Salome deformed elements

DaveStupple

Hi all,

I've started using Salome to generate conformal meshes. It has not been easy, but I'm starting to get the hang of it. I imported a 2D mesh as a .unv and Mecway flagged up two elements as having problems (see picture). The one on the left could lay some claim to being a tri6 (maybe a bit too thin), but the one on the right resembles nothing. In Salome the mesh evaluation reports only tri and quad elements, with no polyhedra. The lower quad region was generated as a sub-mesh using quadrangle mapping with default hypothesis and wire discretization with no. of segments; the rest was generated by quadrangle mapping with the 'standard' quadrangle parameter chosen and wire discretization with local length. Some localized smoothing has been done. I will patch this up easily enough in Mecway, but do any Salome users know what I did wrong and what I might have done to correct it?

Sergio

Dave, can you share your geometry? Maybe with a different partition scheme could lead to a better discretization.

Regards

Victor

Not that I know anything about Salome, but that whole mesh looks like just a first pass without some optimization or shape improvement steps that would normally smear out those perfectly regular grids and high aspect ratio triangles or split them into better shaped elements.

DaveStupple

Sergio, I am sure you are right. I have introduced a degree of partitioning to try to improve the discretization. I think maybe the transition from smaller to larger elements needs to be better managed. I can't share the geometry publicly but I will try to send you the geometry in a message. Thanks for this, I appreciate it.

Victor, I do have some very high aspect ratios, but these are quite localized and not in areas where I expect much to happen. I include another image to show the region in its local context. My intention is to revolve the 2D mesh in Mecway by 0.5° or so about the y-axis with a single subdivision and put frictionless support on the two large surfaces to give me a 3D model representing an assembly with 360° radial symmetry for a thermal stress analysis. In doing so, all the large elements with good aspect ratios in the 2D mesh will acquire horrendous aspect ratios in the 3D revolve, but the small elements where I expect the most action will retain OK aspect ratios. That's the plan anyway. Always happy to take advice on a better approach.

Notice in the image that in the other (lower) large radius is one element that is a different colour, so there's something odd going on there too. I maybe should have been a bit cleverer by excluding some of the radii that are away from the action.

Sergio

The trick is avoid the straight line in the partitions/divisions of the orange surface. The curved partitions must start about the half of the radius

DaveStupple

Sergio, thank you for that, that is an excellent tip. What a fantastic example of how to share an idea quickly and clearly with a simple sketch. We use the expression 'back of-an-envelope' (used to be fag packet, not so common now). Americans use 'back-of-a-napkin', I think. I would not have thought of using curved partitions. Thanks again, keep them coming!

DaveStupple

Sergio, I've gmailed you.