This can be hugely tricky business, particularly if you're
going
generally non-linear. There are a number of convergence parameters
you'll need to juggle just to get the arithmetic to work out. You
can't make a rigid assumption because that implies an infinite
force
acting over zero time. Consequently to make the problem solve you
need to choose time intervals and contact stiffnesses properly.
There
are rules of thumb for it and probably some built-in defaults, but
it
usually takes some playing around. There's the issue of energy
dissipation. In real impacts a fair amount of energy is dissipated
in
stress waves, but this mechanism is usually impractical to model.
You
can use damping to dissipate energy, but Rayleigh damping (the only
thing you've got in time-domain analysis) doesn't work very well
and
has to be tailored to each run. If you neglect the energy
dissipation
the results are hugely conservative. In particular you end up with
all sorts of unrealistic rebounding. You also need to set the
integration time step and the mesh size to balance modeling
practicalities against the reality of an impact.
You might want to give some thought to simulating the collision
loading with a shock spectrum and do the problem with a response
spectrum analysis.
use random vibration analysis for this problem.
You can set up a dynamic analysis in which the explosion is simulated by a pulse force with very short acting time. The issue is calculating the value of the explosion force and the location.
This can be hugely tricky business, particularly if you're going