parameter scaling

Asked by azim on 2018-06-10

Dear all,
I have a question about the effect of parameter scaling in Yade-DEM.
I'm going to decrease the time needed for a simulation to be completed!!
 my current (and only) option is to do scaling (density, gravity, mass, length or ....). I have some questions about this topic:
1) what kind of scaling is possible in Yade-DEM?? (all options? any reference??)
2) has anyone done scaling yet? which kind? (I think: yes) I mean is there any paper or document showing results??
3) what are other options to decrease the simulation time with small effect on convergence??

any help is appreciated in advance
Thanks
Azim

Question information

Language:
English Edit question
Status:
Answered
For:
Yade Edit question
Assignee:
No assignee Edit question
Last query:
2018-06-23
Last reply:
2018-06-25
Jérôme Duriez (jduriez) said : #1

Hi,

1) Yade users have access to any possible properties (density, gravity, length, etc....) of their models, so all options are possible.

2) I personaly used density scaling in [Duriez2018] (from https://www.yade-dem.org/doc/publications.html). However, thinking now in terms of inertial number (see the granular mechanics litterature), I do not think anymore density scaling can be of a real help.

Note that this question has already been discussed on this mailing list, see e.g. https://answers.launchpad.net/yade/+question/626656

Hi,
Please see also [1].
There only two options as far as I know:
1/ If you have very heterogeneous mass/stiffness ratio (like typically with broad size distribution or with multiple materials with very different stiffness), you can use density scaling.
2/ in all cases you can decrease the stiffness/externalPressure ratio to gain speedup, physical implications needs to be checked carefuly - there can be some.

Bruno

[1] https://answers.launchpad.net/yade/+question/249117

azim (mirzavand) said : #3

Hi Jerome and Bruno,
Thanks for your reply. sorry for being late. I had a lot of problems last week And couldn't focus on the project.

--------------------------------------------
Dear Jerome,

>> 1) Yade users have access to any possible properties (density, gravity, length, etc....) of their models, so all options are possible.
.
.
.
I read an article about calibration written by Jan Stranskey[1]. it it possible to analyse the effect of scaling in dimensional analysis!!
I need to proof scaling results theoretically. does all the options you mentioned have the ability to be studied with dimensional analysis in([Stransky2010]) [1]?

>> 2) I personaly used density scaling in [Duriez2018] (from https://www.yade-dem.org/doc/publications.html).
However, thinking now in terms of inertial number (see the granular mechanics litterature), I do not think any more density scaling can be of a real help.
.
.
.

 what do you mean when you say: "I do not think anymore density scaling can be of a real help"?
is it disused? or there is a better way? or there will be no real changes in comparison to non-scale condition??

>> Note that this question has already been discussed on this mailing list, see e.g. https://answers.launchpad.net/yade/+question/626656
.
.
.
I read there>> Bruno Chreyre: Quasistatic means that all inertial effects can be neglected.
in one of my models I can ignore inertial forces, but in the other one inertia has a big effect on my results.

[1]https://www.yade-dem.org/doc/publications.html

---------------------------------------------------

Dear Bruno,

>> 1/ If you have very heterogeneous...with broad size distribution....
yes, I have a broad size distribution.

>> 2/ in all cases you can decrease the stiffness/externalPressure ratio to gain speedup, physical implications needs to be checked carefuly - there can be some.
i don't really get it.
(particles stiffness/confining pressure)? in which conditions do i need to decrease it??loading, packing?
and, which physical implications i need to check??

finally, I need to mention that I want to use one kind of scaling options in DEM-PFV that you are so expert in.
let me know(please), if something relevant exists. or there is something I’m ignorant about.

Thanks
Azim

Jérôme Duriez (jduriez) said : #4

With "I do not think anymore density scaling can be of a real help." in #1, I meant that my current opinion is the following :

for strain controlled simulations, artificial increases in density in order to increase time step eventually do not really help decreasing computation cost (in terms of DEM iterations), because the imposed strain rate has to be decreased in such a proportion that it cancels the effect of a greater time step.

azim (mirzavand) said : #5

Hi Jerome,
thanks.
would you please help me about studying scaling with dimensional analysis? (density, gravity, ....)
Azim

@Jérôme
>"artificial increases in density [...] do not really help decreasing computation cost [...] because the imposed strain rate has to be decreased in such a proportion that ....

I do agree that increasing the density uniformly of all particles has a null effect (strictly).
In this discussion I was guessing that "density scaling" was referring to Yade's NewtonIntegrator::densityScaling, which is a _differential_ scaling and this one should definitely help in some cases (there is even an example script about that called densityScaling.py IIRC).

@azim
> would you please help me about studying scaling with dimensional analysis?

Are you asking someone to explain dimensional analysis? Just found [1], which seems relevant.
[1] http://philsci-archive.pitt.edu/4474/

Bruno

azim (mirzavand) said : #7

Hi Bruno,
thanks for densityScaling.py.
> Are you asking someone to explain dimensional analysis?
of course not. I'm familiar with dimensional analysis.(thanks for the link, I will study that anyway)
I specifically mentioned the article [1] to clarify my question.
I'm searching for an idea for doing dimensional in the way i want.
If,for example, I want to check the effect of density(gravity,...) scaling on macroscopic parameters, what new parameters do I need to add to current ones[1].
note that this is what I know about dimensional analysis, if my expressions are irrelevant I will appreciate your professional comments.

[1] Stransky2010
Thanks
Azim

Jan Stránský (honzik) said : #8

Hi Azim,

in [1] the dimensional analysis is used for static elastic constants, i.e. the case where density plays no role.
For dynamic simulations, density and strain rate (or velocity or something like that) would be another quantities to be taken into account (we neglected them since in the limit of zero strain rate, which was our goal, they has no effect).

My opinion (basically repeating Bruno's answer): with whatever scaling you changes the physics of the problem (you model physically different situation). If it is acceptable or not depends strongly on the specific problem and expected results.

cheers
Jan

azim (mirzavand) said : #9

Hi Jan,
Thanks for your reply.
do you know any paper or reference about dimensional analysis for dynamic simulations in discrete element method?
(gravity, density or .....) not difference, i just need the procedure!!
thanks again
Azim

Jan Stránský (honzik) said : #10

> do you know any paper or reference about dimensional analysis for dynamic simulations in discrete element method?

no, sorry..
Jan

azim (mirzavand) said : #11

Hi Jan
thanks for comments.
I think I need to do it by myself.
I stick your hints(and others) in mind.

thanks Jan,Bruno and Jérôme.
Azim

Can you help with this problem?

Provide an answer of your own, or ask azim for more information if necessary.

To post a message you must log in.