Re: [Yade-dev] [Yade-users] [Question #676841]: the definition of porosity

[Jan Stránský]

Hi Bruno,
sorry for late answer, during Christmas I totally forgot about the topic :-)
yes, reading it once more, it sounds contradictory.. but anyway it make
sense to me and I am sure there are words that would explain/define it
better..

apparently the topic would deserve something like FAQ [1] :-)
Jan

[1] https://answers.launchpad.net/yade/+question/678055



út 18. 12. 2018 v 20:07 odesílatel Bruno Chareyre <
bruno.chare...@3sr-grenoble.fr> napsal:

>
>
> Le mar. 18 déc. 2018 16:37, Jan Stránský  a
> écrit :
>
>>
>>
>> see my previous (slightly modified):
>> - assuming the particles rigid
>> - computing the force under assumption that in reality there would be no
>> overlap and some deformation.
>>
>
> This is self contradictory, no? What would be a model where you assume the
> contrary of what you consider reality and expected result?!
> By this statement you are efficiently killing the rigid assumption.
> Obviously it doesn't fit in the general picture and it it is useless.
> Bruno
>
>
___
Mailing list: https://launchpad.net/~yade-dev
Post to : yade-dev@lists.launchpad.net
Unsubscribe : https://launchpad.net/~yade-dev
More help   : https://help.launchpad.net/ListHelp

Re: [Yade-dev] [Yade-users] [Question #676841]: the definition of porosity

[Bruno Chareyre]

Haha, no worries.
Let's see the reaction of jamespaul.
I was about to send him a link to yade doc (DEM background) in order to 
claim that yade was not assuming overlap (so _he_ would have to justify 
the concept).
Unfortunaltely it is not true: although "normal displacement", "change 
of distance", "normal strain" is used in many places (and refers to 
deformability unequivocaly) the word "overlap" still appears here and 
there. Blame me for not cleaning. :)


Bruno

On 1/24/19 2:44 PM, Jan Stránský wrote:

Hi Bruno,
sorry for late answer, during Christmas I totally forgot about the 
topic :-)
yes, reading it once more, it sounds contradictory.. but anyway it 
make sense to me and I am sure there are words that would 
explain/define it better..


apparently the topic would deserve something like FAQ [1] :-)
Jan

[1] https://answers.launchpad.net/yade/+question/678055



út 18. 12. 2018 v 20:07 odesílatel Bruno Chareyre 
> napsal:




Le mar. 18 déc. 2018 16:37, Jan Stránský
mailto:honzik.stran...@gmail.com>> a
écrit :



see my previous (slightly modified):
- assuming the particles rigid
- computing the force under assumption that in reality there
would be no overlap and some deformation.


This is self contradictory, no? What would be a model where you
assume the contrary of what you consider reality and expected result?!
By this statement you are efficiently killing the rigid
assumption. Obviously it doesn't fit in the general picture and it
it is useless.
Bruno


___
Mailing list: https://launchpad.net/~yade-dev
Post to : yade-dev@lists.launchpad.net
Unsubscribe : https://launchpad.net/~yade-dev
More help   : https://help.launchpad.net/ListHelp


___
Mailing list: https://launchpad.net/~yade-dev
Post to : yade-dev@lists.launchpad.net
Unsubscribe : https://launchpad.net/~yade-dev
More help   : https://help.launchpad.net/ListHelp

Re: [Yade-dev] [Yade-users] [Question #676841]: the definition of porosity

[Bruno Chareyre]

On 12/18/18 8:07 PM, Bruno Chareyre wrote:

it doesn't fit in the general picture and it it is useless.


I hope it didn't sound harsh! What I mean by "useless" is, litteraly, 
"of no use" for justifying/deriving the governing equations.

B




___
Mailing list: https://launchpad.net/~yade-dev
Post to : yade-dev@lists.launchpad.net
Unsubscribe : https://launchpad.net/~yade-dev
More help   : https://help.launchpad.net/ListHelp

Re: [Yade-dev] [Yade-users] [Question #676841]: the definition of porosity

[Bruno Chareyre]

Le mar. 18 déc. 2018 16:37, Jan Stránský  a
écrit :

>
>
> see my previous (slightly modified):
> - assuming the particles rigid
> - computing the force under assumption that in reality there would be no
> overlap and some deformation.
>

This is self contradictory, no? What would be a model where you assume the
contrary of what you consider reality and expected result?!
By this statement you are efficiently killing the rigid assumption.
Obviously it doesn't fit in the general picture and it it is useless.
Bruno
___
Mailing list: https://launchpad.net/~yade-dev
Post to : yade-dev@lists.launchpad.net
Unsubscribe : https://launchpad.net/~yade-dev
More help   : https://help.launchpad.net/ListHelp

Re: [Yade-dev] [Yade-users] [Question #676841]: the definition of porosity

[Jan Stránský]

>> I like my point of view very much :-) ...
> No luck then. :(

it was just explanation why I did the answer. In general I agree with both
options, just wanted to "defend" my approach.
I also think we come to philosophy area in many places here :-) especially
because finally one arrives to the same results and formulas from both
sides.

>>> I would suggest that Cundall/Yade DEM makes no assumption of
rigidity/overlaps.
>> but does not make any assumption on non-rigidity at the same time,
right? :-)
> It does. Starting from an elastic force-displacement equation in the
first place really implies non-rigidity.

see my previous (slightly modified):
- assuming the particles rigid
- computing the force under assumption that in reality there would be no
overlap and some deformation.
Then the elastic force-displacement equation just computes force assuming
some deformation in reality, but the model of particle may remain rigid.

>> Relative motion between bodies says nothing about (non)rigidity of the
bodies, or?
> If you have a sticking (elastic) contact it means two material points
from two different objects are co-moving. If at the same time the reference
position-orientation of the inertial frames attached to each particle are
not undergoing rigid body motion the only solution is that there is an
internal deformation somewhere.

Relative motion is just relative motion. You can define it both for rigid
and deformable particles. In both cases it is just a consequence of the
previous paragraph and decision if particles are or are not rigid :-)

> But my claim is that there is not one context in which the geometrical
overlap would be justified.
> Can you evaluate this volume change by the overlaps? NO! Absolutely not.
> Overlapped volume is not an approximation of anything.

Yes, I agree that from physical point of view, the "purely geometrical"
overlaps does not make much sense. In the launchpad question I just
described how I would do it, but with no experience in the topic..
But it would be nice to actually compare this geometrical and physical
approaches :-) maybe "by chance" the geometrical (non-pysical) overlap
justification is not that far from physical evaluation, even for the first
sight there is no reason for it (I have a fresh experience with something
similar).

cheers
Jan
___
Mailing list: https://launchpad.net/~yade-dev
Post to : yade-dev@lists.launchpad.net
Unsubscribe : https://launchpad.net/~yade-dev
More help   : https://help.launchpad.net/ListHelp

Re: [Yade-dev] [Yade-users] [Question #676841]: the definition of porosity

[Bruno Chareyre]

On 12/17/18 4:26 PM, Jan Stránský wrote:

Hi Bruno,

A) I am currently not so active contributor, so I do not insist on my 
point of view


Still the most active, it seems. :)

B) I like my point of view very much :-) and it makes to me much more 
sense and I wanted to present it here



No luck then. :(
We will have to not mix our replies.
Not a big issue, but an interesting philosophical question.


as I understand DEM and always explained it is that:


I agree, your way to put it is the most common. "Best" not implied.
And actually worst from both scientific and didactic points of view IMO.

> I would suggest that Cundall/Yade DEM makes no assumption of 
rigidity/overlaps.


but does not make any assumption on non-rigidity at the same time, 
right? :-)


It does. Starting from an elastic force-displacement equation in the 
first place really implies non-rigidity.




Relative motion between bodies says nothing about (non)rigidity of the 
bodies, or?


If you have a sticking (elastic) contact it means two material points 
from two different objects are co-moving. If at the same time the 
reference position-orientation of the inertial frames attached to each 
particle are not undergoing rigid body motion the only solution is that 
there is an internal deformation somewhere.




> Rejecting the notion of overlap is I think the only way to escape 
classical ill-posed questions on porosity. "Should overlapped volumes 
be removed?"


Such question always needs a close context. Missing context is IMO 
making it most ill-posed.


Yes, this is context-dependent in general. But my claim is that there is 
not one context in which the geometrical overlap would be justified.



- I have a dense assembly by triaxial compression I squeeze it ti 2/3 
of its original volume.. probably in this case it makes sense to 
somehow treat overlap/deformation.


In that case you certainly need to account for volume change of the 
solid phase, yes.

Can you evaluate this volume change by the overlaps? NO! Absolutely not.
You'll need volume change as a component of the model.  Selecting an 
expression randomly (here cumulative overlapped volume) is not a valid 
replacement for a missing model. And actually overlaps amplify the real 
volume change of standard materials by far. It can even lead to porosity 
>1...



Well, you always have some limit on accuracy.
HM tells the volume change for one sphere-sphere contact, but many 
contacts with "big" overlaps would influence each other and you again 
ends with an approximation.


"Approximation" is always defined with respect to an ideal, more 
accurate, value.

2-sphere HM is only an approximation of multi-contact situation, I agree.
Volume change = 0 is then an approximation of the 2-sphere HM when 
normal disp. tends to zero.
Overlapped volume is not an approximation of anything. It's first 
derivative is way off any physically justified quantity.
It sounds better than zero because it is not constant, but it is 
actually not better. The error by using overlap is larger than by 
keeping zero (this is assuming that Poisson coefficient of the solid 
phase is between 0 and 0.5...).


Cheers

B




___
Mailing list: https://launchpad.net/~yade-dev
Post to : yade-dev@lists.launchpad.net
Unsubscribe : https://launchpad.net/~yade-dev
More help   : https://help.launchpad.net/ListHelp


___
Mailing list: https://launchpad.net/~yade-dev
Post to : yade-dev@lists.launchpad.net
Unsubscribe : https://launchpad.net/~yade-dev
More help   : https://help.launchpad.net/ListHelp

Re: [Yade-dev] [Yade-users] [Question #676841]: the definition of porosity

[Jan Stránský]

Hi Bruno,

A) I am currently not so active contributor, so I do not insist on my point
of view
B) I like my point of view very much :-) and it makes to me much more sense
and I wanted to present it here

as I understand DEM and always explained it is that:
1) DEM represents material as a set of perfectly rigid particles whose
motion is governed by Newton's second law.
2) Based on particles' mutual motion, interparticle forces are computed
according to constitutive laws.
3) Just after there would be a picture of overlapping particles emphasizing
the overlapping volume/area and explanation that in reality there would be
no overlap but a repulsive force, which is computed by the constitutive law
somehow according to that overlap
4) Interparticle forces are put into equations of motion and numerically
integrated.
5) final note that DEM is "nothing more" :-) and that very large amount of
variants are formed using different particle shapes, constitutive laws and
some other extensions (bonded particles.)

I think it is dual to the point of view presented by you and they are
nicely interchangable.
However, I think my point of view is more natural and easier to understand
and accept some other assumptions.

The fact that deformations are not plotted nor taken into account in
NewtonIntegrator IMO supports my point of view.

> It would be better if we could speak the same language in answers

100% agreement

> I would suggest that Cundall/Yade DEM makes no assumption of
rigidity/overlaps.

but does not make any assumption on non-rigidity at the same time, right?
:-)

> The notion of overlap is misleading and should be avoided. I usually
speak of normal displacement wrt. equilibrium state, instead.

This applies to spherical shapes. More general shapes can change their
overlap just by rotation.
On the other hand, rigid particles assumption and word overlap express
exactly what one (or at least me :-) would expect.

> In contact models it is admitted that the bodies are not rigid, since
there can be relative motion between bodies in contact.

Relative motion between bodies says nothing about (non)rigidity of the
bodies, or?

> Hertz-Mindlin is a perfect example, it is directly accounting for
internal deformation, and it is derived on the basis that solid surfaces
*cannot* overlap.
> The other models can be seen as linearizations of HM, and along this line
they don't introduce overlaps either.

see point 3) at the beginning. This can be also achieved using the dual
approach: assuming the particles rigid, but computing the force under
assumption that in reality there would be no overlap.

> The fact is that we never display deformed shapes of particles. We could
in some cases (with HM at least), and then the spheres would appear with
surface deflection instead of overlaps. It would be painful to implement
and rendering would be much slower, but virtually it can be done. Hence why
overlap is just a geometrical artifact. It is not needed in the governing
equations, it only appears as a byproduct of graphical display.

Or a dual view: particle overlap is intrinsic feature of the method and
deformation just computational artifact :-) which can easily be done in a
post-processing stage.

> Rejecting the notion of overlap is I think the only way to escape
classical ill-posed questions on porosity. "Should overlapped volumes be
removed?"

Such question always needs a close context. Missing context is IMO making
it most ill-posed.
- I have a loose packing and by some normal compression make it denser. Do
I need to bother with overlaps while computing porosite? Most likely not
- I have a dense assembly by triaxial compression I squeeze it ti 2/3 of
its original volume.. probably in this case it makes sense to somehow treat
overlap/deformation.

> What is the change of volume of a compressed contact then? Well, HM tells
you exactly the volume change as part of the closed form solution.

Well, you always have some limit on accuracy.
HM tells the volume change for one sphere-sphere contact, but many contacts
with "big" overlaps would influence each other and you again ends with an
approximation.
I see three levels of usage (depending on context):
a) you don't take into account overlaps/deformation at all
b) you compute exactly "geometrical" porosity (applicable only on rigid
body assumption) excluding the overlapping volume of rigid particles
c) you compute porosity based on constitutive law (which may be HM based or
may fall to case b) "by default")

> But in any case, the overlapping volume is irrelevant to physics.

Yes. But in "my" point of view, it is an input for the constitutive law :-)

Cheers
Jan
___
Mailing list: https://launchpad.net/~yade-dev
Post to : yade-dev@lists.launchpad.net
Unsubscribe : https://launchpad.net/~yade-dev
More help   : https://help.launchpad.net/ListHelp

Re: [Yade-dev] [Yade-users] [Question #676841]: the definition of porosity

[Jerome Duriez]

As for me, I subscribe to Bruno's presentation: better speak in terms of 
normal component of relative displacement between centers.


Jérôme

--
Chargé de Recherche / Research Associate
Irstea, RECOVER
3275 route Cezanne – CS 40061 13182 Aix-en-Provence Cedex 5 FRANCE
+33 (0)4 42 66 99 21

On 17/12/2018 13:43, Bruno Chareyre wrote:



On 12/17/18 12:14 PM, Jan Stránský wrote:


yes. In reality (if the particles were soft), there would be no 
overlap but the particles would deform (being in contact, but not 
overlapping).
In Yade, the particles are rigid and do overlap. Repulsive forces are 
somehow computed from this overlap.
Porosity is computed using the original particles volume. This way, 
the computation is very easy and fast.


>From existing overlaps, you can relatively easily compute actual volume
of overlaps and adjust the evaluation if needed.


I beg to disagree. :)
It would be better if we could speak the same language in answers 
(that's why I'm moving the discussion to yade-dev), let's see if we 
find a common ground.
I would suggest that Cundall/Yade DEM makes no assumption of 
rigidity/overlaps. The notion of overlap is misleading and should be 
avoided. I usually speak of normal displacement wrt. equilibrium 
state, instead.


The only rigid-body approximation is in Newton, where we take moment 
of inertia constant (it should change with deformation).
Deformation is neglected from an inertial point of view, that is true, 
but it doesn't mean no deformation anywhere.


In contact models it is admitted that the bodies are not rigid, since 
there can be relative motion between bodies in contact.
Hertz-Mindlin is a perfect example, it is directly accounting for 
internal deformation, and it is derived on the basis that solid 
surfaces  *cannot* overlap.
The other models can be seen as linearizations of HM, and along this 
line they don't introduce overlaps either.


The fact is that we never display deformed shapes of particles. We 
could in some cases (with HM at least), and then the spheres would 
appear with surface deflection instead of overlaps. It would be 
painful to implement and rendering would be much slower, but virtually 
it can be done. Hence why overlap is just a geometrical artifact. It 
is not needed in the governing equations, it only appears as a 
byproduct of graphical display.


Rejecting the notion of overlap is I think the only way to escape 
classical ill-posed questions on porosity. "Should overlapped volumes 
be removed?"
If we agree that there is no overlap there is of course no reason to 
compute overlapping volume. What is the change of volume of a 
compressed contact then? Well, HM tells you exactly the volume change 
as part of the closed form solution. If someone is using a linearized 
form, defining accurate volume change is less clear, it may have to be 
defined as part of the contact model itself. But in any case, the 
overlapping volume is irrelevant to physics.


Bruno




Also please, if the question is about porosity, next time provide the
code you use to compute porosity. There are two of them, [1] using the
computation you described, the other [2] using voxel approximation (but
for your case, computing actual overlaps is not that difficult and much
more precise).

Jan

[1] https://yade-dem.org/doc/yade.utils.html#yade._utils.porosity
[2] https://yade-dem.org/doc/yade.utils.html#yade._utils.voxelPorosity





___
Mailing list: https://launchpad.net/~yade-dev
Post to : yade-dev@lists.launchpad.net
Unsubscribe : https://launchpad.net/~yade-dev
More help   : https://help.launchpad.net/ListHelp



___
Mailing list: https://launchpad.net/~yade-dev
Post to : yade-dev@lists.launchpad.net
Unsubscribe : https://launchpad.net/~yade-dev
More help   : https://help.launchpad.net/ListHelp

Re: [Yade-dev] [Yade-users] [Question #676841]: the definition of porosity

[Bruno Chareyre]

On 12/17/18 12:14 PM, Jan Stránský wrote:


yes. In reality (if the particles were soft), there would be no overlap but the 
particles would deform (being in contact, but not overlapping).
In Yade, the particles are rigid and do overlap. Repulsive forces are somehow 
computed from this overlap.
Porosity is computed using the original particles volume. This way, the 
computation is very easy and fast.

>From existing overlaps, you can relatively easily compute actual volume
of overlaps and adjust the evaluation if needed.


I beg to disagree. :)
It would be better if we could speak the same language in answers 
(that's why I'm moving the discussion to yade-dev), let's see if we find 
a common ground.
I would suggest that Cundall/Yade DEM makes no assumption of 
rigidity/overlaps. The notion of overlap is misleading and should be 
avoided. I usually speak of normal displacement wrt. equilibrium state, 
instead.


The only rigid-body approximation is in Newton, where we take moment of 
inertia constant (it should change with deformation).
Deformation is neglected from an inertial point of view, that is true, 
but it doesn't mean no deformation anywhere.


In contact models it is admitted that the bodies are not rigid, since 
there can be relative motion between bodies in contact.
Hertz-Mindlin is a perfect example, it is directly accounting for 
internal deformation, and it is derived on the basis that solid 
surfaces  *cannot* overlap.
The other models can be seen as linearizations of HM, and along this 
line they don't introduce overlaps either.


The fact is that we never display deformed shapes of particles. We could 
in some cases (with HM at least), and then the spheres would appear with 
surface deflection instead of overlaps. It would be painful to implement 
and rendering would be much slower, but virtually it can be done. Hence 
why overlap is just a geometrical artifact. It is not needed in the 
governing equations, it only appears as a byproduct of graphical display.


Rejecting the notion of overlap is I think the only way to escape 
classical ill-posed questions on porosity. "Should overlapped volumes be 
removed?"
If we agree that there is no overlap there is of course no reason to 
compute overlapping volume. What is the change of volume of a compressed 
contact then? Well, HM tells you exactly the volume change as part of 
the closed form solution. If someone is using a linearized form, 
defining accurate volume change is less clear, it may have to be defined 
as part of the contact model itself. But in any case, the overlapping 
volume is irrelevant to physics.


Bruno




Also please, if the question is about porosity, next time provide the
code you use to compute porosity. There are two of them, [1] using the
computation you described, the other [2] using voxel approximation (but
for your case, computing actual overlaps is not that difficult and much
more precise).

Jan

[1] https://yade-dem.org/doc/yade.utils.html#yade._utils.porosity
[2] https://yade-dem.org/doc/yade.utils.html#yade._utils.voxelPorosity





___
Mailing list: https://launchpad.net/~yade-dev
Post to : yade-dev@lists.launchpad.net
Unsubscribe : https://launchpad.net/~yade-dev
More help   : https://help.launchpad.net/ListHelp