How do you do your hex/quad meshes?

I’m keen to know what the rest of you out there use when meshing hex or quads, to then use in Prepomax.

I’m working on my SALOME skills but
i’m finding it quite complicated to do anything beyond very simple stuff, and not many videos of advanced meshing exists. The docs are good, but not descriptive enough for me to work it out on my own, unless i have A LOT of time. Gmsh seems like even more work, but i dont have much experience with it.

I even enquired about the price for the Meshgems “MG Hex” plugin in SALOME, and they charge 6000usd for a year!

Let me know what you do.

Gmsh is nice but hard to use. I showed some basics of it in one of my tutorials. Salome is also far from being user-friendly but it’s still closer to what you may expect from a GUI-based mesher. It has quite advanced tools including powerful hexa block meshing technique. I will probably include it in one of the future tutorials.

Quad meshes are easier and sometimes can be prepared in PrePoMax using quad-dominated meshing. Otherwise, Gmsh and Salome should help.

1 Like

I made some hexa meshes in Salome, and much more making extrusions or revolutions of quads also in Salome. Lots of works, I agree, that’s why I try to minimize the use of this kind of meshes to where really I need it.

In the past had used I-DEAS, Siemens NX and Abaqus CAE to do pure hexa meshes of rubber parts, all of them were amazing.

1 Like

i’m using Salome, to my understanding to generate structured pure hexahedral need lot of work in partition.

if the model look like extrusion based (translate or rotate) this can be less.

also, someone still can mix between hexahedral and tetrahedral connected by tie constraint or tied contact type if required.

regarding to commercial structured hexahedral mesher, i can not suggest any. but Ansa Beta CAE or Cubit look good, sometime i look their steps to know the capabilities only.

it’s so depend on complexity of geometries. personally, Salome is workable in most of my case.

Yesterday I found PolyCube, an all hex mesher. Saddly is not free for comercial use. I was able to mesh some parts, the first test that I did (rubber mount) amazed me, but after I start to find several issues as scaling, and the mesh gets a litle deformed, flat surfaces that became curved, some models doesn´t mesh at all, and the lack of fine tune of the mesh, all the elements are the same size.

http://www.cs.ubc.ca/labs/imager/tr/2018/HexDemo/

I had to install Meshio, and then the procedure was:
Meshing with first order tets
Export as .msh
Convert from .msh to .mesh with Meshio
Convert from .mesh to vtu with the provided tool mesh2vtu (the .vtu generated by Meshio doesn´t work, that´s why all the ugly conversions)
Run the tool (pipeline.bat)
Import the mesh, and carefully adjust the size, every time the scaling was different!

1 Like

There is a lot of research going on in the field of hexmesh generation from the tetrahedral base mesehs. I am following the research as good as I can to find a possible solution for PrePoMax but did not find one yet. The one proposed by @SergioP1975 looks nice but I don’t thing it keeps the feature edges. Or am I wgong?

2 Likes

it seems an opensource ones available out there from Wenzel Jakob etal called Instant Mesh and Hexa Dominant, i known previously since he also the creator of Mitsuba Renderer (Raytracing) i’ve been use

did not further investigate cause of some personal doubt about the results, is this suitable for Finite Element? some or many element unconnected or distorted.

been waiting GMSH hexa dominate mesh to improve and work again, below picture i get from 2016 versions.

Likely the best solution for potential future implementation in PrePoMax would be Gmsh due to its openness and possibility to work without native GUI. Their hex meshing project is still in development: contrib/hxt · master · gmsh / gmsh · GitLab

Of course, it would be great to have a hex mesher in PrePomax but on the other hand, even software like SolidWorks Simulation supports only tetrahedral meshes so it’s not that hexahedrons are something absolutely necessary for all professional projects.

2 Likes

it seems to be right, GMSH hexa dominate is quite simple by include CAD files (Stp,Igs,Brep) in command script then define algorithm with only few setting.

Merge "file.stp";

Mesh.CharacteristicLengthFactor = 0.5;

Mesh.Algorithm = 9; //8 = delquad or 9= 2D R-tree
Mesh.Algorithm3D = 9; // 3D R-tree
Mesh.Optimize = 1;
//Mesh.Recombine3DAll = 1;
Recombine Volume "*";
Mesh.Smoothing = 0;

someone can experiences above script, unfortunately not working anymore in latest versions.

1 Like

Here’s a script from my YT tutorial PrePoMax (CalculiX FEA) - Tutorial 20 - Hex mesh from Gmsh - YouTube

Merge "Tutorial 20 - hex mesh import.step";
SetFactory("OpenCASCADE");
Transfinite Curve {4, 3, 2, 1} = 6 Using Progression 1;
Transfinite Surface {1};
Extrude {0, 0, 2000} {Surface{1}; Layers {80}; Recombine;}
Physical Volume(13)={1};

It was prepared using Gmsh GUI and generates extruded hex mesh for a rectangular beam, based on its smaller face.

1 Like

Interesting project @SergioP1975 ! What was your conclusion then? That it was not worth pursuing ? Sticking with SALOME?

Nice @synt. Did GMSH use to have HEX dominated mesh options, and then removed it? I wasnt aware of this.

Hmm. Yes, perhaps you’re right. I really don’t like the GUI interaction of Gmsh. Perhaps i should re-evaluate. I honestly just want to find THE RIGHT mesher to get good with, and then stick with it. I just dont want to flip-flop between different ones all the time.

One thing though, i’m not sure i agree with the Solidworks simulation thing. Yes, they don’t do Hex or Quad meshes, but to be honest, i’m not sure i feel that Simulation is really suited for anything serious. I believe the premium version actually used to have Hex meshing, back in 2016 or so, and was then removed. To me at least, it seems quite obvious that Dassault spend no energy on improving Simulation, to instead focus on SIMULIA Abaqus. I have access to Simulation Pro at work, and i prefer PrePoMax/CCX whenever i can. If not for the true shell and beam elements, i would probably not use it at all.

On a side note: i’ve thought about the possibility to buy ABAQUS solver tokens, to be able to compare results from PPM/CCX, without having to buy a license for the flagship stuff. I’ve heard this should be possible, but havent enquired with my retailer. Do any of you do this?

1 Like

It depends on the case. Some geometries are so complex that it doesn’t make sense to struggle with hex meshing and tetrahedral meshes are sufficient as long as they are dense enough. Of course, it’s a good practice to use hex elements whenever possible but it’s just not always possible (or feasible). SW Simulation is enough for many engineers dealing primarily with machine design and using FEA only for rather basic design evaluations.

Yes, you can have solver tokens without CAE (GUI) ones if that’s what you mean.

1 Like

my example script previously generate hex dominate mesh for complex shape of CAD files.

not removing, but due to many improvement and changing the algorithm and make it crash.

i seen latest FEMAP hex dominate shown similar result, it’s interesting the technology apply.

like quad dominate mesh currently possible with Netgen inside PrePoMax. only with few of setting to make all it done.

regarding ideal and pure hexahedral mesh, as i understand only benefit for linear type in large multipart contact analysis (implicit) or extreme large deformation e.g metal forming, rubber self contact (explicit)

on my personal test, even the element is hexahedral and look structured still need to use with caution for distorted shape. automatic tetrahedral quadratic eliminate this problem since it will generate uniform mesh, also mesh refinement at some areas are easy to apply.

previously, if i remember correctly. Ansys also took many benchmark at common problem of tetrahedral and hexahedral quadratic element, result shown no large discrepancy.

That’s right, distorted hex meshes might be worse than good-quality tet meshes. Of course, some types and formulations of elements are particularly susceptible to distortion. Here’s one of the benchmark problems from Abaqus documentation considering also this aspect: https://classes.engineering.wustl.edu/2009/spring/mase5513/abaqus/docs/v6.6/books/bmk/default.htm?startat=ch02s03ach133.html

About PolyCube, by the moment I will not use because:

Unable to control precisely the element quantity at certains parts of the geometry
Final shape is slighty changed from the initial mesh, scale can be adjusted, but the shape of the part can not.
Some minor details (were the stress can be interesting) are loosed

i’m only took simple case from Bischoff etal (2008) as shown below,

2023-02-25 05_45_40

someone can easy to recreates, using 2D shell element or 3D solid by extrusion subjected to in plane / out of plane shear, moment/torque,etc.

even using quadratic element type, still many model results are less accurate compared to uniform triangular or tetrahedral quadratic.

1 Like

@Matej, PolyCube doesn´t keep the feature edges of the original part. At first sight looks amazing/magick, but when you start to check the mesh carefully youl see several issues.

Just for reference, meshing a part as the rubber mount with pure hexas using partitions to carefully respect the geometry (even including draft angles and all the fillets) and put exactly the desired quantity of elements where it needed, using Siemens NX would take at least 3-4 hours if you are lucky and have several years of training and expertise. Surely if you modify the part and want to recreate the mesh probably will be done in half hour, because you know were to put the partitions (for that particular part) to get the mesh, but the initial one takes a lot of thinking.
On the other side, using Polycube took half hour to figure it how to use it and configure Meshio, and then a “similar” hexa mesh was done in 10 minuts, on the second trial. But the fact is that this automatic mesh is not 100% accurate, as the part was deformed, and worst some radius were not well meshed.

This is the kind of mesh that I’m talking about, I did it several years ago while I work for a big automotive industrie.


t

1 Like

I am following the web page https://www.hexalab.net/, where different results of various research papers are presented. Some of them look really promising.

1 Like

i did not know exactly, but it seems something has changed. i do roll back to version 3.0 and found it worked again. mesh option need to be set properly

2023-03-02 00_36_18-Options - Mesh

2023-03-02 00_41_49-Statistics

i exported to readable in PrePoMax as UNV and shown below, it’s solvable but i did not check further.

i will try again to the latest version 4.x and hopes it’s working also with many improvement.

*edited
it seems to be right, an algorithm for hex-dominant mesh with R-tree is not available there.

2023-03-02 01_17_23-Options - Mesh

1 Like

hi @Tor

may i’m wrong, i read their discussion and found this feature being disable in distribution for some reason, but it can be activated by community request. below the quotes,

This is not supported in Gmsh 4, though. Although a new implementation is in the works, if you find the implementation in Gmsh 3 useful, we could temporarily re-instate it in Gmsh 4. Let us know.

it’s a pending to distributed due to experimental states, being improved and prepared for version 5.x

1 Like