I have an intersting part to hex mesh for a non-linear problem. It is a multi-step process we will be doing and we want to hex the part, especially the feature that yields. The part has been simplified to remove all of the non-essential geometry and make most of the part, a 1/6 symmetry tube with some swept features, easily meshed with a sweep. There is a small chamfer, seen on Solid 1 in the attached file, that we need to include in the model, but we were planning on going in and adjusting nodes to match it, thinking it would be easier to mesh without it. We plan on aligning the mesh as best as practical and gluing the top and bottom solids together so we don't have to worry about matching the mesh exactly at the bottom of the cylindrical cutout.
I was hoping to sweep most of the part, then revolve around the area of interst, and extrude the remaining space between those meshes. The problem is the critical location is a cutout, and the axis for the cutout does not align well with the 1/6 symmetry. The axis for the cutout is circled in red below. Revolving anything around it does not produce anything clean on the other face of the revolve. This is made worse by the fact that the other face of the cutout feature is the inside of the larger cylinder, so it is not flat. This makes it difficult to revolve the lower part of the mesh, since it needs more than a 'move node to surface' to get it to match the geometry. Sweeping the elements along a curve doesn't get them in the proper orientation, and using the 'Alignment Curve' option was no help (maybe I'm just not using it correctly?).
Next I tried to Extrude to Surface. It requires high AR elements in some locations, since they are extruding to faces that roll off to near tangent to the extrusion direction, but we can sacrifice some element quality away from the region of interest - which is primarily at the top of the cutout. I created several intermediate cuts to extrude to in order to help keep the mesh at the cutout as clean as practical. At each one I selected the new element faces from the planar location and extruded down, but each time the first element did not extrude / project, leaving holes in the mesh, as shown below.
If I do the extrusion all in one go then I couldn't find any way to control the number of elements extruded - it would have the same number on the shortest edge and the outer edge, which is more than twice as long. Breaking the extrude to any location far down the geometry results in highly skewed elements as the surfaces approach tangential to the extrusion vector, and results in the same 'missing element' when the next step in the extrusion is started.
I have attached a model with the geometry and the mesh shown above. Any tips on getting it to mesh cleanly would be greatly appreciated, since I know we will be doing this again on a similar design in a few weeks.
If geometry is to complicate for conformal hexmesh, try to use Glued connection. Check if your nonlinear solver support glued contact.
Some of your mesh dont look good.
And also parabolic tetra is not so bad.
This may not be the answer that you are looking for, because the mesh does deviate from the geometry slightly. The obvious difficult part of this entire hex meshing challange is the removed spherical area with the small scarf cut. I isolated this area, and then used Combined Curves and Combined Surfaces to create a hex mappable body -
I then set up hex meshing on this now six sided solid, and checked the surface mesh on one end -
Next I set up similar mesh sizing on the transition region next to this, and got these two pieces hex meshed and merged -
From here, I just sweeped faces along the other curves to get to this -
From here, one could do some edge splits, and then move surface nodes closer to the original edges that were suppressed by the combined surfaces and curves and get even closer to the original shape.
Thank you both. I can use glue, and I do to join this geometry to a different component, but since this is the critical interface we wanted to keep the mesh as continuous as possible. We also need to be sure that it matches the geometry. I managed a cleaner mesh with some teniques similar to what Mark recommended. I got rid of the little gaps in the steps and was able to improve the elements some, but not enough. The mesh works for linear, but with the still high skew it is not good for the non-linear analysis. The non-linear aspect is why we want to avoid tets, but I may have to tet the critical geometry and glue it to a swept mesh for the rest of the part.
On re-reading everything I have another idea I want to try tomorrow to see if it cleans up some of the problems a little bit more. I am also using a trial of the Trelis meshing software to try and get a better mesh, but that has been a bit frustrating so far so we will see how it goes.