Time scale issues
I'm using the DesignSPHysics FreeCAD plugin to help with the case pre-processing.
The problem that I am trying to simulate is looking at hypervelocity (~25,000 km/h) impacts in space (like space, space - as in space debris hitting satellites and stuff).
I am trying to simulate 150 microseconds worth of time (150e-6 seconds) with a part time of 1.2 microseconds (1.2e-6 seconds).
When I enter this information in, and also set the inital and minimum timestep to be 1e-12 seconds (for both, respectively), the simulation stops nearly immediately and it thinks that it is complete.
However, if I scale everything up to "seconds" (i.e. 1 second = 1 microsecond), then everything runs ok.
Why is that?
What is the minimum timestep limit that's been hard coded into the programme that is preventing me to run with a timestep of 1e-12 seconds?
Your help is greatly appreciated.
Thank you.
Comments
With these extremely small values wouldn't you perhaps risk seeing floating point errors too? Perhaps that could be part of the issue.
And for my own curiosity, do you still use the typical equation of state given in DualSPHysics? Can you guarantee that it holds true in this enviroment?
Kind regards
In regards to your question about floating point errors - I am not sure, but I would think not if DualSPHysics (whether running it on CPU or GPU) is abiding by the IEEE754 floating point number format.
My thought process is that floating point numbers are represented by the mantissa and the exponent, and therefore; the exponent should be sufficient to cover these small, positive floating point numbers.
If in doubt or if I am worried, I can try switching over to using double precision instead to see if that helps.
re: EOS
My hypothesis (or current assumption) right now is yes because it is basically an aluminum ball (the size of ball bearing) impacting another chunk of aluminum.
My understanding is that DualSPHysics doesn't/won't take the localized frictional heating of the aluminum/aluminum contact surface, but my thought and understanding (for example, with ballistic projectiles) is that the impact energy is perhaps, the dominating factor moreso than the localized heating which is what causes material failure.
I can be wrong, but given the relatively low ambient, absolute temperatures of space, my thought would be that unlike ballistics here on Earth, the impacting projectile would be really cold, hitting something else that is going to be really cold, so even if there was rapid heat generation, the thermal conduction and dissipation should be really fast due to the significant thermal gradient that will arise as a result of the friction.
For now, I am going to assume that the EOS assumption (which looks like it's a Gruneisen model(?)) is applicable and correct until or unless the results show a behaviour that seems to be contrary to the actual, physical findings.
(One of my other team members can do a little bit more research/literature review to see whether the Gruneisen EOS model is still applicable at such low temperatures.)