The predictor attempt to accelerate convergence
by predicting the solution for the first equilibrium iteration of
every substep. The predictor will extrapolate the results of the last
substep to obtain a starting point for the next solution.
If the nonlinear response is smooth (and the
time step sizes are reasonably small) the predictor can accelerate
convergence.
If the nonlinear response is not smooth, or
large rotations are incorporated in the analysis the predictor can
cause divergence.
Do not use the predictor for a large rotation
analysis.
The default with Solution Control turns off the
predictor if there are rotational degrees of freedom in the model, or
if the current time step is reduced by the automatic time stepping
algorithm.
