← Back to team overview

yade-users team mailing list archive

Re: Coupling Yade and Comsol


Ok, I received this e-mail.
I double-checked but couldn't find your previous one.
Regarding a possible collaboration, yes, it would be great. It seems that your experience could help us a lot, and I hope we can help you too. If you can come in grenoble to discuss that, its perfect. Prof. Frederic Donze would participate in the discussions probably, as he is the one who initiated the development of Yade (and he still supervizes it). He does not participate directly to the project on erosion however.

I can see what you are trying to do. The upscaling from Stokes
to Darcy was the topic of my D.E.A., and of part of my Ph.D.
thesis. Your problem seems, however, a bit more complicated
than just that, as the internal erosion might create
wormhole-like fluidized-bed regions where Darcy's law is not
necessarily valid anymore. Have you already developed a
conceptual macroscopic model for this kind of flow?

1/ Assume that the packing is always "dense", so that
2/ the velocity of grains is always low compared to the velocity of the fluid, and 3/ the size of pores is of the same order of magnitude as the size of grains, making low-Reynolds assumption valid. 3/ You can triangulate the packing, and define the geometry for the flow as a series of adjacent "voids" (where each void is a tetrahedron cut by four spheres). 4/ Here comes comsol and the most complex part : let us compute the Stokes flow and bring out the hydraulic properties of a generic void. Given the size of the terahedron, the size and velocity of the four spheres, and the averaged pressures in the four adjacent voids, we need approximated values of the flow rates (a sort of local Darcy law) and hydraulic forces on the four grains. 5/ Mass balance between adjacent voids and the transmitivity equations from 4) will give a big system, which (i hope) will have one solution... 6/ Take this solution, compute for one grain the sum of hydraulic forces from all tetrahedral voids which have this grain as vertex. Do the same for all grains, and go back to the classical DEM scheme taking into account this set of hydraulic forces.

I agree that wormhole formation will be out of the scope of such a model. The concept of small voids between grains is not relevant in this case, neither is the low-Reynolds condition. This model will apply for bulk erosion, clogging effects,... but it will not handle the cases where there is a localization of the erosion process and formation of big holes. But isn't that exactly what you are working on now?

(1) a packing of (rigid) sand grains (of spherical shape) is
hold together by friction and confining stress;
(2) a jet of water is directed against the internal wall of
this borehole to apply normal and tangential stresses to the
individual grains in the packing;
(3) hydrodynamic forces and torques acting on the individual
grains are then passed to a DEM code to verify the breaking of
the grain bonds, remove the grains with a broken link,
recalculate the equilibrium position for the packing,

Some physical issues comes to mind :
1/ the internal wall of the borehole will not be stable if you simulate a dry sand with only friction at contact, even before you apply hydraulic forces (except if you have strong arching effects, but I don't think you will have them). You will need a little bit of cohesion at contact. In a real situation, the stability of a borehole in sand can be due to a little bit of clay or to capillary forces if the sand is not saturated (capillary forces are now implemented in Yade). 2/ You can't consider a one-step procedure for removing the grains. Frictional contacts slip, they don't break. Even if the hydraulic forces cause slip at some contacts when you apply them, you can't be sure that the packing will not stabilize after some small movements. It would sound more reasonable to me to keep the hydraulic forces constant - which is correct as long as there is only small displacements of the grains - for a number of DEM iterations (100 iterations should be enough). Then see which grains are removed from the packing. Perhaps it is what you want to do already...

(4) pass the new geometric configuration to the fluid flow code
to recalculate stresses -> goto (2)

I have modeled the hydrodynamic stresses by means of COMSOL
Multiphysics (see attached), and I now need to pass this information to a DEM

Any help on how to start the coupling with yade will be highly

Look at the class ForceEngine. It takes a series of grains and add external forces on them. You could create a class HydroForceEngine similar to this one. HydroForceEngine would read a file containing grain id's and hydraulic forces on them, then it would add force(i) on grain (i), when ForceEngine add the same force on all grains.

Ciao :)


Chareyre Bruno
Maître de conférence

Institut National Polytechnique de Grenoble
Laboratoire 3S (Soils Solids Structures) - bureau I08
BP 53 - 38041, Grenoble cedex 9 - France Tél :