What would be the best way of meshing a flange-like part with radial holes, which is technically a piece of the compound part. Do the pieces of the compound part all need to be meshed with the same meshing method?
Best case i can see if i could mesh some pieces with the Extrude meshing method and the others with the Transient meshing method.
Currently, the rules are that extruded and revolved meshes can’t be used for compound parts so only surface partitions can be used (unless you mesh each part separately and then connect them with tie constraints). On the other hand, transfinite meshes can be applied to compound parts so you can use volume partitions to obtain proper regions (parts) for hex or hex-dominated meshing. Another rule is that each volume (region) used for transfinite meshing must have 5 or 6 faces with 3 or 4 edges each.
another approach, explode the part in CAD to produce shell instead of solid. Mesh that shell with Quasi-structured quad options, then use thickening feature in PrePoMax. Just remember to make sure the thickness in CAD has the same with input value required.
it’s recommended to use hexahedral element whenever possible. PrePoMax have feature for such purpose by methods of extrude/thicken and revolve, transfinite (partitioned). Only use automatic tetrahedral for complex part. Non-conformal or mismatched node mesh between each part can be connected with Tie constraint or Tied contact type.
However, it all depending on analysis case. Usually an automatic tetrahedral with quadratic type is sufficient for most case, i.e elastic analysis or nonlinear plastic in small strain conditions.
The answer, i guess, comes with experience. The cost of my own time far exceeds the cost of CPU time, ths is the biggest influence on my meshing strategy. For me it results in using hex meshes only where it is obvious to do so, if the geometries are not hex friendly I don’t waste time fighting it.
Some companies require hex meshes in simulations done for them (even when the geometry is complex). But normally yes - it’s often not worth the effort. Of course, it also depends on the mesher’s capabilities.
indeed, in the past i have small conversation by email with Klaus of CalculiX also. Internally use hexahedral quadratic (structured) in their professional works. Latest development version of PrePoMax also given possibility for easiness to achieve this, user only needed to do partition which conveniently depend on their CAD software.
It’s from an older version of Gmsh, right ? Because you’ve mentioned in other threads that this algorithm was removed, then brought back (in 4.12 ?) but doesn’t work so well anymore.
right, older version ones. It’s ran so slow but works and sometimes failed, the newest version seems doing optimization to be faster. Unfortunately, result not good as previously and developer also notified this on progress, prepared or ready in future version (v5.x)
i following their GitLab, Twitter, and author papers. Maybe new question can be start by asking result generates compared to previous.
i still stick to older version, it a bit laggy (frequently freeze) but working at almost my case. Below simple example of previous test, results shown good agreement with another when maximum stress/strain located at boundary layer. Also, it can be better than structured hexahedra due to inappropriateness of partition.
A few days ago i had an awesome idea how this can be done, but in practice it works only partially. I’ve created the flange element and meshed it individually with Transfinite meshing algorithm but when i make a compound part it does not work. It somehow sees that there are elements that do not have 5 or 6 faces.
Wouldn’t it be logical if all the pieces can be meshed with Transfinite algorithm then the whole compound part should be as well?
Here’s a way to do it, albeit not sure how good it is.
It’s not the only requirement. Each face must have 3 or 4 edges. Sometimes even seam edges (the ones added to cylindrical faces in CAD software) can invalidate this.
