Stress/strain evaluation in the uniaxial tension test

Also Jan, what is the most suitable type of material for such type of simulation? (e.g. u mentioned about CohFrictMat). So if it's not CmpMat which i use now, please also say what type of contact law and Interaction Physics creation i should use with this new material.

with regards Alexander

The last question:) when i try to change pysson coefficient the value of displacement along X doesn't change, it remains the same [-0.00152,-0.00152]. why is it so?

with regards Alexander

oh, i posted a wrong picture here

> (made a slice parallel XOY - http://i11.pixs.ru/storage/0/9/1/pic2JPG_2016907_18081091.jpg)

the right ones are: without slice along OXY - http://i11.pixs.ru/storage/8/4/7/pic1JPG_3485627_18082847.jpg with slice - http://i11.pixs.ru/storage/8/5/6/pic2JPG_8047730_18082856.jpg. The problem is that for coarse packing the result is more exact - http://i10.pixs.ru/storage/8/6/4/pic3JPG_6827924_18082864.jpg, i mean the range is closer to 1,5e8. So i think it's because i use young modulus calibrated for finest packing, the displacements for coarse case is also different http://i11.pixs.ru/storage/8/8/4/pic4JPG_4292200_18082884.jpg and they are worse [-0.0063,0.0063] instead of [-0.0059, 0.0059], the ideal range is [-0.006, 0.006].

i also used simple conversion rule for stress values:
s = bodyStressTensors()
stresses = [Matrix3(*[(4.*pi/3.)*s[b.id][i,j]*b.shape.radius**3/(8*b.shape.radius**3) for j in range(3) for i in range(3)]) for b in O.bodies]

So am i right in my discourse?

with regards Alexander

> I mean in generally, if it will be a little bit better with another settings then i don't see any point to check them.

Here i think i'm not correcct:) I would be glad for any advise, especially for example CpmMat has a lot of parameters (sigmaT, sigmaT .ets).

And So if it is the best result that i can achieve, the i will go to the second step of my research, considering more complicated model with 2 different materials. Also it will be cube, not a plate.

 with regards Alexander

Hi Alexander,

As i understand the best and most scientific way is the following:
> 1) for boundary particles without prescribed force use bodyStressTensor,
> but the values will be different in compare with inner particles because
> of boundary particles has same interactions in parallel direction, but
> only half in perpendicular and
> diagonal direction (by the way i plotted interactions -
> http://i11.pixs.ru/storage/0/9/8/pic3JPG_4835644_18081098.jpg :). Also i
> should tell in the final presentation of my research that in limit this
> anomally is a set of measure zero.
> 2) for boundary spheres with prescribed force compute stress manually,
> because of they have only half interactions in parallel and diagonal
> direction. So i gotta take into account prescribe force on them, because of
> force F doesn't belong to any interaction and that's why bodyStressTensor
> doesn't see it.
>
>
well, what is "most scientific" is more a philosophical question, and what
you should or should not is mostly up to you ;-) otherwise I agree with my
own conclusions ;-)

The stress values for inner particles must be equal prescribed pressure
> (1,5e8 in my case) or not?

Because i tried to check Y component of normal stress for inner particles
> and it's not 1,5e8 . For simplicity i set exact stress value 1,5e8 to all
> boundary spheres with prescribe force manually, so as u can see in the
> picture stress inside the model is lower (made a slice parallel XOY -
> http://i11.pixs.ru/storage/0/9/1/pic2JPG_2016907_18081091.jpg), at the
> same time displacements are ok (
> http://i11.pixs.ru/storage/0/8/4/pic1JPG_4520635_18081084.jpg). May be
> for inner particles stress become exact in the infinite case as well as for
> boundary particles without prescribe force?

in the limit case it should converge to the prescribed pressure, but I
would not expect the value to be exactly the same.. The reason is the
boundary :-)

Also Jan, what is the most suitable type of material for such type of
> simulation? (e.g. u mentioned about CohFrictMat). So if it's not CmpMat
> which i use now, please also say what type of contact law and
> Interaction Physics creation i should use with this new material.

if you want to simulate elastic behavior, the specific material does not
matter. I know CpmMat and it is suitable for this simulation. I don't know
other materials very well, but probably there are some, ehish are as
suitable as CpmMat (because of the elastic behavior)

4) For displacements and stresses described above were computes with
> young=1.338e11, but the real value for steel is 2e11. Should i also tell
> in my final presentation, that young and pysson in DEM are just some
> approximation of real physics parameters which is used in FEM.

again, what you should is up to you :-) young parameter of CpmMat is
Young's modulus of one interaction, not directly related to the overall
Young's modulus of final material, that's it. I would concentrate more on
the macroscopic properties, saying the model parameters were calibrated to
fit this and that values. People with DEM background would understand,
others may be surprised that you need calibration for elastic constant..

>
> >For elastic material, you want to fit 2 macroscopic constants, e.g.
> Young's
> >modulus and Poissons's ratio. The main conclusion from the paper:
> >- The Young's modulus is directly proportional to 'young' CpmMat
> >parameters, so it can be fit trivially.
> >- Poisson's ratio does depend on 'poisson' CpmMat parameter, but
> >nonlinearly, so you can run simulations with different values to see the
> >relationship. As it is 1 variable, the optimization is not so difficult.
> Considering your comments is my process of calibration correct?
> - i started calibration from Young's modulus because for the value
> young=2e11 the model stretched lesser than it should be, i mean the extreme
> values of Y component of displacement vector was lower.
> - After young=1.338e11 value was found, i checked extreme values of X
> component of displacement vector, so for my model it was
> [-0.00152,-0.00152], in Ansys it was [-0,0018,0,0018]. So i was satisfied
> with this, that's why left pysson=0,3 like for the real steel.

exactly opposite :-) if you change 'poisson', both overall Young's modulus
and Poisson's ratio would change. If you change 'young', only overall
Young's modulus would change, not influenceing Poisson's ratio. So better
way is first fit 'poisson' to fit macroscipic Poisson's ratio and then
'young' to fit Young's modulus.

i also used simple conversion rule for stress values:
> s = bodyStressTensors()
> stresses = [Matrix3(*[(4.*pi/3.)*s[b.id][i,j]*b.shape.radius**3/(8*b.shape.radius**3)
> for j in range(3) for i in range(3)]) for b in O.bodies]

just a bit more pythonic (multiplying directly stress):
stresses = [4.*pi/3./8.*stress for stress in s]
but yes, it is the correct approach for regular packing

cheers
Jan

2015-07-19 20:56 GMT+02:00 Alexander <email address hidden>:

> Question #269393 on Yade changed:
> https://answers.launchpad.net/yade/+question/269393
>
> Alexander gave more information on the question:
> > I mean in generally, if it will be a little bit better with another
> settings then i don't see any point to check them.
>
> Here i think i'm not correcct:) I would be glad for any advise,
> especially for example CpmMat has a lot of parameters (sigmaT, sigmaT
> .ets).
>
> And So if it is the best result that i can achieve, the i will go to the
> second step of my research, considering more complicated model with 2
> different materials. Also it will be cube, not a plate.
>
> with regards Alexander
>
> --
> You received this question notification because you are a member of
> yade-users, which is an answer contact for Yade.
>
> _______________________________________________
> Mailing list: https://launchpad.net/~yade-users
> Post to : <email address hidden>
> Unsubscribe : https://launchpad.net/~yade-users
> More help : https://help.launchpad.net/ListHelp
>

5 posts for one question, wow! This is becoming confused again. Please Alexander, keep things simple and concise, be specific, and think twice to avoid avalanches of emails on us.

By "be specific", I mean this for instance:
> The stress values for inner particles must be equal prescribed pressure (1,5e8 in my case) or not?
> Because i tried to check Y component of normal stress for inner particles and it's not 1,5e8 .

How far is it? Is the difference balanced by boundary spheres? etc.
My statement was that the average stress (in a row or in the full problem) is equal to the applied stress. I did not say anything about inner/outer IIRC.

Hello Jan and Bruno.

I'm so sorry for my "avalanches" of emails on you, it's because i want to solve my problem very much and your program is my only hope:) I also understand that you spend time on me, now i'll try be more specific.

Thank's a lot Jan, for such a detail answer again:) I will follow your advise when i simulate my final model.

Now Bruno, I'm sorry about reading your messages carelessly:

> My statement was that the average stress (in a row or in the full problem) is equal to the applied stress. I did not say anything > > about inner/outer IIRC.

So i get the average stress equal to applied stress for the whole model But i get this result only after half of prescribed stress was added to the stress value for all boundary spheres with prescribed force.

Do you mean this operation in this question?
>Is the difference balanced by boundary spheres?

with regards Alexander

>So i get the average stress equal to applied stress for the whole model But i get this result only after half of prescribed stress was added to the stress value for all boundary spheres with prescribed force.

I think those boundary spheres should be considered half-spheres, since the applied force are applied at the center of mass. You could reflect that in the volume conversion. The length of the plate should be defined from center to center to. The parts of spheres beyhond this are just pictures that you can draw but they play no mechanical role.

Ok i will check all possible variants after that may be create new question. Now i'm busy with modeling more complex specimen for my simulation. And it will be cube, not a plate.

Thanks a lot for advice

with regards, Alexander