I am trying to perform Finite Element simulations, including linear static and a non linear static analysis, across two different models. The geometries were meshed using the “Full Automatic Tetrahedralization”, with a second order mesh requested.
After various attempts at solving the problems, ending with errors from the solvers (both Code Aster and Calculix), and getting messages pointing to bad Jacobians, it is obvious that the problem is in the meshes, more specifically, the numbering of nodes.
This could or could not be related with the fact that the meshes are quadratic, but definitely is a problem with the algorithm, in my opinion.
My questions are:
Am I right about the cause of the errors?
Is it only me that have this problem? Is it a known bug?
Is there a workaround besides meshing outside simscale and importing the mesh?
Is this a project that you can share publicly? I have had difficulty with second order meshing in the past, but for me it has been much better the last four or five months.
Hello @ggiraldo@cjquijano,
the problem is indeed a meshing issue that we are currently trying to fix. If all goes well we will have an improved version of the second order meshing algorithm online next week.
The actual problem does not come from wrong orientations, but from “collapsed” elements, similar to this:
Sometimes when trying to set the middle nodes on the geometry the algorithm merges those together and creates completely flat element domains.
Actually there is no workaround other than waiting for the next release where the situation should be improved.
It’s due to memory allocation problem. It says : Error during the dynamic allocation. It was not possible to allocate a memory zone length 204226624 (bytes). Either choose a bigger machine or reduce your mesh size. I would recommend to do the later one. With second order elements it will be good to go with a normal mesh.