yade-dev team mailing list archive

[Bruno Chareyre <bruno.chareyre@xxxxxxxxxxxxxxx>]

Hi Raphaël, in my case I get dt=0 strictly...
As previously suspected returning unconditionally at line 58 restores 
previous behaviour.
I updated trunk [1]. I don't feel completely safe with current logic but 
at least it will be the same as before.
Cheers
Bruno

[1] 
https://github.com/yade/trunk/commit/ec68f4513bbb77f2cc1aef2e1e713cf1d71036fb

On 11/6/18 6:13 PM, Raphael Maurin wrote:
> Hi Bruno,
>
> Sorry for the long delay. I am not able to understand how the small 
> change you made affect so much the results.
>
> I believe that you can reproduce the problem by running the example 
> "trunk/examples/HydroForceEngine/twoWayCoupling/sedimentTransportExample_1DRANSCoupling.py" 
> on your computer. If GSTS option viscEl =True (as set by default in 
> this script) GSTS gives a time step of ~10^-310. If you put viscEl = 
> False, then you obtain a more normal a time step of ~10^-5. (The 
> script run well with viscEl = False as the damping is not fundamental 
> in this situation where the restitution coefficient is not too small).
>
> Tell me if you can reproduce that, and if you see any possible reason 
> for such a big change.
>
> Cheers,
>
> Raphael
>
>
>
> Le 17/10/2018 à 18:42, Bruno Chareyre a écrit :
> > Hi Raphaël,
> > Let's use yade-dev for this sort of things.
> > Sorry if I broke something, I was actually about to write to you when 
> > commiting.
> > The check-script is apparently not using timeStepper since "yade 
> > --check" still gives the expected result:
> >
> > /###################################//
> > //running:  checkViscElPM.py//
> > //Status: success//
> > /
> > One thing to clarify before I can suggest something is this: before 
> > my last commit [1] a body with null stiffness (no contact) would not 
> > be asigned a timestep at all, while I can imagine that some particles 
> > have viscous interactions with neighbors even if they don't touch. No?
> > In last commit I thus inserted something to let timestep be computed 
> > with just viscosity part, it leads to line 58 being now conditional:
> > /if (not viscEl) return;/
> >
> > You can remove the "if" to restore previous behaviour. Most of the 
> > other changes I made are only effective if densityScaling=true, which 
> > is not your case I guess. So this "return" is one of very changes 
> > that could cause the regression.
> >
> > Cheers
> >
> > Bruno
> >
> > [1] 
> > https://github.com/yade/trunk/commit/85ae0ab8fea0be21e7f7e58efadc7120552c6f48#diff-d16884b57bf77bd94e7c5043ee35fe82
> >
> >
> >
> > On 10/17/2018 05:52 PM, Raphaël Maurin wrote:
> > > New question #675297 on Yade:
> > > https://answers.launchpad.net/yade/+question/675297
> > >
> > > Hi all,
> > >
> > > I just updated the code from source, and a problem just appeared in GlobalStiffnessTimeStepper, which was working fine before. When considering the example script trunk/examples/HydroForceEngine/twoWayCoupling/sedimentTransportExample_1DRANSCoupling.py, I get a time step of the order of 10^{-310} (!).
> > > This is the case when considering the option viscEl = True, while everything is fine if viscEl = False.
> > >
> > > This seems to come from a recent commit by Bruno (2 days ago), but I am not able to find from where the change come from. Can anyone (Bruno ? ) help me to identify the problem ?
> > >
> > > Thanks !
> > >
> > > Raphael