POLYDATA - disable prediction on velocity field (contact problem)
When defining a simulation involving contact, POLYDATA informs the user that the prediction on the velocity field is disabled. What does it mean actually?
For solving a transient simulation, a predictor-corrector time marching scheme is used. In such a scheme, the predictor is a first guess of the solution at the next time step (n+1), and is usually obtained by extrapolating the unknown fields from past time; the corrector is the actual solution at that step (n+1) resulting from the non-linear solver. The deviation between predictor and corrector is a measure of the accuracy, and is used for selecting subsequent time steps.
In simulation involving contact, this procedure cannot be applied for the velocity field. Indeed, at a contact occurrence, the velocity is suddenly assigned another value controlled by the contact parameters: e.g. it jumps down to zero at the contact occurrence with a fixed mould. In such a circumstance, the predictor-corrector scheme fails, since the velocity would apparently exhibit a high inaccuracy. In order to circumvent this feature, the velocity is simply removed from the list of fields used for the accuracy evaluation: the predictor is not evaluated (or the prediction on the velocity field is disabled). In most contact simulation, the prediction of the velocity field is disabled (as explained above). The accuracy of the transient scheme and the convergence of the iterative solver are evaluated on the basis of the other fields. Sometimes, it can be interesting to include the velocity field for evaluating the convergence of the solver, while stil discarding it for the accuracy. This can easily be done through the use of the keyword CONVELOC to be specified as a POLYDATA option in POLYMAN or to add into the .p3rc file for a POLYDATA session. In the latter case, POLYDATA should then be invoked with the argument -s .p3rc .