yade-users team mailing list archive

[Bruno Chareyre <bruno.chareyre@xxxxxxxxxxx>]

The bias on radii is fixed in last revison (r1865).

Bruno


Bruno Chareyre a écrit :
> chiara.modenese@xxxxxxxxx a écrit :
> > Hello,
> >
> > quick question on one aspect of the Triaxial Test. As a beginner, I 
> > am just trying to run some triaxial tests but I have got this error 
> > message:
> >
> >
> > Exceeded 3000 tries to insert non-overlapping sphere to packing. Only 
> > 1055 spheres was added, although you requested 2000
> >   
> The initial positioning of spheres is done by generating random 
> (x,y,z) and checking if a sphere of radius R (R also randomly 
> generated with respect to a uniform distribution between mean-std_dev 
> and mean+std_dev) can be inserted at this location without overlaping 
> with others.
>
> If it the sphere overlaps, you generate new (x,y,z)'s until you find a 
> suitable place for the new sphere (*). This explains the message you 
> have : after 3000 trial-and-error, the sphere couldn't be placed, so 
> the algorithm stops.
>
> I guess you tried to generate a dense packing, which is not possible 
> with this method. You should keep the default value of porosity.
> The dense state will be obtained in the second step (compaction, see 
> below).
>
> > BTW, could you also explain me the meaningful of the following 
> > variables, since I was unable to find where actually they are 
> > employed in the code.
> > maxMultiplier,
> >
> > finalMaxMultiplier,
> > maxWallVelocity.
> >
> >   
> Compaction is done (1) by moving rigid boxes or (2) by increasing the 
> sizes of the particles. Both algorithm needs numerical parameters to 
> prevent instabilities. For instance, with method (1) maxWallVelocity 
> is the maximum wall velocity, with method (2) finalMaxMultiplier is 
> the max value of the multiplier applied on sizes.
>
>
> (*) I just realized there is a bias in radii generation. If the sphere 
> does not fit at (x,y,z), a new R is generated. Which means small R 
> tend to be a bit more frequent that big R, even though the 
> distribution is supposed to be uniform between Rmin and Rmax. I'll 
> commit a fix in a few minutes.
>
> Bruno
> >   


--

_______________
Chareyre Bruno
Maître de Conférences

Grenoble INP
Laboratoire 3SR - bureau E145
BP 53 - 38041, Grenoble cedex 9 - France
Tél : 33 4 56 52 86 21
Fax : 33 4 76 82 70 43
________________