CGAL users discussion list

["Walker Andreas" <>]

Hello,

I am working on a project where I require a periodic mesh for a structure 
which I use later on in a (mechanical) FEM - context. I generate this mesh 
with a code very similar to the one presented at 
https://doc.cgal.org/latest/Periodic_3_mesh_3/Periodic_3_mesh_3_2mesh_implicit_shape_8cpp-example.html
 - which usually works like a charm. The function I use is a 3D-Interpolation 
from Grid - Data from a decomposition simulation (this simulation is not my 
own work).

Now this works when the grid data is taken from earlier stages of said 
simulation, when the 3D-field shows rather fine-grained structures and the 
values are regularly distributed within the allowed range (which is between 0 
and 1). The structure I'm concerned with is typically "lamellar" i.e. 
consists of many connected lamellae which are rather fine and thin at this 
point. Now I move to later stages where these lamellae become rather thick 
(and fewer). Also, the function values get very close to 0 and 1 and have a 
steep gradient inbetween. Further on, there are also some numerical artifacts 
around these gradients, making the isosurfaces a bit rough around the 
grid-points (not sure if that's of any relevance here).

Now when I try to use the exact same code for such a case I get the following 
error at runtime: 

what():  CGAL ERROR: postcondition violation!
Expr: min_power_distance_to_r < 0.5 && min_power_distance_to_q < 0.5
File: 
/opt/libraries/cgal-4.14.1/include/CGAL/Periodic_3_mesh_triangulation_3.h
Line: 521
Aborted (core dumped)

It took me a while to find a set of mesh criteria under which the code runs 
successfully and produces output, which however only work for rather coarse 
meshes. (To be specific, the cell size needs to be rather large (about twice 
the lamella thickness) and the cell radius to edge ratio higher than 3.5. 

My question is: what causes this error? What can be done to avoid it? Are 
there any features of the input data that cause it?

Thank you in advance and please excuse the lengthy question with the probably 
unnecessary details

Best Regards,

Andreas Walker
Master Thesis Student
m&m Lab
MAVT, ETHZ