Note: This discussion is about an older version of the COMSOL Multiphysics® software. The information provided may be out of date.

Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.

Mesh dependent results

Please login with a confirmed email address before reporting spam

Hi, All

I solved a simple electrostatic problem for computing the potential field (V) in a tube (axisymmetric problem) filled with water. At one end I imposed a voltage of 500V and grounded the other end, one of the wall is defined as axis of symmetry while other has a zero charge/symmetry boundary condition . I calculated the electric forces by writing an expression of the form (f = epsilon*LaplacianV*delV) , however when I am changing the grid density (increasing the mesh density) the value for the electric forces keeps on changing. I am unable to conclude. Is there any other alternative available to calculate the electric forces. I need to calculate them in order to utilize them in Navier-Stokes equations for the CFD part of the problem. I am using version 3.5(a) and I am using mapped meshes.

Thanks

3 Replies Last Post 2 août 2012, 13:43 UTC−4
Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 27 juil. 2012, 04:14 UTC−4
Hi

I can come with some comments, as I have also noticed a high sensitivity to the mesh density. But when you look at your formulas, you see they contain a few derivatives of the dependent variables, Therefore you should also plot these derivatives, to see how they bahave, and to check that your mesh is corretly resolving the gradients (and gradients of the grad) that might build up.

Often you have singularities around sharp corners, sometimes using fillets can improve the result, other times a structured regular coarse mesh wil lgive better results than a finer radom thet mesh, depends on the geoemtry.

Another, way is to increase the discretization (default second order) to third forth or fifth order, this akes longer to solve but your derivatives become smoother, with a risk of some oscillations at the boundaries (to be checked, as its model dependent again.

Finally using another path for the force estimation is a good idea, as verification means but I not sure like that what to propose, and I'm currently far from my reference books and documentation (holyday time ;).

--
Good luck
Ivar
Hi I can come with some comments, as I have also noticed a high sensitivity to the mesh density. But when you look at your formulas, you see they contain a few derivatives of the dependent variables, Therefore you should also plot these derivatives, to see how they bahave, and to check that your mesh is corretly resolving the gradients (and gradients of the grad) that might build up. Often you have singularities around sharp corners, sometimes using fillets can improve the result, other times a structured regular coarse mesh wil lgive better results than a finer radom thet mesh, depends on the geoemtry. Another, way is to increase the discretization (default second order) to third forth or fifth order, this akes longer to solve but your derivatives become smoother, with a risk of some oscillations at the boundaries (to be checked, as its model dependent again. Finally using another path for the force estimation is a good idea, as verification means but I not sure like that what to propose, and I'm currently far from my reference books and documentation (holyday time ;). -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 31 juil. 2012, 06:18 UTC−4
Thanks Dr. Ivar for your suggestions. I will check for higher discretization schemes, however by the word discretization you mean to say the higher order elements for the dependent variable?
Thanks Dr. Ivar for your suggestions. I will check for higher discretization schemes, however by the word discretization you mean to say the higher order elements for the dependent variable?

Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 2 août 2012, 13:43 UTC−4
Hi
indeed you find it under "discretization" of the main physics node (you might need to turn on some of the detailed nodes in the options preferences)
Sorry this is for V4, the naming might be different in 3.5, i have forgotten, the idea and the approach of COMSOL remains the same and has not change w.r.t. this point for the new version

--
Good luck
Ivar
Hi indeed you find it under "discretization" of the main physics node (you might need to turn on some of the detailed nodes in the options preferences) Sorry this is for V4, the naming might be different in 3.5, i have forgotten, the idea and the approach of COMSOL remains the same and has not change w.r.t. this point for the new version -- Good luck Ivar

Note that while COMSOL employees may participate in the discussion forum, COMSOL® software users who are on-subscription should submit their questions via the Support Center for a more comprehensive response from the Technical Support team.