Time step control (5.0 mphase)

I'm trying to make a large-scale 2D simulation of sea ice. Seems I have a problem with time step control. I want the simulation to run for days, with dp something like 500 meters. I'm not interested in things like "sound", so there's no need to tune up the dt for such phenomena.

I have tried setting both DtFixed and DtMin to a large value, but this doesn't seem to have any effect. DualSPHysics seems to do a great number of shorter timesteps regardless. I've had some success setting CoefDtMin to something like 100 000, but this feels like a wrong way to do it.

So, basically I want a simulation of 86400 seconds to run quickly, and stay stable - but for the latter part, I've got some ideas. It's the 1st part that's problematic.

Damn. I'm using 5.0 mphase, sorry for accidentally putting this to the wrong category.


  • Hi!

    Have you tried to tune the speed of sound any way? Remember that the time step control in DualSPHysics depends on the actual value of c, the speed of sound.

    Kind regards

  • Hi!

    I did try to tune the speed of sound without success, but I forget what exactly it was that I did.

    Actually, I remember doing lots and lots of changes to the XML, trying "everything", and eventually DualSPHysics started working. I am not sure, what was the difference. Although I suspect that there was something weird going on with XML comments. I had duplicated several lines and commented out the one with default value, with <!-- -->. I suspect removing these comments was somehow significant in making it work - but it was all a rather non-deterministic affair.

    Thank you though, best regards!

Sign In or Register to comment.