Failed to mesh with using Prepomax new Gmsh options

¿What about a hole tool?

That would be very difficult to acomplish. If the element nodes do not coincide with the hole boundary, remeshing is necessary.

And if the model is build as full (like in the picture ) and the hole is a partition for the mesher to know it has to put the nodes on top of it?.

HI Sergio, Would you be able to provide the stp full part with the holes as a partition?

Dear @Matej, attached a case were the sweep doesn´t respect the predefined edge refinements, in the yellow part should be 20 elements as in the blue one. In a prior test, taking the front planar face of the yellow part for sweping, it tells me that the part was not suitable for sweeping. Don´t know what I have touched but now is swepeable.

When we define edge refinement, if more than two coincident components (not a compound part) are visible, for which one the refinement is defined?


CAGE_02.zip (331.9 KB)

One thing that would be very usefull is to have a special case of mesh refining to apply in edges, and this refining should left a kind of widget in the geometry showing the number of elements adjusted.

thanks for hint and i can confirm, keeping in even numbers of division for quasi-structured quad to be possible to merge.

body partition of bottom part need to satisfy equal face matching to be possible to merge as top part.

The problem is that you selected all sweep edges. This is first not necessary since setting the mesh refinement on a single edge is enough, and secondly, wrong in this case since the edges have different lengths and setting the element size for all edges results in a different number of elements per sweep edge. If multiple number of edges per edge is defined, PrePoMax considers the last one specified. This will work better when I add the Number of elements per edge option, but for now, only try to prescribe the mesh refinement for one sweep edge.

** Edit **
It is also enough to specify the mesh refinement on the source surfaces. The mesh refinements on the target surfaces are never taken into account.

I cannot reproduce this problem.

I will add “Number of elements per edge” mesh setup item. And I agree it would be nice to see it visually. I can display the “Number of elements per edge” distribution and the mesh refinement distribution with edge seeds all the time and only after one or the other is selected in the tree, the default highlight is shown (like the whole edge, surface,…). And it should be possible to hide the edge seeds.

3 Likes

You are right, I confuse the edge size with quantity of elements that should be the same. I apply in all edges because it was not working :slight_smile:

1 Like

I could use a little help with GMSH sweep of the middle section.
I don’t know if I’m doing anything wrong or it’s a bug in GMSH but when I’m doing a sweep in the radial direction, the length of element edges isn’t equidistant in radial directions which make it difficult to merge the node with other parts, so if someone can come up with solution/explanation then it will be nice.
Just to avoid misunderstandings I’m using a compiled version of Prepomax from https://gitlab.com/MatejB/PrePoMax

I have anticipated that the mesh distribution would not be perfect and coincident in both sides, that´s why I write “use a carefully value for stiching the nodes” of the different volumes. You first need that the nodes be the most close possible that your mesher can do, then start to stich the nodes increasing the tolerance until you get all joined. Don´t remember if Prepomax has a tool to check for such unstiched nodes, some kind of “mesh constrained visualization mode”. Salome and other has it.

@SergioP1975 , thanks for the reply, but my hope had been some for me unknown option which could have equidistant elements edges in radial direction.

Prepomax has the ability to merge coincident nodes within a given distance, but such option can cause distorted elements with a negative jabobian depending on the moved distance so it’s not always a solution.

probably an option to use intermediate or mid nodes (average between) coordinate can eliminate the problem in distorted element, dense mesh or quadratic element type also.

The sweep mesh is created by PrePoMax so Gmsh is not to blame, but …

… the mesh at the side surfaces is meshed by Gmsh. It might be the Quasi-structured quad algorithm that is changing the distances. Try some other algorithm to confirm this. On the other hand, the elements on the outer edge are longer than the elements on the inner edge, so maybe Gmsh adjusts the radial edge lengths to go with this.

It can always be a discussion whether the length of the radial edges should be equidistant or not, but I don’t see any reason why the elements should be completely distorted just because I asked for 2 order elements, so I will allow myself to claim that GMSH has a bug in the Quasi-structured quad algorithm.

This algorithm has quite a few bugs. They even made a separare label for them. Can you report this on their GitLab ?

this is maybe a limitation, some algorithm in Gmsh hexahedral element to fit mid-side in geometry. Required to generate linear first then export mesh result to re-open and set to 2nd order (incomplete), but i did not know in details.

indeed, quasi-structured quad is under experimental similar to hex-dominant. Probably some limitation or bugs exist, but still usable.

I tried other algorithms with transfinite meshing, and the spacing in the radial direction is constant. When I was creating the sweep algorithm, I tried to convince the Gmsh to use transfinite meshing on the side surfaces, but I got a feeling that when the Quasi-structures algorithm is used, the side surfaces are meshed differently.

My goal was to test my own abilities of meshing the CAGE_02 with only C3D20 elements.
The cage_02a, -_02c & -_02d will have very nice elements when using “Revolve meshing”, so for me the tricky part is cage_02b. I found the Quasi-structured quad algorithm would have been absolute perfect if it wouldn’t have been for inequality of the radial edge length and the distortion of the 2’order elements.
If anyone should have a simple recipe for meshing the CAGE_02 with C3D20, then please share and I will be happy.