Commit d63e2ca2 authored by Kevin Moreau's avatar Kevin Moreau

improved documentation

git-svn-id: fbbead7c-fb4d-4173-aa67-51132c73c120
parent 5bed31d2
......@@ -106,7 +106,7 @@ To check if several test case works: (for example "dyn_" group)
Good practices
(These are my personnal good practices that I (Kévin Moreau) tried to respect:)
* On the header
......@@ -143,3 +143,34 @@ Other Informations
* Please don't change material parameters directly inside data directory,
data are chosen and used in the script launched by ctest.
If it is changed then reference results are not the same.
* How does restart work?
It is realised by the helper class PreProcessing. We assume that we
restart at step n. Several data are needed in order to restart:
- disp field (step n-1),
- phi field (step n),
- old phi field (step n-1),
- nonlocal domain (element only) (step n),
- computation mesh (if crack already appeared) (step n).
Things to know:
- disp field is step n-1 one, it is given as initial field to the
newton-raphson loop in Algorithm. If the problem is linear and
solved directly (elastic_damage is chosen), there is no need to
know this field, but it is loaded anyway.
- The computation mesh is exported using AOMD implementation which
exports in ASCII. Once reloaded, node locations slightly differ
because of the truncation of node coordinates. To test if restart
restores the exact same state, uncomment the two lines of code
in TLSGeom constructor
and comment
comp_mesh=new xMesh;
pre_pro.loadMesh("comp", *comp_mesh);
This test works only before the emergence of a crack (when d<1).
Without this modification, you cannot restore the exact same
solution unless you code a binary export/import of the mesh. Good
luck if you go for this!
- The integration mesh has to be rebuilt, as well as partitions,
it implies the regeneration of computation mesh sadly.
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment