First of all, make sure that master surface is the one with a coarser mesh.
Tie constraints may attract artificial stress concentrations in addition to structural notches causing stress peaks and singularities. Especially if the meshes are highly dissimilar. You may try making them more aligned.
It’s often best to avoid tie constraints and instead merge the geometries in CAD software or use compound parts in PrePoMax, but this may prevent hexahedral meshing (since only the Transfinite algorithm can be applied to compound parts and it requires volumes with 5/6 faces having 3/4 edges each). Alternatively, you can use the Merge coincident nodes tool in PrePoMax, but it needs closely aligned (almost coincident with a given tolerance) nodes on both sides.
You can try increasing the position tolerance (just make sure that adjustment is disabled). Only the slave nodes within this distance (by default 2.5% of the typical element size) are connected with the master faces.
First of all, make sure that master surface is the one with a coarser mesh.
Yes, i confirm that the coarse mesh is on the master surface
It’s often best to avoid tie constraints and instead merge the geometries in CAD software or use compound parts in PrePoMax, but this may prevent hexahedral meshing (since only the Transfinite algorithm can be applied to compound parts and it requires volumes with 5/6 faces having 3/4 edges each). Alternatively, you can use the Merge coincident nodes tool in PrePoMax, but it needs closely aligned (almost coincident with a given tolerance) nodes on both sides.
I would like to maintain a hexahedral mesh for a better quality result.
You can try increasing the position tolerance (just make sure that adjustment is disabled). Only the slave nodes within this distance (by default 2.5% of the typical element size) are connected with the master faces.
I have already set the tie constraint with a distance of 0.025mm (typical element size equal to 1mm). After that the simulation that i have shared has been done with a distance of 0.2mm (20%) but the problems are the same.
To obtain a aligned mesh between master and slave surface have you got some method to share?
I would like to attach the step file but as a new user i can’t.
You can try with much higher position tolerance, tie constraint can work even on large distances. And it should be fine as long as you don’t have adjustment enabled. Of course, you can later check if it didn’t go too far.
In new PrePoMax versions, you can try setting local refinement with the same number of elements on both sides (previous versions support only size-based local refinement). Unfortunately, it doesn’t work with all meshing algorithms and some are rather stubborn and hard to align for both sides.
You could use some hosting website too. I sent you a message with instructions.
Some parts in your model are not hex meshable out of the box because they don’t meet the criteria for transfinite or extruded/sweep meshing. Of course, you could keep these parts meshed with tetras or add some partitions to allow hex meshing.
When it comes to your question about aligning meshes (to reduce/avoid stress concentrations), you could add some face partitions as shown by the red lines:
This would force the nodes to at least cover the bounds of the blue edges. Then you can specify the number of elements for them and for the other side (if you have an edge in this direction too). Such partitioning should be done in CAD software.
PrePoMax allows me to generate a mesh using a specific meshing strategy only for a single part or a merged compound part.
For example, if I split the shaft into multiple parts, I only have two options to obtain a conformal mesh:
Use Glue Common Nodes, which requires matching meshes on the contact surfaces.
Use Tie Constraints, with the risk that some nodes may not be properly tied together.
In ANSYS, for example, I can mesh each body independently and use a bonded contact definition, which seems to be more tolerant and flexible than a tie constraint when the meshes are not perfectly matching.
The best way is Transfinite Gmsh algorithm. It’s the only hex meshing algorithm in PrePoMax applicable to compound parts. However, as I’ve mentioned before, it requires all subvolumes to have 5 or 6 faces (with 3 or 4 edges each). If you can split your part into such subvolumes then you will get a continuous hex mesh for it. Otherwise, merged nodes or tie constraints are necessary, as you said.
Speaking about that, you could also try tied contact in PrePoMax. It’s similar to tie constraint, but uses contact algorithm. Thus, it’s more computationally expensive and less robust, but may potentially provide better results in terms of spurious stress concentrations, at least for shell-solid connections.
Respect for your patience in dividing parts. I guess I don’t have that much patience in FreeCAD. It seems to me that in your example, the hexagonal elements are very deformed, and the stresses in the weld will be distorted. Will the weld yield better results with wedges? The mesh of connected elements is continuous, so there will be a good fit there.
This isn’t a ready mesh, just a quick example showing how to partition faces to have nodes more aligned at the connections. Of course, that mesh should be refined. And more partitions are needed to get a better mesh also for the cylindrical part.
Being able to partition faces/solids,similar to abaqus would be a huge benefit for prepomax. Hex mesh is great, but a tedious task if you have to go back and forth to freecad ( where these partition tools are also rather own-developed python scripts instead of existing functions).
For simple geometries this may require no loop, but when the geometry gets more complex it is better to have the ability within the preprocessor to find other paths to mesh the object.
Is there a plan for the future, to implement such functionality?
For now. FreeCAD is the best option. It can be rather manual and tedious, but FreeCAD has very good Python scripting interface. You can easily make a macro to automate some partitioning operations, e.g. extended face volume cutting (I did it for one of the cases discussed here).
