Convergence difficulty due to contact separation:
- Use *CONTACT_AUTO_MOVE to remove the initial gap in the contact surface.
- Relax the convergence tolerance until things 'snug up'.
- Use a non-automatic contact which, by default, exhibits 'sticky' behavior in an implicit analysis. This sticky behavior serves to promote convergence.
- Turn on dynamic analysis via *CONTROL_IMPLICIT_DYNAMICS. The resulting inertia will help with convergence.
- Turn on automatic implicit/explicit switching to get over the convergence hump (IMFLAG=4 in *CONTROL_IMPLICIT_GENERAL)
Some ideas for overcoming initial rigid body modes due to contact separation. Two other techniques:
- Use *BOUNDARY_PRESCRIBED_MOTION to bring the parts together, and define it's DEATH time to remove the prescribed motion once contact is established. Then let your force/pressure loads take over. It might take some engineering judgement to decide how far to move the part(s) before the death time.
- Begin the analysis as explicit, then switch to implicit once the gaps are closed. Do this by entering IMFLAG=(-n) on *CONTROL_IMPLICIT_GENERAL, where n is a curve ID. Start the curve with value 0.0 (explicit), hold at 0.0 until the desired switching time, then ramp it sharply up to 1.0 (implicit) and hold at 1.0 for the remainder of the simulation. It may also take some engineering judgement with this method to decide how long to run explicitly before switching. If you run explicitly for too long, you'll build up too much inertia in the model and implicit will have a tough time recovering static equilibrium.
An excellent diagnostic tool is an eigenvalue analysis. Just add these two lines to your implicit model:
This will instruct LS-DYNA to compute the lowest 30 eigenvalues, then stop. View the text output file "EIGOUT", and check for eigenvalues which are nearly zero. Use LS-Prepost to view and animate the "D3EIGV" binary plot database to see the corresponding mode shapes. Hopefully this will show up a missing attachment between two parts, or a part which is unconstrained and causing problems.
A short list of recommendations for implicit...
- Use formulation 16 or 6 for shells
- If a static analysis, ramp the prescribed force/displacement in a linear fashion
- Turn on automatic timestep control (see *CONTROL_IMPLICIT_AUTO)
- Use non-automatic contact types