Faulty elements

DaveStupple

In trying to adapt an old mesh, I isolated the attached set of elements that are a transition region from smaller to larger elements. This appears many times in the larger mesh, but there are just these 6 basic elements that should be filling a 3D rectangular space. Elements 2 (P5), 6 (H8) and 4 (P5) are very odd, each appearing to have at least one curved face. Probably nothing more than selecting nodes wrongly, but would I not expect Mecway to signal an error (v1.1 or v3.0 trial)? 'Open cracks' function doesn't show this up, and no elements show as collapsed. I should say here that the mesh was originally constructed in LISA software.

VMH

I think the reason why Mecway doesn't report error because there seems to be no error in regard to element type and connecting nodes (no open). They are just poor quality/skewed. Mecway currently doesn't check for those. If you mesh in Netgen, you can check for those.

Plot 1 show the element as HEX8 with 8 nodes (poor quality/skewed). Plot show the element PYR5 with 5 nodes (poor quality/skewed).

If we were to mesh in Mecway (Plots 3 and 4), the element quality is much better as compared to the referenced mesh but all elements are tetrahedral elements (linear as shown but doesn't get good results. Quadratic elements (with midside nodes) are much better).

Victor

Yea, I don't think these elements is actually wrong, just quite distorted. Mecway shows an error (red X) if the element is so distorted that the solver will fail (negative Jacobian determinant, non-physical shape like self-intersecting/negative volume, etc.).

I made a similar mesh with better shaped hex8 elements and the temperature field is pretty much the same, so it's still quite a usable mesh for a transition region, you just have to be careful with heat fluxes, or stresses in mechanical analysis.

DaveStupple

Thanks both. Your demonstration, Victor, shows why the thermal results I got were quite plausible. I'll try to do this better next time. Unfortunately for me, nearly all of my meshing is done 'by hand', so I should be able to avoid this kind of mistake. The unusual selection of element types is to grade from smaller to larger elements. Thanks again.