yade-dev team mailing list archive

[Václav Šmilauer <eudoxos@xxxxxxxx>]

>   Just to say to be carefull with formula in Hentz's thesis: the idea 
> was  to use as inputs for contact properties the Young modulus and 
> Poisson ratio that we want to obtain macroscopically for the granular 
> assembly. The "funny term" are use to compute kn and ks with respect to 
> these macroscopic Young modulus and Poisson ratio
Yes, I know. I disregarded that funny term, it is close to 1.0 and
doesn't play any role.

What struct me, though, was that we deviate from any obvious definition
of contact stiffness. Its dimensionality is correct, but it is scaled by
some dimensionless constant (π/2 in our case) away from the "intuitive"
definition (which is the base of what Hentz uses): stiffness of cylinder
with radius min(r₁,r₂) between spheres' centers, with some average
Young's modulus.

I say nothing if such constant is properly documented and supported by
some reasoning, but for me now, even though it works, it is just garbage
code.

> In my opinion, the only important point is to compute kn and ks such as 
> scale effects are avoided, and I think it is well done in the current 
> formula (*) in Yade.

I disagree with the premise that the only important point is to avoid
scale effects. I like quantities to have physical meaning, as much as it
is meaningful with discrete models; the argument (Bruno ;-) ) that
discrete solution doesn't converge (and doesn't approach continuous
solution) if you refine "mesh" (packing) does not justify, in my eyes,
gratuitously introducing random constants to the code.

Besides that, imagine Chiara reading that code (in a few days)... guess
what happens? ;-)

v