yade-users team mailing list archive

[Alexander <question269063@xxxxxxxxxxxxxxxxxxxxx>]

Question #269063 on Yade changed:
https://answers.launchpad.net/yade/+question/269063

    Status: Answered => Open

Alexander is still having a problem:
Hello Jan and Bruno, thank you for your answers?

Now my problems:

1) By your advise I increase the dense of packing (still left it
regular), but after this displacement of spheres is also changed.

Dislplacements along Y axis gotta be around [-0.006, 0.006]

Pay attention! In my simulation i applied force F like pressure.

Test 2 variants, applying force F directly to spheres and multiplying by sphere radius to take into account spheres size.
For each variant different value of young modulus was used and it's calibrated for coarse variant of packing


First variant: coarse - http://i11.pixs.ru/storage/7/5/8/pic1JPG_5947378_18056758.jpg fine - http://i11.pixs.ru/storage/7/6/8/pic2JPG_7134594_18056768.jpg

Second variant: coarse -
http://i11.pixs.ru/storage/7/7/4/pic3JPG_5739457_18056774.jpg fine -
http://i11.pixs.ru/storage/7/7/7/pic4JPG_2331364_18056777.jpg

You can see that desired displacement changed after both variants.

So is it correct to apply force (pressure) to sphere without taking into
account it's size, or i need to multiply by it's radius, diameter or
volume? And what should i do with changing displacement after different
dense of packing was applied?

Well now let's discuss stress values distribution

In this picture i calibrated young to get appropriate displacement for
the fine packing
http://i11.pixs.ru/storage/8/3/6/pic5JPG_4896692_18056836.jpg

The values range of stress is here: http://i11.pixs.ru/storage/8/5/3/pic6JPG_4993977_18056853.jpg
If conversion macroStress = sphereStress*sphereVolume/cubeVolume will be used we get 
something like 3,2321e8 / 2 = 1,69e8 which is similar to 1,5e8.  (like exact variant in Ansys)

As you can see in the following pictures such maximum values have
spheres inside the packing surrounded by one boundary layer of spheres.

whole model -
http://i11.pixs.ru/storage/9/2/3/pic8JPG_2697509_18056923.jpg

slice along XOZ -
http://i11.pixs.ru/storage/9/2/5/pic9JPG_7187525_18056925.jpg

slice along XOY
http://i11.pixs.ru/storage/9/3/0/pic10JPG_8961910_18056930.jpg

So may be i should just create some additional boundary layer of spheres
sorrounded the model and then delete it? Because your advise Jan:

>- you can also put lower force value for conrner particles, as the force is
>transmitted by this nonstandard parallel boundary layer. Do not forget to
>redistribute this jump to the "middle" forces to preserve total force value
>:-) You can again try different "jump" values to see the response, and
>choose the correct "jump" value. The same approach you would choose, if you
>directly apply nodal forces on FEM mesh and you want to get constant
>stress/strain..

couldn't help with spheres where force is applied, they have fixed
applied force on them. (in the pictures they are always blue)

Here yours comment about that:
>- Stress is computed from interactions. The fixed boundary and boundary
>with prescribed force has half parallel interactions, so (very rouhly) you
>get half stress. See [1], you can also compute it analytically by hand :-)

Also Bruno about your formula macroStress =
sphereStress*sphereVolume/cubeVolume. How it will work for non regular
packing? Because in my final model there will be non regular packing
areas inside the plate.

> Does the analytical match ANSYS (certainly not since a sphere array is not isotropic)?
Yes Ansys solution is exact for uniaxial tension

>P2- Does DEM match the analytical expressions? If not there can be a mistake in the definition of boundary condition, a non->static state, or something like that.
What do you mean here, the result with etalon stress and dicplacement in Ansys doesnt match. Honestly i'm still nor ok with DEM theory:)

Do you know some short and easily understood literature for it. For example at the beginning of Jan's article strain and the contact spheres by springs are described quite well, but there's now stress there. Also what type of contact used in Yade? I think the same like in Jan's article. May be you have something else Jan, especially for my type of simulation?
 
Also Jan i will read about https://yade-dem.org/doc/yade.utils.html#yade._utils.bodyStressTensors.

>Two points here:
>- apart from stress, plot also contact forces. You will see, that at the
>boundary (parallel to force) layer, the forces are different. It is because

How to plot contact forces and how much vectors of force will have each
sphere?

> The real question is the number of nodes/gauss points in each element times the number of elements,
You are right that's why i tried more dense packing, but the results is not ok yet

>Ok for diagonal interactions, I didn't notice. It is indeed more isotropic, but on another hand isn't it what explains the special >trend of stress on free boundaries? This boundary effect is still a problem in my eyes (half stress where force is applied is ok, >it is a post processing artifact).
Here i think only Jan can help, he is much more professional than me:) 

>For the moment it is not clear if you struggle with P1, P2, or both of them. It makes the problem solving difficult.
Of course i have struggle with  P2, FEM is fine:) For uniaxial tension which i test now, Ansys  has perfect results (you can see it in the previous comments on pictures)

>So if you are happy with middle stresses and strains,you are on the right way :-)
Jan, I'm not happy yet with middle stresses because they are bigger:) And by the way what do you think about Bruno's proposal to devide  macroStress = sphereStress*sphereVolume/cubeVolume and what should i do in the non-regular packing?
As said in the future there will be non-regular areas inside.

I'm still waiting for help:)

with regards, Alexander

-- 
You received this question notification because you are a member of
yade-users, which is an answer contact for Yade.