best way to check, is there a reasonable way to restrain the problem? overall i prefer to avoid using soft springs and inertial relief unless i back it up with a known solution or that trends match my expectations
Well, there was a factor of sqrt(2) error in my hand calculation, so the inertial relief solution was off by about that much. When I put the right most surface on a slider and took off inertial relief, the solution popped right in within a percent.
I'm a little disappointed, though. Here it makes sense to fix that surface, because it is a separate part which is almost rigid. But frequently there is no rigid surface to fix and I wind up with big phony stress concentrations around the fixtures.
A lot of the vessels we design tend to crumple like oil cans. Any substantial fixture in the analysis really interferes with that. And insubstantial fixtures are not enough to stop the large unbalanced loads (like the one in my original post) from throwing the assembly into large displacement mode.
I had hoped that inertial relief or soft springs was going to be the solution, but they don't seem to work very well. Soft spring pretty much crash every time I turn them on. And inertial relief seems pretty unreliable.
might be one worth throwing to your reseller to see if maybe inertial relief isn't working as expected?
before doing that, maybe just prove it out with an exactly symmetic model to make sure it works?
It depends on how the inertial relief in SW is being carried out. For example, Creo Simulate (formerly Pro/Mechanica) uses a new coordinate system and soft springs in a 3-2-1 constraint setup. This doesn't effect the stress results, but does effect the displacement results. In MSC Nastran, I believe they calculate the rigid body acceleration and apply a counter-acceleration, then does a displacement enforcement to 0 (but don't take my word for it).
At the end of the day, regardless of the method, your stress results are typically good, but your displacement results are usually questionable.
According to the knowledge base, the soft springs option is a spring attached to every node. Inertial relief is a body force. To me, it sounds as if these should give exactly the same result, though with different rigid body displacement. But the KB seems to think the soft springs give more realistic results. That doesn't help me, though, because checking the soft springs box is the same as checking the box that says just crash without attempting to solve this problem.
My results with inertial relief don't seem to be good. It seems to me they are underestimating the stress. Though there are more tests I can do.
mike on the 2d soft springs crash, have you tried a pie slice or similar? it could just be the 2d that is causing it. i don't think i've seen someone try 2d with soft springs, i'm not sure it is supported.
This is a good point.
So, here are some experiment results based on a highly simplified geometry. Basically, everything works fine except 2d with inertial relief and 3d with soft springs. I had long given up on soft springs, but they worked pretty good in the axisymmetric 2d.
The unbalanced load warning that I kept associating with the z-axis reaction listed in the results was actually in the x-axis (the ignored dimension in the 2d simulation). So even though SW will let you turn on inertial relief in 2d, there seems to be a calculation problem.
Closed end tube
External pressure: 10 ksi
Max stress at inside radius:
Roarke thin wall: 62 ksi
Roarke thick wall: 75 ksi
3d with fixed external ring at end cap: 64 ksi
3d with inertial relief: 64 ksi
3d with soft springs: crash
2d with fixed external ring at end cap: 65 ksi
2d with inertial relief: 34 ksi (warning large unbalanced load in x axis i.e., normal to screen)
2d with soft springs: 65 ksi (warning large displacement, though expected)