simpleFoam no convergence
Hi,
I'm trying to set up a case with simpleFoam solver without turbolence. The geometry is a pipe with a diameter of 3mm and a length of 117mm. U: The volume flow rate at inflow is 5 ml/s, so the velocity field is Code:
inflow The density rho is 1065kg/m^3 (blood) The dynamic viscosity is 3.4e-3 Pa*s (blood) So my cinematic viscosity is: Code:
nu nu [0 2 -1 0 0 0 0 ] 3.2e-6 pressure boundary condictions are Code:
inflow I'd like to use a laminar model, so I set: Code:
simulationType laminar; The solution doesn't converge but if I change the inflow velocity to Code:
inflow Code:
nu nu [0 2 -1 0 0 0 0 ] 3.2e-3 What is wrong in my case? |
Hi,
you've reduced inlet velocity and increased viscosity, so Re went from 663.15 to 0.00066315, that's why it converged quickly. What's the number of time steps in your original simulation (with higher Re), what's in your SIMPLE dictionary? |
thank you for the reply
after 1000 timesteps the residuals are horizontal but there is no convergence this is the SIMPLE dictionary Code:
SIMPLE |
Well...
1. what are the values of residuals for pressure and velocity (flat ones)? 2. I suppose there's no error in the direction of the velocity (i.e. it really goes along the pipe not in normal direction) 3. is it cylindrical pipe? can you show checkMesh output? |
1) residuals:
Code:
smoothSolver: Solving for Ux, Initial residual = 0.000598572, Final residual = 4.88836e-05, No Iterations 3 3) checkMesh output: Code:
Mesh stats |
What's in your fvSolution? What if you switch GAMG/smoothSolver to PCG/PBiCG?
|
What means change GAMG/smoothSolver to PCG/PBiCG is obscure for me (I've to study the solver before run the simulation!! I know).
Please can you help me to set up the dictionary? The second question is: I've another geometry with a variable diameter. The lower value is 1.2mm and the reynolds at that section is about 1600. I think I've to use a turbolence model. How change the dictionaries in order to activate this feature? Thank you very much this is my fvSolution. Code:
solvers |
Well,
you can just take fvSchemes/fvSolution dictionaries from $FOAM_TUTORIALS/incompressible/simpleFoam/pitzDaily. In the same tutorial you'll find all necessary modifications to run case with turbulence. by switching from GAMG/smoothSolver to PCG/PBiCGI meant changing: Code:
solvers Code:
solvers |
It doesn't converge.
But the residuals are low! or not? :( Do you think the problem is in the mesh? these are the residuals: Code:
DILUPBiCG: Solving for Ux, Initial residual = 0.000595124, Final residual = 1.49812e-07, No Iterations 1 |
Are you sure your mesh is 3 mm in diameter? checkMesh thinks it is 3 m in diameter
Code:
Overall domain bounding box (0 -1.4918 -1.5) (50 1.4918 1.5) in this case, you've got Re = UD/Nu = 663145.3125 and it's obviously non-laminar case and therefore you've got no convergence as you're running without turbulence. |
now the mesh is scaled
Code:
Overall domain bounding box (0 -0.0014918 -0.0015) (0.05 0.0014918 0.0015) Code:
DILUPBiCG: Solving for Ux, Initial residual = 0.000565348, Final residual = 2.30843e-05, No Iterations 1 |
1 Attachment(s)
attached here there is the plot of the residuals
|
1 Attachment(s)
I don't know what I'm doing wrong but attached case converged in 143 iterations. I've taken velocity and viscosity from your post. Two points that are different from your case:
1. Fully hexagonal mesh. 2. van Leer scheme for velocity discretisation. (if you'd like to run attached case, you'll need Gmsh to create mesh) |
1 Attachment(s)
thank you Alexey!
I'll try with gmsh. Attached here you can find my polymesh directory. |
With attached polyMesh directory case converged in 42 iterations.
|
1 Attachment(s)
and here the case
|
There's not convergence with "limitedLinear 1.0" scheme but if I change it to upwind it converges in 42 iteration (64 iterations with GammaV, 68 iteration with vanLeerV, 65 with linear).
|
with your fvSolution and fvScheme it converges in 67 iterations.
Where can I change the scheme from "limitedLinear 1.0" scheme to upwind in order to reduce the number o iterations? A lower number of iterations is very important because I need to run the case recursively because I'll couple it with a 0D model. |
Your current fvSchemes:
Code:
divSchemes Code:
divSchemes |
Ok. I'll try this evening with the different divSchemes.
Now It converges. Thank you very much Alexey for your help!! I've lost a lot of hours with this problem! |
All times are GMT -4. The time now is 22:36. |