timestep-bug

[Christian Jakob]

Timesteps increase potentially until the maximum of double is reached.
The bug can be related to makeCloud and/or GlobalStiffnessTimeStepper.
See attachment for further details. After generating the particle cloud click on "Play" button to see the effect.

[Anton Gladky (gladky-anton)]

Christian,

try to put your "engine" section _after_ creating bodies.

Anton

[Christian Jakob (jakob-ifgt)]

Thanks Anton, problem solved.

[Bruno Chareyre (bruno-chareyre)]

Precisely, it is only that part that must be after bodies creations: "defaultDt=0.3*utils.PWaveTimeStep()", since the PWave speed of nothing is... nothing!
And note that "utils." is not needed in the most common functions. PWaveTimeStep() alone works as well.

[Christian Jakob (jakob-ifgt)]

Another problem occurs, when I remove numbers of particles (num=-1).
example:

sp=pack.SpherePack();
sp.makeCloud((x_cl,y_cl,z_cl),(x_cu,y_cu,z_cu),psdSizes=dia,psdCumm=phi_,distributeMass=True)

Then only particles with the minimum particle size (dia[0]) are generated (after >1 hour generation time when using the script from above).

Is this normal, or a bug?

[Anton Gladky (gladky-anton)]

Anyway, I think it is a bug.
Because PWaveTimeStep should not return "inf"-value.

It was fixed in r2979. PWaveTimeStep gives a warning now and sets dt=1.0

Christian, please, close this bug.

"numbers of particles" is perhaps another issue.

Anton

[Bruno Chareyre (bruno-chareyre)]

>Then only particles with the minimum particle size (dia[0]) are generated

There are so many parameters combinations in makeCloud that I overlooked one. How did you manage to hit the one that was not expected? ;)
2D/3D * periodic * num * poro * psdSizes * distributeMass * hSize = 2^7 = 128 possible usage (ok, some of them are forbidden).

It is fixed in r2980, thanks for notifying. Note that less particles will be generated, this is expected behaviour (when num is defined, we can generate the big ones first, else it is 100% stochastic and it is more difficult to place big spheres).

[Christian Jakob (jakob-ifgt)]

> There are so many parameters combinations in makeCloud that I overlooked one.
> How did you manage to hit the one that was not expected? ;)

There are so many users of Yade. Why is it me who is trapping in all bugs?

I was wondering, because the one who wrote .../scripts/test/psd.py should have noticed that there is a bug.

Thanks for your help.

[Bruno Chareyre (bruno-chareyre)]

>the one who wrote .../scripts/test/psd.py

It used to work, it's been broken after the initial script.
Actually, the normalization I included in r2980 was not at the appropriate line, I moved it down and now (r2981) everything looks correct from psd.py: only the 20k spheres PSD is scaled down, the one with 5k is unchanged.

p.s. You are exploiting the code more than the average user (e.g. coupling hertz and capillary forces) and you write scripts from scratch, so you hit more bugs, sorry for that. It helps us to push the limits of Yade further.