Discussion Forum

Hi

indeed I have also some issues with the unsymmetrical of the meshing, it's worth to send the question to support.

I agree that the boundary mesh does not look nice, but I'm not sure its changing that much the results

If you really need a symmetric mesh, its worth to use a sweep or revolve mesh,

A few other comments to allow you to gain time:

when generating the geometry, you can make 3 concentric circels + a rectangle on the left (r<0) side and do a difference, add the two largest, suppress the left square and the smallest one

Then to make several points, you generate one point and use the rotate or array operations

By the way if you add those points for the mesh as hard points, then its better to define that in the mesh section. You can mesh the edge with a regular or irregular (i.e. exponential progressing) mesh

You can also copy meshing on edges surfaces ...


--
Good luck
Ivar

Hi Ivar,

Thank you for your reply.

Unfortunately the mesh affects the results. That is in the mph file I had attached before, if you swap the boundary condition in the central half circle and in both cases you calculate the volume integral of the dependent variable u, you will see that the values are no the same while they should be as we have just mirrored the geometry with no change in the physics of the problem. It is a concern that the asymmetry in the mesh affects the final results.

Thanks also for the rest of your comments. Would you please let me know if there is any tutorial that tells me about "a sweep or revolve mesh" (if it exist in 2D), "use the rotate or array operations",

About the points on the inner half circle, actually I added them to make width of the elements shorter when I am using the boundary layer. I don't know how to set the width of each boundary layer mesh, so I added multiple points on the boundary so that it forces to reduces the width.

Sorry for so much questions, but I don't know how to do these "By the way if you add those points for the mesh as hard points, then its better to define that in the mesh section. You can mesh the edge with a regular or irregular (i.e. exponential progressing) mesh. You can also copy meshing on edges surfaces ..."

The idea of concentric circels + a rectangle on the left (r<0) side was very interesting.

Thanks,

Amir

Hi

I do not know about any particular tutorial, there is the excellent course on advanced meshing by COMSOL (see the main web site) but then I learned mostly by meshing many cases, including looking at the examples here on the forum and answering. I would say that I'm continuously improving by spending an hours time on the forum, each day ;)

Check the edge meshing, its better to work on the meshing separately, if you start to mesh an edge then you give "seed points" for the next mesh level. Setting "hard points" as you were doing, on the geometry is handy for post processing and analysis purpose, but not ideal for meshing

Depending on what you are doing, check the frame you are integrating on, as this often gives larger errors, than some meshing irregularities.

But I agree, I have to some models not solving to the level of required precision, because of difficulties to symmetrise the mesh, for a symmetric part, but with a unsymmetrical load


--
Good luck
Ivar

Hi Ivar,

I finally couldn't make the symmetric mesh, however, I learned many things from your comments!

Thanks,

Amir