Best Strategy to Generate Hex Meshes within Mecway

rwhirley

I am a relatively new user of Mecway, but have some nonlinear FEA experience with other codes.

What is the best strategy for generating hex meshes within Mecway? I tried creating a 3D model within FreeCad and exporting the STEP file, but when I used "Automesh3D" I got all tetrahedral elements.

As another approach, I tried to export just the 2D sketch from FreeCad (for an axisymmetric part) with the idea maybe I could create a quad 2-D mesh in Mecway and then revolve it to get the 3D part, but the 2D sketch did not seem to import into Mecway at all (I'm not sure it even exported correctly from FreeCad).

What is the best strategy to create a hex mesh with Mecway? My problems usually involve extensive plasticity and contact, and will be solved in CCX. Tet elements don't perform well in plasticity, so I'd really like to end up with a hex mesh.

I'd welcome guidance on how best to approach this.

Thanks,
-Robert

Victor

Hello Robert. Mecway doesn't have a hex automesher. Here are some other ways that might do:

Make a surface in FreeCAD and export it (not the sketch) as STEP and import to Mecway. Under Meshing Parameters, check Surface mesh and Quad dominant. That won't always be 100% quads, but mostly. From there, you can extrude it into hex's. See picture of loop.

Export a 2D shape as DXF. Not sure if/how FreeCAD does this. Then import to Mecway and use Automesh 2D to make a quad dominant mesh.

If it's a simple shape, make it directly in Mecway. For example, you can start with Quick cube then refine and delete elements to carve out the geometry you want. (see attached picture) If it's curved, make a coarse mesh using quad8 elements placed by hand with curved sides, then use Mesh tools -> Refine custom to effectively make a mapped quad-only mesh out of them. (see attached picture)

There are also free hex meshers. I haven't tried them but Gmsh and IA-FEMesh (https://www.ccad.uiowa.edu/MIMX/projects/IA-FEMesh) are two.

Sergio

I have tried IA-FEMesh this week, for "biological shapes" works ok, I have tried a few simple cases (bones) and they create very amaizing full hexa 3d meshes with very few clics. This bones doesn't has sharp edges, I will try some more complex shapes with edges to see if is usefull for my work.

Regards

rwhirley

Thanks. I'll give the above ideas a try.

-Robert

Sergio

Hi Robert, just reading the complete thread, I realize that I'm in the same situation, complex hyperelastic materials, big displacements and contact everywhere.

What I have found productive for simple parts is export the solid as IGS (faces) from the CAD suite, and then mesh with pure cuads in Roshaz one face of the geometry and then export as Abaqus inp and import on Mecway to extrude/revolve/simetrie and have nices hexas....

At the first tries you will find a little ugly the Roshaz interface, but after doing the same work several times you will start to understand it.

Regards!

rwhirley

Salome seems to have some powerful hex mesh tools, although I haven't personally had much experience with them on complex parts. If I can find a good way to get those meshes into Mecway it might be possible to leverage it also.

Since I use the FEA to drive design iterations, I don't like to get too many steps or too much manual labor in the CAD->FEA process or it really slows everything down. The manual labor is really the biggest issue, so if I can easily create the mesh somewhere and bring it into Mecway for loads, BCs, and generation of the CCX input file that may be a good solution.

Sergio

Can you share some samples of the parts that you need to mesh with hexas? I have being meshing with pure hexas since 2001, on I-DEAS and NX. Maybe I could give some advice.

Regards

VMH

IA-FEMesh seems to be able to hex mesh mechanical parts. Sergio, have you tried? What file export does IA-FEMesh offer? Can we import the mesh into Mecway for FEA?

(Images attached from IA-FEMesh manual)

Thanks

Sergio

IA-FEMMesh export in Abaqus inp file, you can create even groups of nodes/faces/elements for creating boundary conditions, so is easy to follow the work on Mecway or CCX. I did'n try for mechanical parts yet, but the joint looks great.

Sergio

Victor, can be meshed only some faces of a imported solid? I meant, if this were possible we could just particionate the solid in the CAD in "swepeables volumes" and then in Mecway try to mesh some faces and extrude/revolve it. Maybe the end side of one volume mesh can be manually copied to the start of the second volume to use as seed for the next volume/mesh. Then the coincident nodes at nexted faces could be merged.

For simple parts can be a kind of "almost fully hexa mesh".

Victor

That's an interesting idea for a feature "mesh selected surfaces". For now you can still do it by meshing all the surfaces then deleting the ones you don't want. To aid in deleting them, you can make a named selection of the faces to keep, then switch to Select elements mode, invert the selection and delete what's now selected (all the unwanted surfaces' elements).

Sergio

I have tried some simple parts, and for very simpe could work (see . I have added a little of complexity and then arise some troubles:

1) We need a wireframe view and possibility of select hiden surfaces to then invert selection and delete elements. Guess that the best will be every surface identified and their mesh also identified separately. As was really dificult (or impossible) to select the hiden surfaces, I has solved exporting in the CAD only the surfaces that I need to mesh (this adds an extra step of planning the mesh secuence)
2) For the bosses of the second part I try by revolute the elements, we need a possibility of select the position of the revolution axis (now is only possible in the global coordinate system axis, but for features not in the center the rotation can't be done.
3) I have tried to swep in place of rotate, it works when I use 2D element edges as guide/paths, but not when the guides are 3D element edges (I have solved creating a set of 1D elements over the edge of the 3D elements.... now how I could delete it?) Even that the sweep don't conect the nodes that are not in the path or guide, and the faces of the last element loose the normal direction, so I make a lot of manual node reposition to merge with the existent elements.
4) Still I didn't find a way to make a symetric copy of the mesh as the symetry plane is not an ortoghonal one. I have tried the point option but it makes a rotated copy, not simmetric

This is my first tries with hexa meshes in mecway, surely the practice would improve. Guess that having the option of just mesh by selected surfaces would improve a lot, and if can be with only quads even better.

Having hexa mesher capacity will be a great plus for Mecway, even more now that people will start to use as prepro for CCX.

Sergio

The first simple is the yellow, the second (not in the pics or model) was a try to make the next part with bosses and draft importing the solid (but I was not able to select the hiden faces), the third (grey) is then the same model but just only importing the faces needed to mesh

Sergio

Tomorrow I will try some parts with radius to see if it can be done.

Regards

Sergio

Hi, following with my tests of the extructured meshing in Mecway, I took the last example and add some radius to the bosses and the tube, places were normally stress will be concentrated and need to have a good mesh. I didn't put attention on mesh quality, just is a test to see if is possible to mesh in a extructured way.

Happily I was able to mesh all the part, and found very usefull the sweep command (more than extrude or revolve) as it permit to keep the element spacing along the path. Extrude or revolve will make perfectly evenly separate elements, but sweep use the path as reference for spacing, and this important to match with the nexted meshed volumes.

Again, will be a great help to have the possibility of using 3d element edges as path/guide and not have to add extra 2d elements. The best would be some kind of "autoselect" for the element edges by tangency or conectivity to avoid selecting one by one.
Another improvement would be that when the 2D meshes were created, is the possiblity of have every surface mesh grouped automaticly (is just create a named selection for every surface), so then sweep/extrude/revolve operations become more easy

The grey elements are those that I was forced to delete and recreate (on 2D meshing before sweeping) to keep the node conectivity. First I tried to move the nodes manually, but is quicker, easier and precise (to don't lose the plane) to delete and recreate manualy.

I'm really impressionated by the things that can be done with Mecway!

Regards

Sergio

Until we convince (hey, we need more people, please make noise in this thread!) Victor to add fully 2D quad mesh in Mecway, in order to have a full hexa mesh the parts (faces in fact) can be meshed (2d seed quad meshes) in Roshaz and then export to Mecway to do all the sweeps and extrusions. I will make this same part in this way to see the difference.

Regards

Sergio

Until we convince (hey, we need more people, please make noise in this thread!) Victor to add fully 2D quad mesh in Mecway, in order to have a full hexa mesh the parts (faces in fact) can be meshed (2d seed quad meshes) in Roshaz and then export to Mecway to do all the sweeps and extrusions. I will make this same part in this way to see the difference.

Regards

Sergio

This is the same parte now with quad mesh in Roshaz and then sweeped in Mecway...the procedure was really straighfull, and now is full hexa.

Guess that the next step will be meshing in IA-FEMMesh also.

VMH

If we can do all hexa mesh in one step with IA-FEMMesh than that improve the workflow quite a bit especially when we have to make some geometry changes per design requirement or even client request the last minutes.

Sergio

In my experience hexamesh is a slow process and very manual, at least for complex parts. The first times that you mesh with bricks will take a time to understand the process and tools availables, and learn the mesh patterns for your commons features.

Then, in production, when you mesh a part for first time you must plane how the model must be cleaned (some features will not be possible to mesh with hexas) and divided in order to get the "swepable" volumes, and do the mesh. In the next desing iteration the mesh must be done from the begining (yes, normally the "update mesh" just strugle your model), but now as you know hot to do it so it takes a fraction of the time.

On the other side, this meshes are for complex analysis (contact, no lineal, hyperelastic materials...), so not switable for last minuts requests :-)

I have tried to mesh in IA-FEMMEsh this same part (obviolusy usin multiblock), but I didn't get a usefull mesh in the few minuts that I have to try.

Regards!

[Deleted User]

I never heard of IA-FEMesh. I tried it out and had no luck. Tons of elements with a negative volume every time. Most FEA programs that I know of rely on the libraries from Visual Kinematics. This is the link to their meshing library; http://www.vki.com/2013/Products/VisToolsMesh.html. It does everything you would want. However, I'm sure if Victor were to use libraries from Visual Kinematics, the price of Mecway would skyrocket. I'm personally fine with Netgen. It does a great job and it's free. Having a quad dominate mesh has never done anything great for me. It may look nicer but the results are the same. Depending on the model it may reduce the equation count some, but not enough to matter in any case I have run. You should be able to solve any model you wish with beams, shells, or tets. Quad dominate meshs aren't that important, in my experience at least.

VMH

I also been meshing in Ansys with its Multi-zone meshing feature to get all hex mesh but haven't seen any significant different in the results compared to Tet10 using linear static analysis with linear elastic material model. The hex mesh look much nicer though. I haven't tried other material model and other type of analysis yet.

[Deleted User]

Here is another company that offers meshing tools for auto quad dominate surface meshes and automatic volume hex meshes; http://www.meshgems.com/index.html
Don't know how much it costs. I think a lot of these tools are cost prohibitive. I like that Mecway is low cost for home use. Adding features like this are likely to be very expensive. Pretty sure these are the meshing libraries ANSYS has used for a long time.

[Deleted User]

I was testing out Netgen 6.1 Experimental and was surprised to see it can now auto generate 3d hex dominate meshes.

Victor

That's amazing news!

[Deleted User]

Darn, bringing the mesh into Mecway, I see it is still only a surface mesh.

[Deleted User]

What tripped me up is you also have to change another mesh option. The last step needs to be changed to Optimize Volume. I attached a picture of the settings.

svingegneria

Hello to all,
Im preparing a simple workshop to import structured mesh from salome platform, exporting in .med then converting in .msh in GMSH. Please let me know if this can be of your interest.
Best Regards,
Stefano

Sergio

Awesome mesh Stefano! I'm interest not only in how to translate to Mecway but also in how was done in Salome-Plataform.

Regards

Sergio

Just to give another option to hexameshing, I have found in this new release 6 (not sure if is a new feature), an option to convert tet4 elements into four hexa8 elements, that is, a mesh of tetras to full hexas. They are not so perfectly shaped as if they were extruded or revoluted quads, but they will give very razonable results for the cost (time to mesh it). The tool is under Mesh Tools/Change Element Shape

This is the example part uploaded by UGMENTALCASE, first meshed with tet4 and after converted to hexa8 with Mecway.

I have several comparations using tets vs regular hexas and this "converted hexas", and they perform much better than tets in terms of stiffness and stress for big displacement analysis.

kwip

I tested this approach some time ago as a customer followed the rule to accept hexaedral meshes only. Make sure the tets are in in good shape.
For reference:
http://onlinelibrary.wiley.com/doi/10.1002/cnm.485/abstract