure, and also whit some non-orthogonal correctors (5). Then the divergence problem appear just in the Ux, Uy, Uz and the time continuity error, but probably the last is a consequence of the velocity divergence, or? Code:
Time = 18 1. http://www.cfd-online.com/Forums/ope...implefoam.html Quote:
http://www.cfd-online.com/Forums/ope...ady-state.html Quote:
Do someone know a good book, papers whatever that help me to understand a bite more about the different schemes that i can choose? Any suggestion will be rally appreciated. Thanks! Carles |
Dear Carles,
Well it seems like you problems do lie elsewhere. Did you check your mesh, running checkMesh? Does it look ok? If not, than I think you may need to look into your boundary conditions. I do not think the nonOrthogonalityCorrectors are necessary unless you have nonOrthogonality problems reported by checkMesh. Also in that case I would suggest you use the Code:
Gauss linear limited 0.333 I also do not think it is necessary to have a maxIter larger than the standard 1000. I usually only set it to a lower value if there is some instability in the first couple of iterations, just to speed up the entire process. Furthermore I would suggest in your fvSolution to just use a relTol of 0.01 for p and maybe just 0.1 for the other variables. Than I think your numerics set-up should be ok, maybe only play around with lower relaxation factors for the first few iterations. Yes, you can change stuff in the fvSystem/fvSolution file during runtime and it will be taken into account. The initialization may work, but it depends on the problem. First make sure your mesh and boundary conditions are ok, otherwise initialization won't do the trick. So I would suggest to first run checkMesh and maybe check again your boundary conditions, before trying to change your numerics. Good luck, Tom |
Dear Tom,
I think that the mesh is ok. The only prooblems that checkMesh report me is: Code:
***Wedge patch frontAndBackPlanes_pos not planar. Point (0.268943 0.0632487 0.00275887) is not in patch plane by 8.66739e-07 meter. Code:
Mesh non-orthogonality Max: 25.1038 average: 4.18976 Thanks, Carles |
Dear Carles,
Ah so it is an axis-symmetrical case. How many cells does it have? Not a lot I would guess? Could you maybe post your entire checkMesh result and your boundary conditions for U and p? Also a picture of your geometry (preferably showing the mesh) would help a lot. Regards, Tom |
3 Attachment(s)
Dear Tom,
Of course I can, checkMesh Code:
Mesh stats Code:
dimensions [0 2 -2 0 0 0 0]; Code:
dimensions [0 1 -1 0 0 0 0]; Code:
// Average |
5 Attachment(s)
And some screen shots.
|
Carles,
I am wondering a bit about the problem you are trying to solve. If I read your checkMesh file correctly you have a patch GAPTOP that has only 10 faces and it seems from your boundary conditions you want to set this as your outlet. From the pictures I can't really tell what way the flow is meant to go. I think you specify an inlet at the bottom left, but I do not see where the outlet is exactly and also I do not see a patch of only 10 faces. Are you sure that this GAPTOP patch is defined correctly? I think you may want to try to first perform a simulation with a fixed inflow value instead of the mapped one, and maybe generate a coarser mesh to find the correct setup for your problem. If that is found gradually increase the complexity of the simulation until you achieve the correct simulation. I won't be near a computer next couple of days, so I hope you can work out your problem, of someone else can help you. Cheers, Tom |
1 Attachment(s)
Hi Tom,
Yes you're right about the inlet. itīs at the bottom left. The outlet is located at the top right, I forget to attach a picture. Quote:
Thanks for your time!! greetings, carles |
I could find a solution, the problem was in the fvSolution file. Now iīm using this settings:
Code:
solvers |
All times are GMT -4. The time now is 19:22. |