I have modelled an experiment consisted of a specimen under a four point bending test, using LS-DYNA (FEM Software). At the keyword *CONTROL_IMPLICIT_GENERAL I have defined that an implicit solution should be followed. I have noticed that by changing the number of time steps, i.e. by changing the option DTMAX at *CONTROL_IMPLICIT_AUTO, I get different results. My termination time is 10 seconds and by defining DTMAX=0.2, DTMAX=0.1 and DTMAX=0.02 I get 50, 100 and 500 time steps respectively. I attach a plot of the results where the difference is obvious. Why are the results different?
The accuracy of results are actually sensitive to the time steps you are using in implicit methods.When you use an implicit method with larger time step, you may get a stable solution, but the accuracy will be reduced and there will be some artificial damping present.
Thank you for your reply. Could you please elaborate a little bit more, why the accuracy is reduced? How is this artificial damping being introduced and how can I control it? I would appreciate an explanation or some links where I can search more into depth.
Did you try a tighter convergence criterion? If you still get different results then you must check if the implicit method is applied only globally or also at the integration point level.
In my humble opinion, the time step shall not have an influence on the solution as far as the solution converges within the specified tolerance. I assume that there is no time dependency in the constitutive behavior.
Otherwise, I use explicit method with a quasi static condition. Have a look http://www.dynalook.com/international-conf-2006/34SimulationTechnology.pdf
- Badges
- Science topic