1
$\begingroup$

In a given simulation in MSC Adams, if I use a 1 or 0.1 step increment, I get the (favorable) result I intuitively assumed would occur, but if I use step increments of 0.01 or 0.001 a different (and unfavorable) result occurs. One would assume the finer-detailed result would generate fewer simulation false-result-inducing artifacts. Should I always assume the more granular simulation will result the more realistic and reliable result?

$\endgroup$
1
  • 1
    $\begingroup$ Depends on what you simulate, if you need to simulate contact then you may need more steps as even the adaptive nature of the solver wont help. $\endgroup$
    – joojaa
    Oct 11 '20 at 20:28
2
$\begingroup$

This is a common issue on all numerical methods (not just multibody problems), i.e. the tradeoff between the following three types of error:

  1. Round off error
  2. Discretization
  3. Truncation (sometimes the 2 and 3 are included in the same term).

All of them are forms of quantization error.

These three types have a dependence on mesh size (either spatial or temporal). You can see their dependence, in the following image.

enter image description here

As you reduce the time step (or mesh size) the discretization error decreases, however then what happens is that you get the round off error to increase. Usually, it's the balance between those two that dominates the total error.

$\endgroup$
3
  • $\begingroup$ There is no mesh in a multibody system, but the description is more or less the same. $\endgroup$
    – joojaa
    Oct 11 '20 at 20:27
  • $\begingroup$ After running new series, I found the 0.1 step and the 0.001 and 0.0001 (yes weeks running) step runs agreed to the very same result (the 0.1 just giving a chunkier curve). I thus assume the disagreeing 0.01 result was an artifact and may safely discard this result, would you agree? $\endgroup$ Nov 19 '20 at 13:24
  • $\begingroup$ The fact that you have [Good, bad, good] results doesn't give me much confidence. Have you tried using 0.02 or 0.12 or 0.09 to check whether that is indeed some strange combination of parameters? $\endgroup$
    – NMech
    Nov 19 '20 at 14:13

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.