Hello SimScale team,
I am having issues in generating finite element mesh for my mechanical model. Also the model is not converging even with simplest of material properties and boundary conditions. It is the last model Smaller 20km_elasticMantle_SW 15km_DOC on this project link.
Thank you
Viven
Hi @sharm760, thanks for posting on the forum,
Since your model is made up of extrudable parts, a good idea might be to use an extrusion mesh refinement in each of them. I’ve done this in this project link I’ve shared with you: SimScale Login.
Also, I see that you’re applying a force in a surface to which a symmetry plane is assigned. This won’t allow dislocation to happen in the direction normal to the plane and the force won’t have any effects.
Also, let me know if we can share your simulation to the public so that we can post screenshots/public projects here!
Best,
Igor
Hi,
I have been using the same setting for the past 10 months or so and had not run into any issue until now. It seems that the engineers have changed something under the hood changing how the mesh processing is done or how it is evaluated.
This thing concerns me the most about the non open source computational platforms (such as SimScale) where from my side I cannot know or control that what has been changed with the processing pipeline.
For example with the same settings previously mesh was getting generated in just 1-2 min with 32 cores but the same is taking over 10 min now (see attached pic).
As a result I have been just spending time trying to generate decent mesh instead of running models and analyzing results just because most likely something changed at your end.
Viven
The settings from the latest mesh are quite different from some of the earlier meshes. Local element size refinements are finer, there’s a larger gap refinement factor, there is a manual “curvature” definition for the local element size/global mesh settings, etc.
PS: you will virtually never need to use more than 4 cores to generate a mesh for a structural simulation. If it gets to a point where you need 8 cores to generate a mesh, you will likely end up with a massive 20M+ nodes mesh, which you probably won’t be able to run.
Hi,
My question remains simple, did SimScale’s mesh processing pipeline change over the past month or not?
As I said that I have been using same settings to generate meshes for my simulation for the last 10 months or so and never ran into such issues.
The differences you pointed out were only because I had to try other things to make a decent mesh.
No matter what settings I choose I am not able to get the aspect ratio lower than 8 in my geometry. This is the same geometry I have been using previously also.
I need to sort out these things before I can make any further progress in my research project.
Viven
Yes, there have been updates in the meshing tool over the last 10 months.
You can basically replicate the meshes from 10 months ago by using extrusion mesh refinements.
PS: the poor cells are likely generated due to a dirty CAD model (see the extremely small faces below in red).
I’d suggest cleaning those up to increase the quality of the mesh.
Now, it would have been better if you would have released those updates as an optional newer version to SimScale meshing. This is because when we address research problem we want to keep things standardized at different levels of simulation setups. For example I had already run convergence tests with the previous settings to make sure that those discretization was fine for my research problem. Now with this change I may have to do it again.
This is the reason why companies release updates in a newer version. For example Mathworks releases multiple versions every year so as to not cause problems with MATLAB or Simulink models that were designed with previous versions.
Hope this makes sense.
BTW, thanks to your suggestions I am now able to fix the mesh but still the model is not running even with simplest of boundary conditions.
Thanks
Viven
