@Matej this looks awesome. Did you think in some way to allow to clean the geometry before meshing? Very often even if they are partitioned to mashables volumes, they have some edges in the lateral faces of the sweep that prevent the swepping. Is a kind of “delete edge” and “sew edge” for the division that may left. And some times you need the reverse, split and edge, and ad artificial edges to help/guide the sweeping in the laterals walls.
In Abaqus, you either handle that with more partitions (never enough of them
) or use Virtual Topology. It’s a simple yet powerful feature that allows you to ignore some edges and faces (particularly if they are small and irrelevant).
There is already a built-in split shell feature, but the cleanup features will come in the next step.
Sweeping is also improved, so it has fewer geometric constraints. Multi-to-one sweeping is supported, and the side edges forming the sweep wall can be split into more than one segment. It is not perfect yet and will not be when released, since it is highly dependent on edge cases rather than on theory alone. But I’ve already tested it in many cases (time is the limit). Also, defining seed edges - mesh refinement might be problematic in some cases. But I feel it is good enough to be released and tested.
If cells are symmetric, they are automatically meshed as such (it is a user’s setting).
Great, let’s run the upcoming dev version through some thorough testing then. It looks really amazing and if it’s indeed as close to Abaqus workflow then indeed even quite complex models should be hex meshable with a few clicks (finally without having to struggle with tedious plane/surface partitioning in FreeCAD).
My favorite and most commonly used partitioning tool is Extend face and I see that you already added that.
Mine to. And in PrePoMax multiple faces can be used for splitting at the same time.
There will be a way to specify that the different cells should be conformal? I did some meshes with the last version of Prepomax using diferent sweep directions, but I has to force the node count in the edges and after merge the nodes to get the final mesh.
Yes, the point is that the entire compound is meshed with a single command. The source and target surfaces of the entire compound are automatically detected and assigned a direction. Then the compound is meshed with a coincident mesh, resulting in a single solid part.
Impressive this mesh.
Impressive your work.
My intention is to try to organise to serve as helpfull: minimum to the enormous job you are doing.
I am going to organise to start doing Calculix and FEAMaster validation models. On the other hand, have you thought to have the option of CODE-Aster?
Bad news: OpenRadioss project is discontinued. Now is Radioss Simcenter ![]()
It would be good to set a GitHub repo and maybe a different communication channel. How about Discord ? It’s much more efficient for such brainstorming. FreeCAD also uses it for development nowadays instead of a regular forum.
Incredible
, without any clear explanation, a shame. But if they were open source for a while, could it be that some one fork the code?
FreeCAD may have it in the future, there’s an open PR already adding some basic features (shells and composites). The problem is very specific syntax.
Typical for such large companies. It was too good to be true. I guess they just wanted community help for a while before closing it.
But was there any announcement ? I only see that the repo is gone.
Yeah, they even admit it:
After four years of successful community-driven research, Siemens is entering a new chapter for the Radioss solver.
Found it: The future of Radioss is Simcenter | Siemens
I guess there might be some existing forks (and I should have an older copy of the code), but otherwise, we are stuck with CalculiX’s explicit dynamics now ![]()
This looks really impressive. The sweep seems to be really stable, even though there are no full partions of some structure (green lines). Will the sweep really be possible, when the target face has a subdivided surface?
@JJJC @FEAnalyst
About OpenRadioss:
The last accessible snapshot is from March 2026:
GitHub - OpenRadioss/OpenRadioss: OpenRadioss is a powerful, industry-proven finite element solver for dynamic event analysis
The last version I was able to download was from july 2025:
web.archive.org/web/20250923074545/https://github.com/OpenRadioss/OpenRadioss/archive/refs/tags/latest-20250717.zip
Yes, very, very, typical. Imagine that I work with Autodesk products… I think that Altair followed a strategy (possible to be more visible the company), and Siemens didn’t like it.
But they cannot change the publicly released licenses for the past?
Wed I designed it, it was possible. Many-to-many with no constraints on the number of patches. But then I partially removed this possibility. The problem is that I was not able to project the edges of the target face to the source face so that the source face has the correctly split elements. So when the sweep was applied, the geometric lines on the target disappeared. And I understand this as a topology change. So I reduced it to cases where the internal target edges can be recognized/mapped to the internal source edges.
In the image, the side-sweep edges of the filets are missing, but the sweep can still be made automatically without splitting the geometry. And since this is automatically detected, the user does not need to specify any mesh setup items; they only click Mesh to get the default result.
What about contacting to the mainteiners of this site? they could become the hosts for the validation cases for any solver, they have several cases well documented and start to work on NAFEM as well





