CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   OpenFOAM Meshing & Mesh Conversion (http://www.cfd-online.com/Forums/openfoam-meshing/)
-   -   Amount of cells needed? (http://www.cfd-online.com/Forums/openfoam-meshing/116389-amount-cells-needed.html)

erichu April 18, 2013 03:53

Amount of cells needed?
 
1 Attachment(s)
Hi,

I have previuosly posted here related to a nozzle mesh problem which has been solved with a fine mesh, good schemes as well as suitable turbulence model.

However, I am now extending my model to the complete geometrical entity which is a 1.2 meter long lance with varying radius(minimum radius 1mm, max radius 4mm). My goal is to calculate the flow rate of water (velocity, density=constant) of the complete lance using known pressures at inlet and outlet, 50 mbarg and 0 barg, respectively. Reynolds number is approximatley 200 (from experimental resutls) so I use laminar flow settings.


At the moment while using simpleFoam, my convergence is very slow and approaches constant residual higher than my required tolerance. Using relaxation factors for U, p around 0.01 - 0.1 leads to a continuing minimisation of the residual, however, very slow. Looking at late intermediate iterations in paraView, it appears that the velocity profile isn't correct as the flow has not a shape of a parabola (maybe explained by the fact that the mesh has a lot of cells in the center of the tube).

It is becoming clearer and clearer that my mesh is most likely not good enough, however, what that exactly means in my case, I am not sure of.

My current mesh is based on the geometry seen in the attached file, consisting of about 100 000 hex cells.

My question is maybe too general but what should my mesh look like to improve convergence and a good result that compares to the experimental data?

Thanks for your input

A.Wendy April 24, 2013 05:10

Quote:

Originally Posted by erichu (Post 421439)

My question is maybe too general but what should my mesh look like to improve convergence and a good result that compares to the experimental data?

you can never say this in general. if you want to be sure that your mesh is ok in a numerical point you have run it with a mewsh and refine it until there is no change in the results. Than your result is independent of the mesh.

100.000 cells are not much. when your mesh is base on blockMesh and SHM just douple the cell numbers in the blockMeshDict and try to compute.

best wishes

Andreas

fredo490 April 25, 2013 04:21

Just to confirm, your mesh is 3D ?
The best thing to find if you mesh is good is to look at your data with paraview or any other software. Try to find the high gradient regions and check if you have enough cells in the gradient direction. To build a good mesh from the first time, you need to have a good guess of the flow structure and you also need to use all your previous simulation experiences.


All times are GMT -4. The time now is 09:16.