Rigid body constraints shouldn’t be used with shells in nonlinear analyses, this leads to non-convergence and even an immediate error in CalculiX 2.23.
Regarding the solid model, you have C3D6 and C3D8 elements - they can overstiffen the model in bending. Try at least C3D8R or C3D8I, but I would just switch to second-order elements altogether. Some global refinement and better meshing algorithm (Gmsh with hex-dominated mesh generation via extrusion or sweep with quasi-structured quad algorithm) may also help. In your model, it seems that you just used the Thicken Shell Mesh tool on the ready shell mesh, right ? This is ok if you generate a good shell mesh using Shell Gmsh and surface partitions to make use of the transfinite 3- and 4-sided faces.
What supports do you want to model here ? Why do you leave the rotational DOFs of the rigid body rerefence nodes unconstrained and why don’t you apply the BCs directly on the edges or faces of the holes ? Is there a description of the model from another FEM software available for reference anywhere ? And are the reference results from linear or nonlinear analysis (what nonlinearities ?) ?
You could also utilize symmetry here and model just a quarter of this plate. Just be careful with symmetry BCs on shells - fixing the drilling DOF leads to overconstraint and even non-convergence.
There might be differences in how supports are defined. Here, you are using a rigid body constraint, so making the edge/surface of the hole completely rigid and following a single reference node whose translations are fixed (but rotations aren’t). Some software may use flexible spiders instead (CalculiX has them too; they are called distributing couplings, but they have various limitations and aren’t available in PrePoMax yet).
I assume that material properties are the same in all cases (you used a material from PrePoMax’s library, that’s why I mention it).
Then, for nonlinear cases, you are only considering geometric nonlinearity ?
You could try with this model (refined 2nd order mesh, using symmetry):
I’m aware of the RIGID BODY limitation. I replaced the hole with a rigid plate and TIE connection. The deformation results are acceptable. The model cannot use symmetry because it’s training for asymmetric tasks.
They may improve if you refine the mesh and use second-order elements or at least ones with reduced integration.
Right, you may not be able to open those files in older versions.
Yeah, I’ve noticed the name of the file “szklane”. I’m Polish too - if you want to talk in this language, we can move to private chat (on the public forum we should keep using English).
I downloaded version 2.4. You’ve fixed the hole, so there should be free support in the rotula. I know how to do this and I use support, but that’s not the point. Interesting. Are you using a “Quasi-structured quad” without Rercombine and have squares in the slab?
I fixed the edge/surface of the hole and, as you said, thought you could try BCs more appropriate for your desired setup.
Yeah, that’s the only 2D algo not requiring recombine. Others need it. Also, it provides the best structured quad meshes in most cases, but it can’t be used with extruded meshes (you would need a sweep mesh, but it has some additional limitations) to generate hexahedral solid elements.
Also, for verification of linear cases with analytical results, you can take the portion of the plate between the holes and treat it as a plate fixed at corners: Plate Supported On All Corners With Uniform Load
This yields the deflection of around 7.2 mm in your case.
