convergence very very slow

Asked by wangxiaoliang on 2012-03-27

Hi all

In example , the box is usually 10*10*10 or 1*1*1, and particles size is usually a bit larger than the size of real sand.

However, when I tried to model grenoble sand (d50 = 0.33mm), I use two contact model

1, hertz model

2, Moment material model

in both model, after the isotropic compression, in the deviatoric compression process, usually the strain response is

very very small. 1000 step just make the strain in y direction go forward about 1e-6 or 1e-7.

So it needs maybe month or year to achieve a final strain in y of 0.3, say.

In my opinion, a high rigid modulus (usually quartz GPa), induce a high P wave speed, and a smaller size makes dt = r/ vp even

much more smaller than the usual sample in the tutorial(usually mod. = 1e5~6Pa, and particle size .5m)

So I do not know how to deal with this ques.

Question information

Language:
English Edit question
Status:
Answered
For:
Yade Edit question
Assignee:
No assignee Edit question
Last query:
2012-03-28
Last reply:
2012-03-28

Hi,

I hope to interpret correctly what you are saying. Your time step is too small and thus your test runs very slowly. It is a well known issue and you can workaround it by increasing the mass of your particles, for instance.

HTH,
Chiara

On 27 Mar 2012, at 03:55, wangxiaoliang wrote:

> New question #191800 on Yade:
> https://answers.launchpad.net/yade/+question/191800
>
> Hi all
>
> In example , the box is usually 10*10*10 or 1*1*1, and particles size is usually a bit larger than the size of real sand.
>
> However, when I tried to model grenoble sand (d50 = 0.33mm), I use two contact model
>
> 1, hertz model
>
> 2, Moment material model
>
> in both model, after the isotropic compression, in the deviatoric compression process, usually the strain response is
>
> very very small. 1000 step just make the strain in y direction go forward about 1e-6 or 1e-7.
>
> So it needs maybe month or year to achieve a final strain in y of 0.3, say.
>
> In my opinion, a high rigid modulus (usually quartz GPa), induce a high P wave speed, and a smaller size makes dt = r/ vp even
>
> much more smaller than the usual sample in the tutorial(usually mod. = 1e5~6Pa, and particle size .5m)
>
> So I do not know how to deal with this ques.
>
> --
> 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

Jérôme Duriez (jduriez) said : #2

Hello,

On 27 Mar 2012, at 03:55, wangxiaoliang wrote:

>
> 1000 step just make the strain in y direction go forward about 1e-6 or 1e-7.
>

I think you can also change (increase) the strain rate :https://www.yade-dem.org/doc/yade.wrapper.html#yade.wrapper.TriaxialTest.strainRate !

>a high rigid modulus (usually quartz GPa), induce a high P wave speed, and a smaller size makes dt = r/ vp even much more smaller than the usual sample in the tutorial

This statement is correct.
The thing that puzzles me in your question is you don't mention the number of particles, which is for sure the first parameter to consider.

>So I do not know how to deal with this ques.

Use a smaller number of particles?

The number of particles aside, changing mass and/or strain rate does not help in general (unless strain rate is excessively small initially, of course), because higher mass needs slower strain if you don't want to get inertial effects in your results (similarly, faster strain needs smaller mass).
The only way to speed things up is to decrease stiffness.

"Then how can I simulate 10e10 particles with stiffness 10e12", you may ask. You can't, or you have to be very patient.

Also note that GlobalStiffnessTimeStepper will help you to find the largest possible timestep, while PWave only gives a rough estimate.

On 27 Mar 2012, at 12:05, Chareyre wrote:

> Question #191800 on Yade changed:
> https://answers.launchpad.net/yade/+question/191800
>
> Chareyre proposed the following answer:
>> a high rigid modulus (usually quartz GPa), induce a high P wave speed,
> and a smaller size makes dt = r/ vp even much more smaller than the
> usual sample in the tutorial
>
> This statement is correct.
> The thing that puzzles me in your question is you don't mention the number of particles, which is for sure the first parameter to consider.
>
>> So I do not know how to deal with this ques.
>
> Use a smaller number of particles?
I agree that this could be one possible solution.
>
> The number of particles aside, changing mass and/or strain rate does not help in general (unless strain rate is excessively small initially, of course), because higher mass needs slower strain if you don't want to get inertial effects in your results (similarly, faster strain needs smaller mass).
> The only way to speed things up is to decrease stiffness.
I have to disagree with this last statement. Decreasing the inter-particle stiffness (the Young's modulus, in case of Hertzian contacts) won't solve your problem because your results are likely going to be affected by the reduced stiffness. In other words, you would be simulating a different material. This is the case for shear type tests which I suppose are what you are after? In other scenarios, I reckon that the stiffness is less important (like in flow kinematics).

Chiara

>
> "Then how can I simulate 10e10 particles with stiffness 10e12", you may
> ask. You can't, or you have to be very patient.
>
> Also note that GlobalStiffnessTimeStepper will help you to find the
> largest possible timestep, while PWave only gives a rough estimate.
>
> --
> 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

Of course, stiffness is a physical parameter and changing it may affect (or not...) the results. Ultimately the number of particle is the only parameter you can play with.

However, if wangxiaoliang is targeting a final strain of 0.3, he is probably not focusing on the elastic regime.
If the confining pressure is small compared to the stiffness of silicates (10e11), stiffness can most likely be decreased by a few orders of magnitude without any change in peak strength, residual strength, dilatancy, etc. It's all a matter of modelling methodology and wise compromize. The best model is useless if it can't be computed...

On 27/03/12 13:35, Chareyre wrote:
> Question #191800 on Yade changed:
> https://answers.launchpad.net/yade/+question/191800
>
> Chareyre proposed the following answer:
> Of course, stiffness is a physical parameter and changing it may affect
> (or not...) the results. Ultimately the number of particle is the only
> parameter you can play with.
I agree. I would also suggest to wangxiaoliang to always check that the
results are not affected by the change in stiffness or any other
physical parameters.
> However, if wangxiaoliang is targeting a final strain of 0.3, he is probably not focusing on the elastic regime.
> If the confining pressure is small compared to the stiffness of silicates (10e11), stiffness can most likely be decreased by a few orders of magnitude without any change in peak strength, residual strength, dilatancy, etc. It's all a matter of modelling methodology and wise compromize. The best model is useless if it can't be computed...
Correct... But then wangxiaoliang (btw, is this your name?) should
reconsider the use of Hertzian law. The sensitivity with this law is
high to the stiffness level, in terms of strains (small and large),
dilatancy, peak strength etc. If you are interested, there is some nice
published work by C. Thornton where the behaviour of soft and hard
spheres (Hertzian type) is compared and discussed (700MPa versus 70GPa
of Young's modulus).

Cheers,
Chiara

Yes, please, give Thornton's reference. :)
Modulus should always be compared to confining pressure, else it can't
be called high or low.

wangxiaoliang (wangxiaoliang) said : #8

thanks all. First of all, wangxiaoliang is my Chinese name (Chinese style name) In English version( Xiaoliang Wang)

as the number of particles, now I just have about 1100 in my triaxial test engine.

Increasing mass density is useless to my ques.

that you said decreasing the stiffness maybe a way.

I will have a test first

On 28/03/2012 04:05, wangxiaoliang wrote:
> Question #191800 on Yade changed:
> https://answers.launchpad.net/yade/+question/191800
>
> Status: Answered => Open
>
> wangxiaoliang is still having a problem:
> thanks all. First of all, wangxiaoliang is my Chinese name (Chinese
> style name) In English version( Xiaoliang Wang)
>
> as the number of particles, now I just have about 1100 in my triaxial
> test engine.
>
> Increasing mass density is useless to my ques.
It is not useless if your question was how to speed up the calculation
(obviously you need to test it). This is not only my opinion, but a
method that is quite often applied in the literature (e.g. Thornton, 2000).

Chiara
> that you said decreasing the stiffness maybe a way.
>
> I will have a test first
>

Can you help with this problem?

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

To post a message you must log in.