yade-users team mailing list archive

[Václav Šmilauer <eudoxos@xxxxxxxx>]

> 
> I have never used UniaxialEngine but from the last Vaclav's comment we
> can get a better idea of how it works.
> If you run a compression test, you end up creating a number of chains
> in the models through which you somehow transmit the load. 
> If we look at the triaxial, for instance, in the triaxial phase we
> apply a load on the upper wall to compress the specimen and if we
> check the stresses at the bottom wall, they are more or less the same
> (if the strain rate is small enough).
> I am actually thinking how it would work in a tension example... If
> you have any further explanations I am quite curious..

Nothing exceptional, interactions are created at the first step with
enlarged boundign boxes, then you load in tension by displacing boundary
particles.

Relevant paragraphs from the future thesis:

> We were running simple strain-controlled tension/compression
> (\yclass{UniaxialStrainer}) test on a 1:1:2 cuboid-shaped specimen of
> 2000 spheres.\footnote{Later, the test was being done on
> hyperboloid-shaped specimen, to pre-determine fracturing area, while
> avoiding boundary effects.} Straining is applied in the direction of
> the longest dimension, on boundary particles; they are identified, on
> the ``positive'' and ``negative'' end of the specimen, by distance
> from bounding box of the specimen; as result, rougly one layer of
> spheres is considered as support on each side. Distance between (some)
> two spheres on each end along the strained axis determines the
> reference length $l_0$; specimen elongation is computed from their
> current distance divided by $l_0$ during subsequent simulation.
> Straining imposes displacement on support particles along strained
> axis, symetrically on either end of the specimen (half on the
> ``positive'' and half on the ``negative'' boundary particles), while
> all their other degrees of freedom are kept free, including
> perpendicular translations, leading to simulation of frictionless
> supports.
> 
> Axial force $F$ is computed by averaging sums of forces on support
> particles from both supports $F^+$ and $-F^-$. Divided by specimen
> cross-section $A$, average stress is obtained. The cross-section area
> is estimated as either cross-section of the specimen's bounding box
> (for cuboid specimen) or as minimum of several areas $A_i$ of convex
> hull around particles intersecting perpendicular plane at different
> coordinates along the axis (for non-prismatic specimen) -- see fig.~
> \ref{fig-cpm-uniax-specimen}.
> 
> Such tension/compression test can be found in the
> \ysrc{examples/concrete/uniax.py} script

(there referenced figure is attached)

Cheers, vaclav

Attachment: uniax-specimen.pdf
Description: Adobe PDF document