Hi,
I have recently started using Griddle for meshing. And I found that the mesh I generate have a bad planarity of zone faces compared to other generation methods (which makes sens in a way as Griddle is used to handle more complex geometries).
Overall the meshes I generate are of good quality based on the Griddle log file, but when checking them in FLAC3D if find that have a large portion of the zones with planarity above 0.1. This confuses me because this planarity criterion does not seem to align with the shape quality criterion that Griddle uses.
Therefore, I have te following questions:
- Is there a way to constraint the planarity of zones in GVol? Based on my reading of the documentation of Griddle 3.0 it does not seem to be the case as the shape quality criterion used by Griddle seems to use a slightly different way of computing warping of faces.
- How does the Griddle criterions align with the flac3D criterions.
- How bad is a planarity between 0.1 and 0.3? The FLAC3D documentation is fairly sparse about it. It only says planarity should be << 1.
- Are there papers covering this topics? I could only find this 2013 paper (11-02.pdf) and it limited to investigating orthogonality and aspect ratio.
Thank you in advance!
Hello, Stephane,
Looks like we missed this post earlier, but let me reply to some of your questions now.
Griddle can and often does generate Hex-dominant meshes with non-planar quadrilateral zone faces - there is no way to restrict this. This, however, does not introduce any problem in FLAC3D due to the concept of mixed discretization in which zones are further subdivided into tetrahedral subzones ( Grid Discretization in FLAC — Itasca Software 9.6 documentation ). Now each of such subzones has only planar faces. Note that all stresses and strains are calculated in each of such subzones and then averaged in a special way for the whole zone. This increases accuracy of stress calculations and avoids some of the numerical artifacts (to some extent this is similar to the higher-order integration in FEA).
In general, non-planar quadrilateral faces do cause small differences between quantities calculated and averaged in each overlay, compared to the case of planar faces (imagine splitting a non-planar quadrilateral over one or another diagonal - in one case you’ll get somewhat more concave geometry and in another - more convex). However, when averaging all the quantities for the whole zone and then considering all the connected zones, the overall effect of non-planar faces will be negligible if faces are not too far from planar (in general, Griddle never creates badly non-planar faces, unless provided geometry / surface meshes contain such faces).
Griddle incorporates face planarity into the single compound equation for element shape quality (See Griddle 3.0 manual, page 41). A single metric is used for simplicity of comparing meshes generated when using different parameters. FLAC3D, on the other hand, historically used separate criteria and also reported the worst among them. In general, face planarity is calculated using somewhat different equations in Griddle and FLAC3D but both are internally normalized to the interval 0 to 1, so they produce comparable results.
Overall, no matter which criteria you use, the main idea is to increase mesh quality as much as possible and preferably use any shapes but tetrahedrons in FLAC3D simulations.
Dear Andrey,
Thank you for the reply. This clarify the relation between Griddle meshing and FLAC3D.