yade-dev team mailing list archive

[Bruno Chareyre <bruno.chareyre@xxxxxxxxxxx>]

Hi,

I'm committing  a first update in minutes (3.6. Periodic boundary
conditions). Do you agree Vaclav with those sentences (else feel free to
fix or suggest changes) :

(1) "It is believed that the first method is generally more convenient,
since it will let  :yref:`Cell.trsf` reflect only the transformation
produced during the simulation, independently of the initial period
geometry."

(2) "In all cases, the period geometry should not be modified during a
simulation, be it via Hsize, trsf, or refSize. The velocity gradient
:yref:`Cell.velGrad` is the only variable that let the period
deformation be correctly accounted for in constitutive laws and Newton
integrator"

Side comment on (1) : it will need to change the behaviour of setTrsf
one day if we really want to uncouple trsf and Hsize. For instance, one
could like to type trsf=identity after a deformation process (i.e. after
some iterations) in order to set the current state as the reference for
further deformation. It should have no effect on Hsize. Currently, it
would mess the simulation big time.

For (2), am I too restrictive?

Additional comments :
- be carefull with cumulated deformations, it is a product, not a sum
(the equation is fixed in doc). There are very few cases when (dF+I)F =
dF+F (small strain assumption would make it true for instance). I've
seen that flipCell is using a sum, it could well be one of the few cases
giving sum/product equality, but I'm not really sure. Using a product
would be on the safe side.
- It seems you forgot to update the doc when you replaced
Newton::homotheticResize by Cell::homoDeform, but I love extra work too. ;-)

Bruno

-- 
_______________
Bruno Chareyre
Associate Professor
ENSE³ - Grenoble INP
Lab. 3SR
BP 53 - 38041, Grenoble cedex 9 - France
Tél : +33 4 56 52 86 21
Fax : +33 4 76 82 70 43
________________