PRED
PRED, Sskey, --, Lskey
Activates a predictor in a nonlinear analysis.
SskeySubstep predictor key:
AUTO
—
The program uses a predictor but, within certain exceptions, automatically switches prediction off. This behavior is the default; see "Command Default" for details.
OFF
—
No prediction occurs.
LINEAR (or ON)
—
Use the linear predictor on all substeps after the first.
QUADRATIC
—
Use the quadratic predictor on all substeps after the second.
--Unused field.
LskeyLoad step predictor:
OFF
—
No prediction across load steps occurs. This is the default behavior.
ON
—
Use a predictor also on the first substep of the load step. (
Sskey= ON is required.)
Command Default
The default command behavior is to use prediction
(Sskey = AUTO). The AUTO option chooses to either use the
linear predictor or to turn the predictor OFF. However, prediction does not occur if
one or more of these conditions exist:
Over prediction occurs due to a large residual force or excessive element distortion.
You are mapping (MAPSOLVE) variables to a new mesh during rezoning. (Prediction does not occur for any MAPSOLVE substeps, nor for the first substep afterwards.)
You have steady-state analysis defined (SSTATE), and contact elements exist in the model.
Notes
Activates a predictor in a nonlinear analysis on the degree-of-freedom solution for the first equilibrium iteration of each substep.
When using the arc-length method (ARCLEN, ARCTRM), you cannot issue the DOF solution predictor command (PRED), the automatic time stepping command (AUTOTS), or the line search command (LNSRCH). If you activate the arc-length method after you set PRED, AUTOTS, or LNSRCH, a warning message appears. If you elect to proceed with the arc-length method, the program disables your DOF predictor, automatic time stepping, and line search settings, and the time step size is controlled by the arc-length method internally.
When using step-applied loads, such as TUNIF, BFUNIF, etc., or other types of non-monotonic loads, the predictor may adversely affect the convergence. If the solution is discontinuous, the predictor may need to be turned off.
When performing a nonlinear analysis involving large rotations, the predictor may require using smaller substeps. If the model has rotational degrees-of-freedom, the quadratic predictor could work more efficiently than the linear predictor.
This command is also valid in PREP7.