Local mesh refinement - hanging nodes

Altair Forum User

When I attempt to generate a 2D mesh using the automesh tool (QI optimize method), I have difficulty producing a mesh with local refinement. If I mesh the surfaces for which I desire local mesh refinement and then mesh all other surfaces, the result is often essentially two separate meshes. To put it another way, it often looks like a superposition of the coarse mesh and the local mesh with many hanging nodes. In some regions there will be a region of transition and proper connectivity as desired but it is a fairly rare occurrence with my methodology.

 

I have set the quality criteria for the coarse mesh such that the smallest permissible element size (failure) is smaller than the smallest element created in the initial, refined mesh of certain features. The desired element size is not far off between the two meshes. For the refined surfaces I am attempting to obtain an average element size of 1mm. For the remaining surfaces I am attempting to obtain an average element size of 2mm.

 

Any tips on obtaining acceptable interfaces between two surface meshes with different average element sizes would be very much appreciated.

Altair Forum User

Using the size and bias 2D Automesh method yields perfect coarse-fine mesh connectivity but the mesh quality is much poorer. However, the QI optimize method does not yield a perfect mesh either and so either method is logically followed by efforts to improve the mesh quality. So I seem to have a solution.

 

If I may, I recommend efforts by Hypermesh developers to investigate mesh connectivity when using the QI optimize method of the 2D Automesh tool as described.

 

A couple of other suggestions:

 

When I save surfaces, or any other entity (elements, etc.), I cannot retrieve these entities after closing and re-opening the model. For my current task, for which I select a subset of the surfaces of the model for a finer mesh, this is a major annoyance. My solution is to leave Hypermesh running at all times but eventually the workstation will be restarted. I will try to save the relevant section of the command file but in my experience most users, especially novice users, do not use the command file.

 

The Geometry Auto Clean-up function does not seem to recognize that two (nearly) parallel edges separated by a very small distance (less than the minimum acceptable characteristic length of an element) will yield poor mesh quality. The Automatic Geometry Clean-up tool does not toggle one of these edges.

 

Similarly, two edges which intersect with a very small angle (much smaller than the minimum acceptable angles for tri/quad/tet elements, for example) will yield a poor quality mesh at this location. There is a connection in the software between the Automatic Geometry Clean-up and the mesh quality criteria but the Geometry Clean-up tool ignores the basic relationships described above. I would rather toggle a couple of edges to re-activate these entities than toggle a large number to de-activate them. Surely the software could decide which edge to 'de-activate' by attempting to implement the described functions and maximize the length of interconnected, active edges. Software would undoubtedly do a much job than us error-prone humans.

 

The command file seems to capture nearly every action by the user. Would it be so hard to implement a GUI (element) to store and allow for manipulation of a history of commands? At the very least an 'undo' function is long overdue. Even if it would be necessary to start with the original imported geometry and repeat every command, requiring some not insignificant CPU time to do so, I and many other users would surely prefer this crude implementation to the dearth of options we have now (I must admit I am only using HW12 and I am ignorant of the changes made to newer versions).

 

Hopefully this is seen to be less of a rant than I suspect it appears to be. Hypermesh/works is quite user friendly in comparison to many other CAE pre-/post-processors and solvers.