← Back to team overview

yade-dev team mailing list archive

Re: Questions about Leap-Frog position integrator.

 

Jerome Duriez a écrit :
And how are velocities (in both schemes) computed ?

If they are computed by the same mean, I can still not understand at all what is the difference between these two shemes, except the fact that "2nd order" sounds more serious than "1st order"...

Héhé...

In first order scheme, you would use current velocity for p(t+dt) = p(t) + v(t)*dt, THEN you would compute v(t+dt)=..., in 2nd order, you use the updated (second) value instead. A while ago, Yade was using the 1st order scheme for rotations, and it was the reason why we needed so small time steps.

It seems many people have problems with this simple 2nd order explicit scheme. I send the equations in the attached file. I hope (should I?) it will stop endless discussions in my office about this. ;)

You will find the exact same scheme in any Cundall's paper, even if he doesn't explain it exactly the same way perhaps.

Bruno






Jerome

Bruno Chareyre a écrit :
The answer is (just for the record, I discussed that with Vincent) :

In a 2nd order finite difference scheme for Newtons law, positions and forces are defined at times t+Ndt, while velocities are defined at times t+(N+1/2)dt. If everything was defined at times Ndt, it would be a first order sheme.

v(t+0.5*dt) = v(t-0.5*dt) + a(t)*dt
p(t+dt) = p(t) + v(t+0.5*dt)*dt

Bruno



Vincent Richefeu a écrit :
(I re-send this mail because it didn't appear in yade-dev list... sorry if you receive it for a second time)


Hi everybody,

This morning I had a look at the file LeapFrogPositionIntegrator.cpp and I a little scarred of what I saw (I exaggerate of course).
Can someone enlighten me on that point:
If I translate the c++ code into human readable operations, I've got:

v(t+dt) = v(t) + a(t)*dt
p(t+dt) = p(t) + v(t+dt)*dt + getMove

1. The first thing that seems odd to me is the use of v(t+dt) in the evaluation of p(t+dt). There's no great danger if the displacements are small and slow. However, it is know that even a quasistatic load may be accompanied by small localized dynamic crises. If there is indeed a problem here, no big deal because reducing the time step is enough to correct it. But it's a pity!

2. The second that I don't understand is the method getMove that seems to return a force (!) I know, I am probably wrong. Can anyone give me details on what this function returns ?

Thank you in advance for your helpful comments,
VR

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




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



--

_______________
Chareyre Bruno
Maître de Conférences

Grenoble INP
Laboratoire 3SR - bureau E145
BP 53 - 38041, Grenoble cedex 9 - France
Tél : 33 4 56 52 86 21
Fax : 33 4 76 82 70 43
________________

Attachment: note.pdf
Description: Adobe PDF document


References