yade-dev team mailing list archive

[François <francois.kneib@xxxxxxxxx>]

2013/12/17 Bruno Chareyre <bruno.chareyre@xxxxxxxxxxx>

>  On 17/12/13 10:20, François wrote:
>
>  Ok I'm going to do the second option.
>  I talked with Raphael and I remember that his work was mostly focused on
> the timestepper. But what I can see in the timestepper now is that he
> didn't prepared the case of coexistent viscous and "pure" elast contacts in
> the same simulation. So if the user sets the flag "viscEl=1" in the
> timestepper, the code will do :
>
> *ViscElPhys* viscPhys = YADE_CAST<ViscElPhys*>(contact->phys.get());*
>
>  for each contact, which is a problem for all non viscous contact -> segt
> fault.
>
>
> If they all inherit from elastMat this problem will disappear, no?
>

No. In fact ViscElPhys already inherits from FrictPhys. We can cast a
ViscElPhys into a FrictPhys, but not the opposite so I must change
something here.

>
>
>
>  About K and Young, my precise idea is :
>
>    - Ip2_FrictMat_FrictMat_FrictPhys : will use young and poisson
>    directly, or use kn and ks if they exist and are set by converting them
>    with : Young = 0.5 * Kn * R and Poisson = Ks/Kn. If both modulus and
>    stiffnesses are defined for one material, print a message that gently
>    recommend the user to review his script.
>     - Ip2_ViscElMat_ViscElMat_ViscElPhys : will use Kn and Ks directly if
>    they are set, convert and use Young and Poisson otherwise. If both modulus
>    and stiffnesses are defined for one material, print a message that gently
>    recommend the user to review his script.
>
>   Excellent.
>
>
>   About your warning, should I take care about something more ? I think
> that kwargs are automatically handled  between python/c++ by boost.
>
> I'm not sure how you will do it, and not sure about possible difficulties.
> If you use "add_property" to handle the different sort of assignments, then
> I'm not sure it will work with kwargs. There is also a subtle difference in
> what really happens:
>
> "mat.young =" ... is simple assignement.
>  "ViscElMat(.....,young=...)" is triggering postLoad().
>

My idea is to use the Ip2 to choose between K or Young and compute the
stiffness of contact. So there is no need to add some new attributes, but
sometimes (in Ip2_FrictMat_FrictMat_FrictPhys) I will have to check whether
the FrictMats are ViscElMat too or not. If they are, I will use its Kn and
Ks if they are defined (note that the default value of Kn and Ks is "NaN",
so it's easy to know if the user have set this attribute).
By doing this, the save/load operation is already ok as the stiffnesses of
the contact is saved/loaded.

>
> Last possible difficulty, the kn/Young decision must be preserved if you
> save and reload.
>
> I'm not saying it is difficult. Just a few things to keep in mind.
>
> Bruno
>
>
>
>
>
> 2013/12/16 Bruno Chareyre <bruno.chareyre@xxxxxxxxxxx>
>
>>  My vote for 2°).
>> The independancy of the "Visc" classes makes no sense. It is inherited
>> from very early ages of the code.
>>
>> Raphael partly fixed the problem regarding kn vs. Young. You should talk
>> to him (I don't remember now why he did not make the inheritance Frict ->
>> ViscEl - to minimize the noise most probably).
>> There is now a choice to decide if kn should be translated like Young or
>> not (IIRC).
>> Your suggestion is to make it more transparent, the flag being set
>> depending on which attribute is defined. I like the idea.
>>
>> Be carefull it must work with different syntax:
>> mat.young = ...
>> m=ViscElMat(.....,young=...)
>>
>> B
>>
>>
>> On 16/12/13 08:56, François wrote:
>>
>>   Hi everybody,
>>
>> I would like in my simulations to make possible the interaction between
>> (Coh)FrictMat and ViscElMat spheres. I don't want to activate the cohesion
>> so I would let CohFrictMat become FrictMat by inheritance. The idea is to
>> study the granular flow of viscous grains with one of the boundary made of
>> cohesive spheres who sticks together.
>>
>> I think there are two solutions : the first one seems easier because it
>> just adds a new class, but the second one (which modifies a little the
>> viscous classes logic) may deserves our consideration. The choice between
>> one of them is the subject of my question.
>>
>> 1°) Create a new class : Ip2_FrictMat_ViscElMat_FrictPhys(). This functor
>> would make an interaction physic just like
>> Ip2_FrictMat_FrictMat_FrictPhys(), ignoring the viscosity and converting
>> the young modulus of the frictional sphere to a stiffness.
>>
>> 2°) ViscElMat, unlike its name lets believe, doesn't inherit from
>> ElastMat. The main reason I think is because a ViscElMat has stiffnesses
>> {kn,ks} while an ElastMat has a young modulus and a poisson ratio
>> {young,poisson}. FrictMat inherits from ElastMat, and the only additional
>> attribute it has is the friction coefficient. As ViscElMat has a friction
>> coefficient too, don't you think we could make ViscElMat a child of
>> FrictMat ? To avoid breaking people's scripts, we can imagine that if the
>> user sets {kn,ks}, those parameters will override {young,poisson} in the
>> calculation of resulting contact stiffnesses. With this inheritance, and if
>> the user sets at least {young,poisson} for both his FrictMat and ViscElMat,
>> the contact will be possible between :
>> - FrictMat + FrictMat => FrictPhys
>> - ViscElMat + ViscElMat => ViscElPhys
>> - FrictMat + ViscElMat(FrictMat by inheritance) => FrictPhys
>>
>> Can people working with/on visco-elasticity give their opinion ? (Anton,
>> Raphaël ...)
>>
>> Thanks,
>>
>> François
>>
>>
>>  _______________________________________________
>> Mailing list: https://launchpad.net/~yade-dev
>> Post to     : yade-dev@xxxxxxxxxxxxxxxxxxx
>> Unsubscribe : https://launchpad.net/~yade-dev
>> More help   : https://help.launchpad.net/ListHelp
>>
>>
>>
>> --
>> _______________
>> Bruno Chareyre
>> Associate Professor
>> ENSE³ - Grenoble INP
>> Lab. 3SR
>> BP 53
>> 38041 Grenoble cedex 9
>> Tél : +33 4 56 52 86 21
>> Fax : +33 4 76 82 70 43
>> ________________
>>
>>
>> _______________________________________________
>> Mailing list: https://launchpad.net/~yade-dev
>> Post to     : yade-dev@xxxxxxxxxxxxxxxxxxx
>> Unsubscribe : https://launchpad.net/~yade-dev
>> More help   : https://help.launchpad.net/ListHelp
>>
>>
>
>
> _______________________________________________
> Mailing list: https://launchpad.net/~yade-dev
> Post to     : yade-dev@xxxxxxxxxxxxxxxxxxx
> Unsubscribe : https://launchpad.net/~yade-dev
> More help   : https://help.launchpad.net/ListHelp
>
>
>
> --
> _______________
> Bruno Chareyre
> Associate Professor
> ENSE³ - Grenoble INP
> Lab. 3SR
> BP 53
> 38041 Grenoble cedex 9
> Tél : +33 4 56 52 86 21
> Fax : +33 4 76 82 70 43
> ________________
>
>
> _______________________________________________
> Mailing list: https://launchpad.net/~yade-dev
> Post to     : yade-dev@xxxxxxxxxxxxxxxxxxx
> Unsubscribe : https://launchpad.net/~yade-dev
> More help   : https://help.launchpad.net/ListHelp
>
>