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.
Interpolation of geometry according to chosen element type
Posted Sep 11, 2017, 10:03 a.m. EDT Mesh, Structural Mechanics Version 5.2a 6 Replies
Please login with a confirmed email address before reporting spam
I am new to Comsol and have a basic question.
When modeling a curved geometry (solid mechanics) meshed with a coarse mesh with linear interpolation, the curved parts gets edgy due to the limited number of elements. To fix this, I would like to choose a higher order of interpolation, i.e. cubic serendipity or lagrange elements. I do this under 'Discretization' under the Solid Mechanics node.
However, it seems that I can only achieve this for the displacement field alone, which leaves me with an edgy geometry, with only the displacements interpolated. I would like the intermediate nodes of the elements to be fitted to the geometry according to the shape functions defined by the element. How do I achieve this?
David
Please login with a confirmed email address before reporting spam
When higher order elements are used, the intermediate nodes are in fact fitted to the curved boundaries although they are displayed as if their edges were straight. I believe it's to make the rendering faster.
Best,
Jeff
Please login with a confirmed email address before reporting spam
Thank you for the swift reply.
Is there a way to bypass this feature? For my application purposes I usually prefer to use fewer higher order elements rather than a lot of lower level elements, however it's quite difficult to assess wether the mesh is fine enough, if I cannot see how the elements fit the geometry...
I've attaches a simple example where it's clear that the meshed geometry is edgy. However, the stress field also seems quite strange, as some part of the curved part of the geometry seems to be fitted to the intermediate nodes of the elements, and other parts seems merely linear. Am I doing something wrong?
Thank you.
David
Attachments:
Please login with a confirmed email address before reporting spam
If you look at the message pane, you will see warning messages like
"Used linear geometry shape in 2 mesh elements to avoid inverted curved elements."
Your mesh is inadequate. Several elements will have 0 or 180 degree angles in the limit of higher order representation.
Regards,
Henrik.
Please login with a confirmed email address before reporting spam
Thank you for your reply.
To be honest, I hadn't seen the warnings. I suppose an inverted element means something like the elements gets distorted to an hourglass-like shape, relating the mapping between coordinate systems with a negative area? I'll do a re-meshing in that case.
Also, is there a way to bypass the inaccurate rendering of the meshed geometry when using higher order elements?
Thank you!
David
Please login with a confirmed email address before reporting spam
In graphics in the Mesh node, the edges are always shown as straight, since the meshing does not know which discretization the physics will use (it could even differ between two physics interfaces in the same model, or between different study steps for a single physics interface).
In a result plot, the edges are shown with curvature. In that case, it is the Resolution setting in the Quality section which controls the graphics. This is clearly seen in your model. If you change to No refinement, those elements which were forced to have a linear geometry (according to the warning messages) are not affected, whereas those who have cubic parameterization are straightened.
Regards,
Henrik
Please login with a confirmed email address before reporting spam
Henrik and Jeff, Thank you very much for your help and precise explanations. This concludes my inquiry.
David